优化消息队列服务持久化

优化rankservice错误返回
优化性能监控与rankservice持久化
2026-02-06 16:14:45 +08:00 · 2023-05-17 11:26:29 +08:00 · 2023-05-09 14:34:33 +08:00 · 2023-05-09 14:06:17 +08:00 · 2023-05-04 17:53:42 +08:00 · 2023-05-04 17:35:40 +08:00
43 changed files with 5388 additions and 918 deletions
--- a/README.md
+++ b/README.md
@@ -1,10 +1,10 @@
 origin 游戏服务器引擎简介
-==================
-
+=========================

 origin 是一个由 Go 语言（golang）编写的分布式开源游戏服务器引擎。origin适用于各类游戏服务器的开发，包括 H5（HTML5）游戏服务器。

 origin 解决的问题：
+
 * origin总体设计如go语言设计一样，总是尽可能的提供简洁和易用的模式，快速开发。
 * 能够根据业务需求快速并灵活的制定服务器架构。
 * 利用多核优势，将不同的service配置到不同的node，并能高效的协同工作。
@@ -12,12 +12,16 @@ origin 解决的问题：
 * 有丰富并健壮的工具库。

 Hello world!
---------------
+------------
+
 下面我们来一步步的建立origin服务器,先下载[origin引擎](https://github.com/duanhf2012/origin "origin引擎"),或者使用如下命令：
+
 ```go
 go get -v -u  github.com/duanhf2012/origin
 ```
+
 于是下载到GOPATH环境目录中,在src中加入main.go,内容如下：
+
 ```go
 package main

@@ -29,16 +33,20 @@ func main() {
 	node.Start()
 }
 ```
+
 以上只是基础代码，具体运行参数和配置请参照第一章节。

 一个origin进程需要创建一个node对象,Start开始运行。您也可以直接下载origin引擎示例:
+
 ```
 go get -v -u github.com/duanhf2012/originserver
 ```
+
 本文所有的说明都是基于该示例为主。

 origin引擎三大对象关系
---------------
+----------------------
+
 * Node:   可以认为每一个Node代表着一个origin进程
 * Service:一个独立的服务可以认为是一个大的功能模块，他是Node的子集，创建完成并安装Node对象中。服务可以支持对外部RPC等功能。
 * Module: 这是origin最小对象单元，强烈建议所有的业务模块都划分成各个小的Module组合，origin引擎将监控所有服务与Module运行状态，例如可以监控它们的慢处理和死循环函数。Module可以建立树状关系。Service本身也是Module的类型。
@@ -46,7 +54,8 @@ origin引擎三大对象关系
 origin集群核心配置文件在config的cluster目录下，如github.com/duanhf2012/originserver的config/cluster目录下有cluster.json与service.json配置：

 cluster.json如下：
---------------
+------------------
+
 ```
 {
    "NodeList":[
@@ -70,21 +79,26 @@ cluster.json如下：
        }
    ]
 ```
---------------
+
+---
+
 以上配置了两个结点服务器程序:
+
 * NodeId: 表示origin程序的结点Id标识，不允许重复。
 * Private: 是否私有结点，如果为true，表示其他结点不会发现它，但可以自我运行。
 * ListenAddr:Rpc通信服务的监听地址
 * MaxRpcParamLen:Rpc参数数据包最大长度，该参数可以缺省，默认一次Rpc调用支持最大4294967295byte长度数据。
 * NodeName:结点名称
 * remark:备注，可选项
-* ServiceList:该Node将安装的服务列表
---------------
+* ServiceList:该Node拥有的服务列表，注意：origin按配置的顺序进行安装初始化。但停止服务的顺序是相反。
+
+---

 在启动程序命令originserver -start nodeid=1中nodeid就是根据该配置装载服务。
 更多参数使用，请使用originserver -help查看。
 service.json如下：
---------------
+------------------
+
 ```
 {
 "Global": {
@@ -103,7 +117,7 @@ service.json如下：
 			"Keyfile":""
 		}
 		]
-		
+	
 	  },
 	  "TcpService":{
 		"ListenAddr":"0.0.0.0:9030",
@@ -160,10 +174,12 @@ service.json如下：
 }
 ```

---------------
+---
+
 以上配置分为两个部分：Global,Service与NodeService。Global是全局配置，在任何服务中都可以通过cluster.GetCluster().GetGlobalCfg()获取，NodeService中配置的对应结点中服务的配置，如果启动程序中根据nodeid查找该域的对应的服务，如果找不到时，从Service公共部分查找。

 **HttpService配置**
+
 * ListenAddr:Http监听地址
 * ReadTimeout:读网络超时毫秒
 * WriteTimeout:写网络超时毫秒
@@ -172,6 +188,7 @@ service.json如下：
 * CAFile: 证书文件，如果您的服务器通过web服务器代理配置https可以忽略该配置

 **TcpService配置**
+
 * ListenAddr: 监听地址
 * MaxConnNum: 允许最大连接数
 * PendingWriteNum：发送网络队列最大数量
@@ -180,20 +197,21 @@ service.json如下：
 * MaxMsgLen:包最大长度

 **WSService配置**
+
 * ListenAddr: 监听地址
 * MaxConnNum: 允许最大连接数
 * PendingWriteNum：发送网络队列最大数量
 * MaxMsgLen:包最大长度
---------------
-
-

+---

 第一章：origin基础:
---------------
+-------------------
+
 查看github.com/duanhf2012/originserver中的simple_service中新建两个服务，分别是TestService1.go与CTestService2.go。

 simple_service/TestService1.go如下：
+
 ```
 package simple_service

@@ -223,7 +241,9 @@ func (slf *TestService1) OnInit() error {


 ```
+
 simple_service/TestService2.go如下：
+
 ```
 import (
 	"github.com/duanhf2012/origin/node"
@@ -263,6 +283,7 @@ func main(){
 ```

 * config/cluster/cluster.json如下：
+
 ```
 {
    "NodeList":[
@@ -279,6 +300,7 @@ func main(){
 ```

 编译后运行结果如下：
+
 ```
 #originserver -start nodeid=1
 TestService1 OnInit.
@@ -286,13 +308,15 @@ TestService2 OnInit.
 ```

 第二章：Service中常用功能:
---------------
+--------------------------

 定时器:
---------------
+-------
+
 在开发中最常用的功能有定时任务，origin提供两种定时方式：

 一种AfterFunc函数，可以间隔一定时间触发回调，参照simple_service/TestService2.go,实现如下：
+
 ```
 func (slf *TestService2) OnInit() error {
 	fmt.Printf("TestService2 OnInit.\n")
@@ -305,10 +329,11 @@ func (slf *TestService2) OnSecondTick(){
 	slf.AfterFunc(time.Second*1,slf.OnSecondTick)
 }
 ```
+
 此时日志可以看到每隔1秒钟会print一次"tick."，如果下次还需要触发，需要重新设置定时器

-
 另一种方式是类似Linux系统的crontab命令，使用如下：
+
 ```

 func (slf *TestService2) OnInit() error {
@@ -327,27 +352,29 @@ func (slf *TestService2) OnCron(cron *timer.Cron){
 	fmt.Printf(":A minute passed!\n")
 }
 ```
-以上运行结果每换分钟时打印:A minute passed!

+以上运行结果每换分钟时打印:A minute passed!

 打开多协程模式:
 ---------------
+
 在origin引擎设计中，所有的服务是单协程模式，这样在编写逻辑代码时，不用考虑线程安全问题。极大的减少开发难度，但某些开发场景下不用考虑这个问题，而且需要并发执行的情况，比如，某服务只处理数据库操作控制，而数据库处理中发生阻塞等待的问题，因为一个协程，该服务接受的数据库操作只能是一个
 一个的排队处理，效率过低。于是可以打开此模式指定处理协程数，代码如下：
+
 ```
 func (slf *TestService1) OnInit() error {
 	fmt.Printf("TestService1 OnInit.\n")
-	
+
 	//打开多线程处理模式，10个协程并发处理
 	slf.SetGoRoutineNum(10)
 	return nil
 }
 ```
-为了


 性能监控功能:
---------------
+-------------
+
 我们在开发一个大型的系统时，经常由于一些代码质量的原因，产生处理过慢或者死循环的产生，该功能可以被监测到。使用方法如下：

 ```
@@ -382,6 +409,7 @@ func main(){
 }

 ```
+
 上面通过GetProfiler().SetOverTime与slf.GetProfiler().SetMaxOverTimer设置监控时间
 并在main.go中，打开了性能报告器，以每10秒汇报一次，因为上面的例子中，定时器是有死循环，所以可以得到以下报告：

@@ -390,10 +418,11 @@ process count 0,take time 0 Milliseconds,average 0 Milliseconds/per.
 too slow process:Timer_orginserver/simple_service.(*TestService1).Loop-fm is take 38003 Milliseconds
 直接帮助找到TestService1服务中的Loop函数

-
 结点连接和断开事件监听:
---------------
+-----------------------
+
 在有些业务中需要关注某结点是否断开连接，可以注册回调如下：
+
 ```
 func (ts *TestService) OnInit() error{
 	ts.RegRpcListener(ts)
@@ -408,13 +437,14 @@ func (ts *TestService) OnNodeDisconnect(nodeId int){
 }
 ```

-
 第三章：Module使用:
---------------
+-------------------

 Module创建与销毁:
---------------
+-----------------
+
 可以认为Service就是一种Module，它有Module所有的功能。在示例代码中可以参考originserver/simple_module/TestService3.go。
+
 ```
 package simple_module

@@ -476,7 +506,9 @@ func (slf *TestService3) OnInit() error {
 }

 ```
+
 在OnInit中创建了一条线型的模块关系TestService3->module1->module2，调用AddModule后会返回Module的Id，自动生成的Id从10e17开始,内部的id，您可以自己设置Id。当调用ReleaseModule释放时module1时，同样会将module2释放。会自动调用OnRelease函数，日志顺序如下：
+
 ```
 Module1 OnInit.
 Module2 OnInit.
@@ -484,14 +516,16 @@ module1 id is 100000000000000001, module2 id is 100000000000000002
 Module2 Release.
 Module1 Release.
 ```
+
 在Module中同样可以使用定时器功能，请参照第二章节的定时器部分。

-
 第四章：事件使用
---------------
+----------------
+
 事件是origin中一个重要的组成部分，可以在同一个node中的service与service或者与module之间进行事件通知。系统内置的几个服务，如：TcpService/HttpService等都是通过事件功能实现。他也是一个典型的观察者设计模型。在event中有两个类型的interface，一个是event.IEventProcessor它提供注册与卸载功能，另一个是event.IEventHandler提供消息广播等功能。

 在目录simple_event/TestService4.go中
+
 ```
 package simple_event

@@ -535,6 +569,7 @@ func (slf *TestService4) TriggerEvent(){
 ```

 在目录simple_event/TestService5.go中
+
 ```
 package simple_event

@@ -590,19 +625,24 @@ func (slf *TestService5) OnServiceEvent(ev event.IEvent){


 ```
+
 程序运行10秒后，调用slf.TriggerEvent函数广播事件，于是在TestService5中会收到
+
 ```
 OnServiceEvent type :1001 data:event data.
 OnModuleEvent type :1001 data:event data.
 ```
+
 在上面的TestModule中监听的事情，当这个Module被Release时监听会自动卸载。

 第五章：RPC使用
 ---------------
+
 RPC是service与service间通信的重要方式，它允许跨进程node互相访问，当然也可以指定nodeid进行调用。如下示例：

 simple_rpc/TestService6.go文件如下：
-```
+
+```go
 package simple_rpc

 import (
@@ -627,6 +667,7 @@ type InputData struct {
 	B int
 }

+// 注意RPC函数名的格式必需为RPC_FunctionName或者是RPCFunctionName，如下的RPC_Sum也可以写成RPCSum
 func (slf *TestService6) RPC_Sum(input *InputData,output *int) error{
 	*output = input.A+input.B
 	return nil
@@ -635,6 +676,7 @@ func (slf *TestService6) RPC_Sum(input *InputData,output *int) error{
 ```

 simple_rpc/TestService7.go文件如下：
+
 ```
 package simple_rpc

@@ -709,11 +751,82 @@ func (slf *TestService7) GoTest(){
 }

 ```
+
 您可以把TestService6配置到其他的Node中，比如NodeId为2中。只要在一个子网，origin引擎可以无差别调用。开发者只需要关注Service关系。同样它也是您服务器架构设计的核心需要思考的部分。

-第六章：配置服务发现
+
+第六章：并发函数调用
 ---------------
+在开发中经常会有将某些任务放到其他协程中并发执行，执行完成后，将服务的工作线程去回调。使用方式很简单，先打开该功能如下代码：
+```
+	//以下通过cpu数量来定开启协程并发数量，建议:(1)cpu密集型计算使用1.0  (2)i/o密集型使用2.0或者更高
+	slf.OpenConcurrentByNumCPU(1.0)
+	
+	//以下通过函数打开并发协程数，以下协程数最小5，最大10，任务管道的cap数量1000000
+	//origin会根据任务的数量在最小与最大协程数间动态伸缩
+	//slf.OpenConcurrent(5, 10, 1000000)
+```
+
+使用示例如下:
+```
+
+func (slf *TestService13) testAsyncDo() {
+	var context struct {
+		data int64
+	}
+
+	//1.示例普通使用
+	//参数一的函数在其他协程池中执行完成，将执行完成事件放入服务工作协程，
+	//参数二的函数在服务协程中执行，是协程安全的。
+	slf.AsyncDo(func() bool {
+		//该函数回调在协程池中执行
+		context.data = 100
+		return true
+	}, func(err error) {
+		//函数将在服务协程中执行
+		fmt.Print(context.data) //显示100
+	})
+
+	//2.示例按队列顺序
+	//参数一传入队列Id,同一个队列Id将在协程池中被排队执行
+	//以下进行两次调用，因为两次都传入参数queueId都为1，所以它们会都进入queueId为1的排队执行
+	queueId := int64(1)
+	for i := 0; i < 2; i++ {
+		slf.AsyncDoByQueue(queueId, func() bool {
+			//该函数会被2次调用，但是会排队执行
+			return true
+		}, func(err error) {
+			//函数将在服务协程中执行
+		})
+	}
+
+	//3.函数参数可以某中一个为空
+	//参数二函数将被延迟执行
+	slf.AsyncDo(nil, func(err error) {
+		//将在下
+	})
+
+	//参数一函数在协程池中执行，但没有在服务协程中回调
+	slf.AsyncDo(func() bool {
+		return true
+	}, nil)
+
+	//4.函数返回值控制不进行回调
+	slf.AsyncDo(func() bool {
+		//返回false时，参数二函数将不会被执行; 为true时，则会被执行
+		return false
+	}, func(err error) {
+		//该函数将不会被执行
+	})
+}
+```
+
+
+第七章：配置服务发现
+--------------------
+
 origin引擎默认使用读取所有结点配置的进行确认结点有哪些Service。引擎也支持动态服务发现的方式，使用了内置的DiscoveryMaster服务用于中心Service，DiscoveryClient用于向DiscoveryMaster获取整个origin网络中所有的结点以及服务信息。具体实现细节请查看这两部分的服务实现。具体使用方式，在以下cluster配置中加入以下内容：
+
 ```
 {
 	"MasterDiscoveryNode": [{
@@ -727,8 +840,8 @@ origin引擎默认使用读取所有结点配置的进行确认结点有哪些Se
 		"ListenAddr": "127.0.0.1:8801",
 		"MaxRpcParamLen": 409600
 	}],
-	
-	
+
+
 	"NodeList": [{
 		"NodeId": 1,
 		"ListenAddr": "127.0.0.1:8801",
@@ -741,6 +854,7 @@ origin引擎默认使用读取所有结点配置的进行确认结点有哪些Se
 	}]
 }
 ```
+
 新上有两新不同的字段分别为MasterDiscoveryNode与DiscoveryService。其中:

 MasterDiscoveryNode中配置了结点Id为1的服务发现Master，他的监听地址ListenAddr为127.0.0.1:8801，结点为2的也是一个服务发现Master，不同在于多了"NeighborService":["HttpGateService"]配置。如果"NeighborService"有配置具体的服务时，则表示该结点是一个邻居Master结点。当前运行的Node结点会从该Master结点上筛选HttpGateService的服务，并且当前运行的Node结点不会向上同步本地所有公开的服务，和邻居结点关系是单向的。
@@ -748,14 +862,13 @@ MasterDiscoveryNode中配置了结点Id为1的服务发现Master，他的监听
 NeighborService可以用在当有多个以Master中心结点的网络，发现跨网络的服务场景。
 DiscoveryService表示将筛选origin网络中的TestService8服务，注意如果DiscoveryService不配置，则筛选功能不生效。

+第八章：HttpService使用
+-----------------------

-
-
-第七章：HttpService使用
---------------
 HttpService是origin引擎中系统实现的http服务，http接口中常用的GET,POST以及url路由处理。

 simple_http/TestHttpService.go文件如下：
+
 ```
 package simple_http

@@ -825,15 +938,16 @@ func (slf *TestHttpService) HttpPost(session *sysservice.HttpSession){
 }

 ```
+
 注意，要在main.go中加入import _ "orginserver/simple_service"，并且在config/cluster/cluster.json中的ServiceList加入服务。

+第九章：TcpService服务使用
+--------------------------

-
-第七章：TcpService服务使用
---------------
 TcpService是origin引擎中系统实现的Tcp服务，可以支持自定义消息格式处理器。只要重新实现network.Processor接口。目前内置已经实现最常用的protobuf处理器。

 simple_tcp/TestTcpService.go文件如下：
+
 ```
 package simple_tcp

@@ -901,9 +1015,9 @@ func (slf *TestTcpService) OnRequest (clientid uint64,msg proto.Message){
 }
 ```

+第十章：其他系统模块介绍
+------------------------

-第八章：其他系统模块介绍
---------------
 * sysservice/wsservice.go:支持了WebSocket协议，使用方法与TcpService类似
 * sysmodule/DBModule.go:对mysql数据库操作
 * sysmodule/RedisModule.go:对Redis数据进行操作
@@ -912,9 +1026,9 @@ func (slf *TestTcpService) OnRequest (clientid uint64,msg proto.Message){
 * util:在该目录下，有常用的uuid,hash,md5,协程封装等工具库
 * https://github.com/duanhf2012/originservice: 其他扩展支持的服务可以在该工程上看到，目前支持firebase推送的封装。

-
 备注:
---------------
+-----
+
 **感觉不错请star, 谢谢!**

 **欢迎加入origin服务器开发QQ交流群:168306674，有任何疑问我都会及时解答**
@@ -924,6 +1038,7 @@ func (slf *TestTcpService) OnRequest (clientid uint64,msg proto.Message){
 [因服务器是由个人维护，如果这个项目对您有帮助，您可以点我进行捐赠，感谢！](http://www.cppblog.com/images/cppblog_com/API/21416/r_pay.jpg "Thanks!")

 特别感谢以下赞助网友：
+
 ```
 咕咕兽
 _
--- a/cluster/cluster.go
+++ b/cluster/cluster.go
@@ -26,7 +26,7 @@ type NodeInfo struct {
 	Private           bool
 	ListenAddr        string
 	MaxRpcParamLen    uint32   //最大Rpc参数长度
-	ServiceList       []string //所有的服务列表
+	ServiceList  	  []string //所有的有序服务列表
 	PublicServiceList []string //对外公开的服务列表
 	DiscoveryService  []string //筛选发现的服务，如果不配置，不进行筛选
 	NeighborService   []string
@@ -110,15 +110,13 @@ func (cls *Cluster) DelNode(nodeId int, immediately bool) {
 			break
 		}

-		rpc.client.Lock()
 		//正在连接中不主动断开，只断开没有连接中的
 		if rpc.client.IsConnected() {
 			nodeInfo.status = Discard
-			rpc.client.Unlock()
 			log.SRelease("Discard node ", nodeInfo.NodeId, " ", nodeInfo.ListenAddr)
 			return
 		}
-		rpc.client.Unlock()
+
 		break
 	}

@@ -194,20 +192,17 @@ func (cls *Cluster) serviceDiscoverySetNodeInfo(nodeInfo *NodeInfo) {
 	if _, rpcInfoOK := cls.mapRpc[nodeInfo.NodeId]; rpcInfoOK == true {
 		return
 	}
+
 	rpcInfo := NodeRpcInfo{}
 	rpcInfo.nodeInfo = *nodeInfo
-	rpcInfo.client = &rpc.Client{}
-	rpcInfo.client.TriggerRpcEvent = cls.triggerRpcEvent
-	rpcInfo.client.Connect(nodeInfo.NodeId, nodeInfo.ListenAddr, nodeInfo.MaxRpcParamLen)
+	rpcInfo.client =rpc.NewRClient(nodeInfo.NodeId, nodeInfo.ListenAddr, nodeInfo.MaxRpcParamLen,cls.triggerRpcEvent)
 	cls.mapRpc[nodeInfo.NodeId] = rpcInfo
-
 }

 func (cls *Cluster) buildLocalRpc() {
 	rpcInfo := NodeRpcInfo{}
 	rpcInfo.nodeInfo = cls.localNodeInfo
-	rpcInfo.client = &rpc.Client{}
-	rpcInfo.client.Connect(rpcInfo.nodeInfo.NodeId, "", 0)
+	rpcInfo.client = rpc.NewLClient(rpcInfo.nodeInfo.NodeId)

 	cls.mapRpc[cls.localNodeInfo.NodeId] = rpcInfo
 }
@@ -253,8 +248,9 @@ func (cls *Cluster) checkDynamicDiscovery(localNodeId int) (bool, bool) {
 	return localMaster, hasMaster
 }

-func (cls *Cluster) appendService(serviceName string, bPublicService bool) {
-	cls.localNodeInfo.ServiceList = append(cls.localNodeInfo.ServiceList, serviceName)
+func (cls *Cluster) AddDynamicDiscoveryService(serviceName string, bPublicService bool) {
+	addServiceList := append([]string{},serviceName)
+	cls.localNodeInfo.ServiceList = append(addServiceList,cls.localNodeInfo.ServiceList...)
 	if bPublicService {
 		cls.localNodeInfo.PublicServiceList = append(cls.localNodeInfo.PublicServiceList, serviceName)
 	}
@@ -298,11 +294,10 @@ func (cls *Cluster) SetupServiceDiscovery(localNodeId int, setupServiceFun Setup

 	//2.如果为动态服务发现安装本地发现服务
 	cls.serviceDiscovery = getDynamicDiscovery()
+	cls.AddDynamicDiscoveryService(DynamicDiscoveryClientName, true)
 	if localMaster == true {
-		cls.appendService(DynamicDiscoveryMasterName, false)
+		cls.AddDynamicDiscoveryService(DynamicDiscoveryMasterName, false)
 	}
-	cls.appendService(DynamicDiscoveryClientName, true)
-
 }

 func (cls *Cluster) FindRpcHandler(serviceName string) rpc.IRpcHandler {
@@ -358,10 +353,10 @@ func (cls *Cluster) IsNodeConnected(nodeId int) bool {
 	return pClient != nil && pClient.IsConnected()
 }

-func (cls *Cluster) triggerRpcEvent(bConnect bool, clientSeq uint32, nodeId int) {
+func (cls *Cluster) triggerRpcEvent(bConnect bool, clientId uint32, nodeId int) {
 	cls.locker.Lock()
 	nodeInfo, ok := cls.mapRpc[nodeId]
-	if ok == false || nodeInfo.client == nil || nodeInfo.client.GetClientSeq() != clientSeq {
+	if ok == false || nodeInfo.client == nil || nodeInfo.client.GetClientId() != clientId {
 		cls.locker.Unlock()
 		return
 	}
@@ -383,7 +378,6 @@ func (cls *Cluster) triggerRpcEvent(bConnect bool, clientSeq uint32, nodeId int)
 	}
 }

-
 func (cls *Cluster) TriggerDiscoveryEvent(bDiscovery bool, nodeId int, serviceName []string) {
 	cls.rpcEventLocker.Lock()
 	defer cls.rpcEventLocker.Unlock()
--- a/cluster/dynamicdiscovery.go
+++ b/cluster/dynamicdiscovery.go
@@ -60,6 +60,21 @@ func (ds *DynamicDiscoveryMaster) addNodeInfo(nodeInfo *rpc.NodeInfo) {
 	ds.nodeInfo = append(ds.nodeInfo, nodeInfo)
 }

+func (ds *DynamicDiscoveryMaster) removeNodeInfo(nodeId int32) {
+	if _,ok:= ds.mapNodeInfo[nodeId];ok == false {
+		return
+	}
+
+	for i:=0;i<len(ds.nodeInfo);i++ {
+		if ds.nodeInfo[i].NodeId == nodeId {
+			ds.nodeInfo = append(ds.nodeInfo[:i],ds.nodeInfo[i+1:]...)
+			break
+		}
+	}
+
+	delete(ds.mapNodeInfo,nodeId)
+}
+
 func (ds *DynamicDiscoveryMaster) OnInit() error {
 	ds.mapNodeInfo = make(map[int32]struct{}, 20)
 	ds.RegRpcListener(ds)
@@ -103,6 +118,8 @@ func (ds *DynamicDiscoveryMaster) OnNodeDisconnect(nodeId int) {
 		return
 	}

+	ds.removeNodeInfo(int32(nodeId))
+
 	var notifyDiscover rpc.SubscribeDiscoverNotify
 	notifyDiscover.MasterNodeId = int32(cluster.GetLocalNodeInfo().NodeId)
 	notifyDiscover.DelNodeId = int32(nodeId)
--- a/concurrent/concurrent.go
+++ b/concurrent/concurrent.go
@@ -0,0 +1,93 @@
+package concurrent
+
+import (
+	"errors"
+	"runtime"
+
+	"github.com/duanhf2012/origin/log"
+)
+
+const defaultMaxTaskChannelNum = 1000000
+
+type IConcurrent interface {
+	OpenConcurrentByNumCPU(cpuMul float32)
+	OpenConcurrent(minGoroutineNum int32, maxGoroutineNum int32, maxTaskChannelNum int)
+	AsyncDoByQueue(queueId int64, fn func() bool, cb func(err error))
+	AsyncDo(f func() bool, cb func(err error))
+}
+
+type Concurrent struct {
+	dispatch
+
+	tasks     chan task
+	cbChannel chan func(error)
+}
+
+/*
+cpuMul 表示cpu的倍数
+建议:(1)cpu密集型 使用1  (2)i/o密集型使用2或者更高
+*/
+func (c *Concurrent) OpenConcurrentByNumCPU(cpuNumMul float32) {
+	goroutineNum := int32(float32(runtime.NumCPU())*cpuNumMul + 1)
+	c.OpenConcurrent(goroutineNum, goroutineNum, defaultMaxTaskChannelNum)
+}
+
+func (c *Concurrent) OpenConcurrent(minGoroutineNum int32, maxGoroutineNum int32, maxTaskChannelNum int) {
+	c.tasks = make(chan task, maxTaskChannelNum)
+	c.cbChannel = make(chan func(error), maxTaskChannelNum)
+
+	//打开dispach
+	c.dispatch.open(minGoroutineNum, maxGoroutineNum, c.tasks, c.cbChannel)
+}
+
+func (c *Concurrent) AsyncDo(f func() bool, cb func(err error)) {
+	c.AsyncDoByQueue(0, f, cb)
+}
+
+func (c *Concurrent) AsyncDoByQueue(queueId int64, fn func() bool, cb func(err error)) {
+	if cap(c.tasks) == 0 {
+		panic("not open concurrent")
+	}
+
+	if fn == nil && cb == nil {
+		log.SStack("fn and cb is nil")
+		return
+	}
+
+	if fn == nil {
+		c.pushAsyncDoCallbackEvent(cb)
+		return
+	}
+
+	if queueId != 0 {
+		queueId = queueId % maxTaskQueueSessionId+1
+	}
+
+	select {
+	case c.tasks <- task{queueId, fn, cb}:
+	default:
+		log.SError("tasks channel is full")
+		if cb != nil {
+			c.pushAsyncDoCallbackEvent(func(err error) {
+				cb(errors.New("tasks channel is full"))
+			})
+		}
+		return
+	}
+}
+
+func (c *Concurrent) Close() {
+	if cap(c.tasks) == 0 {
+		return
+	}
+
+	log.SRelease("wait close concurrent")
+
+	c.dispatch.close()
+
+	log.SRelease("concurrent has successfully exited")
+}
+
+func (c *Concurrent) GetCallBackChannel() chan func(error) {
+	return c.cbChannel
+}
--- a/concurrent/dispatch.go
+++ b/concurrent/dispatch.go
@@ -0,0 +1,196 @@
+package concurrent
+
+import (
+	"sync"
+	"sync/atomic"
+	"time"
+
+	"fmt"
+	"runtime"
+
+	"github.com/duanhf2012/origin/log"
+	"github.com/duanhf2012/origin/util/queue"
+)
+
+var idleTimeout = int64(2 * time.Second)
+const maxTaskQueueSessionId = 10000
+
+type dispatch struct {
+	minConcurrentNum int32
+	maxConcurrentNum int32
+
+	queueIdChannel chan int64
+	workerQueue    chan task
+	tasks          chan task
+	idle           bool
+	workerNum      int32
+	cbChannel      chan func(error)
+
+	mapTaskQueueSession map[int64]*queue.Deque[task]
+
+	waitWorker   sync.WaitGroup
+	waitDispatch sync.WaitGroup
+}
+
+func (d *dispatch) open(minGoroutineNum int32, maxGoroutineNum int32, tasks chan task, cbChannel chan func(error)) {
+	d.minConcurrentNum = minGoroutineNum
+	d.maxConcurrentNum = maxGoroutineNum
+	d.tasks = tasks
+	d.mapTaskQueueSession = make(map[int64]*queue.Deque[task], maxTaskQueueSessionId)
+	d.workerQueue = make(chan task)
+	d.cbChannel = cbChannel
+	d.queueIdChannel = make(chan int64, cap(tasks))
+
+	d.waitDispatch.Add(1)
+	go d.run()
+}
+
+func (d *dispatch) run() {
+	defer d.waitDispatch.Done()
+	timeout := time.NewTimer(time.Duration(atomic.LoadInt64(&idleTimeout)))
+
+	for {
+		select {
+		case queueId := <-d.queueIdChannel:
+			d.processqueueEvent(queueId)
+		default:
+			select {
+			case t, ok := <-d.tasks:
+				if ok == false {
+					return
+				}
+				d.processTask(&t)
+			case queueId := <-d.queueIdChannel:
+				d.processqueueEvent(queueId)
+			case <-timeout.C:
+				d.processTimer()
+				if atomic.LoadInt32(&d.minConcurrentNum) == -1 && len(d.tasks) == 0 {
+					atomic.StoreInt64(&idleTimeout,int64(time.Millisecond * 10))
+				}
+				timeout.Reset(time.Duration(atomic.LoadInt64(&idleTimeout)))
+			}
+		}
+
+		if atomic.LoadInt32(&d.minConcurrentNum) == -1 && d.workerNum == 0 {
+			d.waitWorker.Wait()
+			d.cbChannel <- nil
+			return
+		}
+	}
+}
+
+func (d *dispatch) processTimer() {
+	if d.idle == true && d.workerNum > atomic.LoadInt32(&d.minConcurrentNum) {
+		d.processIdle()
+	}
+
+	d.idle = true
+}
+
+func (d *dispatch) processqueueEvent(queueId int64) {
+	d.idle = false
+
+	queueSession := d.mapTaskQueueSession[queueId]
+	if queueSession == nil {
+		return
+	}
+
+	queueSession.PopFront()
+	if queueSession.Len() == 0 {
+		return
+	}
+
+	t := queueSession.Front()
+	d.executeTask(&t)
+}
+
+func (d *dispatch) executeTask(t *task) {
+	select {
+	case d.workerQueue <- *t:
+		return
+	default:
+		if d.workerNum < d.maxConcurrentNum {
+			var work worker
+			work.start(&d.waitWorker, t, d)
+			return
+		}
+	}
+
+	d.workerQueue <- *t
+}
+
+func (d *dispatch) processTask(t *task) {
+	d.idle = false
+
+	//处理有排队任务
+	if t.queueId != 0 {
+		queueSession := d.mapTaskQueueSession[t.queueId]
+		if queueSession == nil {
+			queueSession = &queue.Deque[task]{}
+			d.mapTaskQueueSession[t.queueId] = queueSession
+		}
+
+		//没有正在执行的任务，则直接执行
+		if queueSession.Len() == 0 {
+			d.executeTask(t)
+		}
+
+		queueSession.PushBack(*t)
+		return
+	}
+
+	//普通任务
+	d.executeTask(t)
+}
+
+func (d *dispatch) processIdle() {
+	select {
+	case d.workerQueue <- task{}:
+		d.workerNum--
+	default:
+	}
+}
+
+func (d *dispatch) pushQueueTaskFinishEvent(queueId int64) {
+	d.queueIdChannel <- queueId
+}
+
+func (c *dispatch) pushAsyncDoCallbackEvent(cb func(err error)) {
+	if cb == nil {
+		//不需要回调的情况
+		return
+	}
+
+	c.cbChannel <- cb
+}
+
+func (d *dispatch) close() {
+	atomic.StoreInt32(&d.minConcurrentNum, -1)
+
+breakFor:
+	for {
+		select {
+		case cb := <-d.cbChannel:
+			if cb == nil {
+				break breakFor
+			}
+			cb(nil)
+		}
+	}
+
+	d.waitDispatch.Wait()
+}
+
+func (d *dispatch) DoCallback(cb func(err error)) {
+	defer func() {
+		if r := recover(); r != nil {
+			buf := make([]byte, 4096)
+			l := runtime.Stack(buf, false)
+			errString := fmt.Sprint(r)
+
+			log.SError("core dump info[", errString, "]\n", string(buf[:l]))
+		}
+	}()
+
+	cb(nil)
+}
--- a/concurrent/worker.go
+++ b/concurrent/worker.go
@@ -0,0 +1,79 @@
+package concurrent
+
+import (
+	"sync"
+
+	"errors"
+	"fmt"
+	"runtime"
+
+	"github.com/duanhf2012/origin/log"
+)
+
+type task struct {
+	queueId int64
+	fn      func() bool
+	cb      func(err error)
+}
+
+type worker struct {
+	*dispatch
+}
+
+func (t *task) isExistTask() bool {
+	return t.fn == nil
+}
+
+func (w *worker) start(waitGroup *sync.WaitGroup, t *task, d *dispatch) {
+	w.dispatch = d
+	d.workerNum += 1
+	waitGroup.Add(1)
+	go w.run(waitGroup, *t)
+}
+
+func (w *worker) run(waitGroup *sync.WaitGroup, t task) {
+	defer waitGroup.Done()
+
+	w.exec(&t)
+	for {
+		select {
+		case tw := <-w.workerQueue:
+			if tw.isExistTask() {
+				//exit goroutine
+				log.SRelease("worker goroutine exit")
+				return
+			}
+			w.exec(&tw)
+		}
+	}
+}
+
+func (w *worker) exec(t *task) {
+	defer func() {
+		if r := recover(); r != nil {
+			buf := make([]byte, 4096)
+			l := runtime.Stack(buf, false)
+			errString := fmt.Sprint(r)
+
+			cb := t.cb
+			t.cb = func(err error) {
+				cb(errors.New(errString))
+			}
+
+			w.endCallFun(true,t)
+			log.SError("core dump info[", errString, "]\n", string(buf[:l]))
+		}
+	}()
+
+	w.endCallFun(t.fn(),t)
+}
+
+func (w *worker) endCallFun(isDocallBack bool,t *task) {
+	if isDocallBack {
+		w.pushAsyncDoCallbackEvent(t.cb)
+	}
+
+	if t.queueId != 0 {
+		w.pushQueueTaskFinishEvent(t.queueId)
+	}
+}
--- a/event/event.go
+++ b/event/event.go
@@ -7,7 +7,6 @@ import (
 	"sync"
 )

-
 //事件接受器
 type EventCallBack func(event IEvent)

@@ -229,7 +228,6 @@ func (processor *EventProcessor) EventHandler(ev IEvent) {
 	}
 }

-
 func (processor *EventProcessor) castEvent(event IEvent){
 	if processor.mapListenerEvent == nil {
 		log.SError("mapListenerEvent not init!")
@@ -246,3 +244,4 @@ func (processor *EventProcessor) castEvent(event IEvent){
 		proc.PushEvent(event)
 	}
 }
+
--- a/event/eventpool.go
+++ b/event/eventpool.go
@@ -0,0 +1,24 @@
+package event
+
+import "github.com/duanhf2012/origin/util/sync"
+
+// eventPool的内存池,缓存Event
+const defaultMaxEventChannelNum = 2000000
+
+var eventPool = sync.NewPoolEx(make(chan sync.IPoolData, defaultMaxEventChannelNum), func() sync.IPoolData {
+	return &Event{}
+})
+
+func NewEvent() *Event{
+	return eventPool.Get().(*Event)
+}
+
+func DeleteEvent(event IEvent){
+	eventPool.Put(event.(sync.IPoolData))
+}
+
+func SetEventPoolSize(eventPoolSize int){
+	eventPool = sync.NewPoolEx(make(chan sync.IPoolData, eventPoolSize), func() sync.IPoolData {
+		return &Event{}
+	})
+}
--- a/event/eventtype.go
+++ b/event/eventtype.go
@@ -12,7 +12,11 @@ const (
 	Sys_Event_WebSocket       EventType = -5
 	Sys_Event_Node_Event      EventType = -6
 	Sys_Event_DiscoverService EventType = -7
-
+	Sys_Event_DiscardGoroutine EventType = -8
+	Sys_Event_QueueTaskFinish  EventType = -9
+	
 	Sys_Event_User_Define EventType = 1
+
+
 )

--- a/go.mod
+++ b/go.mod
@@ -23,8 +23,8 @@ require (
 	github.com/xdg-go/scram v1.0.2 // indirect
 	github.com/xdg-go/stringprep v1.0.2 // indirect
 	github.com/youmark/pkcs8 v0.0.0-20181117223130-1be2e3e5546d // indirect
-	golang.org/x/crypto v0.0.0-20201216223049-8b5274cf687f // indirect
+	golang.org/x/crypto v0.1.0 // indirect
 	golang.org/x/sync v0.0.0-20201020160332-67f06af15bc9 // indirect
-	golang.org/x/text v0.3.6 // indirect
+	golang.org/x/text v0.4.0 // indirect
 	gopkg.in/yaml.v2 v2.4.0 // indirect
 )
--- a/go.sum
+++ b/go.sum
@@ -58,8 +58,9 @@ go.mongodb.org/mongo-driver v1.9.1/go.mod h1:0sQWfOeY63QTntERDJJ/0SuKK0T1uVSgKCu
 golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
 golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
 golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
-golang.org/x/crypto v0.0.0-20201216223049-8b5274cf687f h1:aZp0e2vLN4MToVqnjNEYEtrEA8RH8U8FN1CU7JgqsPU=
 golang.org/x/crypto v0.0.0-20201216223049-8b5274cf687f/go.mod h1:jdWPYTVW3xRLrWPugEBEK3UY2ZEsg3UU495nc5E+M+I=
+golang.org/x/crypto v0.1.0 h1:MDRAIl0xIo9Io2xV565hzXHw3zVseKrJKodhohM5CjU=
+golang.org/x/crypto v0.1.0/go.mod h1:RecgLatLF4+eUMCP1PoPZQb+cVrJcOPbHkTkbkB9sbw=
 golang.org/x/mod v0.2.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
 golang.org/x/mod v0.3.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
 golang.org/x/net v0.0.0-20190311183353-d8887717615a/go.mod h1:t9HGtf8HONx5eT2rtn7q6eTqICYqUVnKs3thJo3Qplg=
@@ -79,8 +80,8 @@ golang.org/x/term v0.0.0-20201117132131-f5c789dd3221/go.mod h1:Nr5EML6q2oocZ2LXR
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
 golang.org/x/text v0.3.5/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
-golang.org/x/text v0.3.6 h1:aRYxNxv6iGQlyVaZmk6ZgYEDa+Jg18DxebPSrd6bg1M=
-golang.org/x/text v0.3.6/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
+golang.org/x/text v0.4.0 h1:BrVqGRd7+k1DiOgtnFvAkoQEWQvBc25ouMJM6429SFg=
+golang.org/x/text v0.4.0/go.mod h1:mrYo+phRRbMaCq/xk9113O4dZlRixOauAjOtrjsXDZ8=
 golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20190531172133-b3315ee88b7d/go.mod h1:/rFqwRUd4F7ZHNgwSSTFct+R/Kf4OFW1sUzUTQQTgfc=
 golang.org/x/tools v0.0.0-20191119224855-298f0cb1881e/go.mod h1:b+2E5dAYhXwXZwtnZ6UAqBI28+e2cm9otk0dWdXHAEo=
--- a/network/processor/pbrawprocessor.go
+++ b/network/processor/pbrawprocessor.go
@@ -78,7 +78,6 @@ func (pbRawProcessor *PBRawProcessor) SetRawMsgHandler(handle RawMessageHandler)
 func (pbRawProcessor *PBRawProcessor) MakeRawMsg(msgType uint16,msg []byte,pbRawPackInfo *PBRawPackInfo)  {
 	pbRawPackInfo.typ = msgType
 	pbRawPackInfo.rawMsg = msg
-	//return &PBRawPackInfo{typ:msgType,rawMsg:msg}
 }

 func (pbRawProcessor *PBRawProcessor) UnknownMsgRoute(clientId uint64,msg interface{}){
--- a/network/processor/processor.go
+++ b/network/processor/processor.go
@@ -17,17 +17,11 @@ type IProcessor interface {
 }

 type IRawProcessor interface {
-	SetByteOrder(littleEndian bool)
-	MsgRoute(clientId uint64,msg interface{}) error
-	Unmarshal(clientId uint64,data []byte) (interface{}, error)
-	Marshal(clientId uint64,msg interface{}) ([]byte, error)
+	IProcessor

+	SetByteOrder(littleEndian bool)
 	SetRawMsgHandler(handle RawMessageHandler)
 	MakeRawMsg(msgType uint16,msg []byte,pbRawPackInfo *PBRawPackInfo)
-	UnknownMsgRoute(clientId uint64,msg interface{})
-	ConnectedRoute(clientId uint64)
-	DisConnectedRoute(clientId uint64)
-
 	SetUnknownMsgHandler(unknownMessageHandler UnknownRawMessageHandler)
 	SetConnectedHandler(connectHandler RawConnectHandler)
 	SetDisConnectedHandler(disconnectHandler RawConnectHandler)
--- a/network/slicepool.go
+++ b/network/slicepool.go
@@ -34,7 +34,6 @@ func (areaPool *memAreaPool) makePool() {
 	for i := 0; i < poolLen; i++ {
 		memSize := (areaPool.minAreaValue - 1) + (i+1)*areaPool.growthValue
 		areaPool.pool[i] = sync.Pool{New: func() interface{} {
-			//fmt.Println("make memsize:",memSize)
 			return make([]byte, memSize)
 		}}
 	}
--- a/network/tcp_client.go
+++ b/network/tcp_client.go
@@ -22,11 +22,7 @@ type TCPClient struct {
 	closeFlag       bool

 	// msg parser
-	LenMsgLen    int
-	MinMsgLen    uint32
-	MaxMsgLen    uint32
-	LittleEndian bool
-	msgParser    *MsgParser
+	MsgParser
 }

 func (client *TCPClient) Start() {
@@ -69,14 +65,24 @@ func (client *TCPClient) init() {
 		log.SFatal("client is running")
 	}

+	if client.MinMsgLen == 0 {
+		client.MinMsgLen = Default_MinMsgLen
+	}
+	if client.MaxMsgLen == 0 {
+		client.MaxMsgLen = Default_MaxMsgLen
+	}
+	if client.LenMsgLen ==0 {
+		client.LenMsgLen = Default_LenMsgLen
+	}
+	maxMsgLen := client.MsgParser.getMaxMsgLen(client.LenMsgLen)
+	if client.MaxMsgLen > maxMsgLen {
+		client.MaxMsgLen = maxMsgLen
+		log.SRelease("invalid MaxMsgLen, reset to ", maxMsgLen)
+	}
+
 	client.cons = make(ConnSet)
 	client.closeFlag = false
-
-	// msg parser
-	msgParser := NewMsgParser()
-	msgParser.SetMsgLen(client.LenMsgLen, client.MinMsgLen, client.MaxMsgLen)
-	msgParser.SetByteOrder(client.LittleEndian)
-	client.msgParser = msgParser
+	client.MsgParser.init()
 }

 func (client *TCPClient) GetCloseFlag() bool{
@@ -120,7 +126,7 @@ reconnect:
 	client.cons[conn] = struct{}{}
 	client.Unlock()

-	tcpConn := newTCPConn(conn, client.PendingWriteNum, client.msgParser,client.WriteDeadline)
+	tcpConn := newTCPConn(conn, client.PendingWriteNum, &client.MsgParser,client.WriteDeadline)
 	agent := client.NewAgent(tcpConn)
 	agent.Run()

--- a/network/tcp_conn.go
+++ b/network/tcp_conn.go
@@ -1,11 +1,12 @@
 package network

 import (
+	"errors"
 	"github.com/duanhf2012/origin/log"
 	"net"
 	"sync"
+	"sync/atomic"
 	"time"
-	"errors"
 )

 type ConnSet map[net.Conn]struct{}
@@ -14,7 +15,7 @@ type TCPConn struct {
 	sync.Mutex
 	conn      net.Conn
 	writeChan chan []byte
-	closeFlag bool
+	closeFlag int32
 	msgParser *MsgParser
 }

@@ -49,7 +50,7 @@ func newTCPConn(conn net.Conn, pendingWriteNum int, msgParser *MsgParser,writeDe
 		conn.Close()
 		tcpConn.Lock()
 		freeChannel(tcpConn)
-		tcpConn.closeFlag = true
+		atomic.StoreInt32(&tcpConn.closeFlag,1)
 		tcpConn.Unlock()
 	}()

@@ -60,9 +61,9 @@ func (tcpConn *TCPConn) doDestroy() {
 	tcpConn.conn.(*net.TCPConn).SetLinger(0)
 	tcpConn.conn.Close()

-	if !tcpConn.closeFlag {
+	if atomic.LoadInt32(&tcpConn.closeFlag)==0 {
 		close(tcpConn.writeChan)
-		tcpConn.closeFlag = true
+		atomic.StoreInt32(&tcpConn.closeFlag,1)
 	}
 }

@@ -76,12 +77,12 @@ func (tcpConn *TCPConn) Destroy() {
 func (tcpConn *TCPConn) Close() {
 	tcpConn.Lock()
 	defer tcpConn.Unlock()
-	if tcpConn.closeFlag {
+	if atomic.LoadInt32(&tcpConn.closeFlag)==1 {
 		return
 	}

 	tcpConn.doWrite(nil)
-	tcpConn.closeFlag = true
+	atomic.StoreInt32(&tcpConn.closeFlag,1)
 }

 func (tcpConn *TCPConn) GetRemoteIp() string {
@@ -104,7 +105,7 @@ func (tcpConn *TCPConn) doWrite(b []byte) error{
 func (tcpConn *TCPConn) Write(b []byte) error{
 	tcpConn.Lock()
 	defer tcpConn.Unlock()
-	if tcpConn.closeFlag || b == nil {
+	if atomic.LoadInt32(&tcpConn.closeFlag)==1 || b == nil {
 		tcpConn.ReleaseReadMsg(b)
 		return errors.New("conn is close")
 	}
@@ -133,14 +134,14 @@ func (tcpConn *TCPConn) ReleaseReadMsg(byteBuff []byte){
 }

 func (tcpConn *TCPConn) WriteMsg(args ...[]byte) error {
-	if tcpConn.closeFlag == true {
+	if atomic.LoadInt32(&tcpConn.closeFlag) == 1 {
 		return errors.New("conn is close")
 	}
 	return tcpConn.msgParser.Write(tcpConn, args...)
 }

 func (tcpConn *TCPConn) WriteRawMsg(args []byte) error {
-	if tcpConn.closeFlag == true {
+	if atomic.LoadInt32(&tcpConn.closeFlag) == 1 {
 		return errors.New("conn is close")
 	}

@@ -149,7 +150,7 @@ func (tcpConn *TCPConn) WriteRawMsg(args []byte) error {


 func (tcpConn *TCPConn) IsConnected() bool {
-	return tcpConn.closeFlag == false
+	return atomic.LoadInt32(&tcpConn.closeFlag) == 0
 }

 func (tcpConn *TCPConn) SetReadDeadline(d time.Duration)  {
--- a/network/tcp_msg.go
+++ b/network/tcp_msg.go
@@ -11,62 +11,36 @@ import (
 // | len | data |
 // --------------
 type MsgParser struct {
-	lenMsgLen    int
-	minMsgLen    uint32
-	maxMsgLen    uint32
-	littleEndian bool
+	LenMsgLen    int
+	MinMsgLen    uint32
+	MaxMsgLen    uint32
+	LittleEndian bool

 	INetMempool
 }

-func NewMsgParser() *MsgParser {
-	p := new(MsgParser)
-	p.lenMsgLen = 2
-	p.minMsgLen = 1
-	p.maxMsgLen = 4096
-	p.littleEndian = false
-	p.INetMempool = NewMemAreaPool()
-	return p
-}

-// It's dangerous to call the method on reading or writing
-func (p *MsgParser) SetMsgLen(lenMsgLen int, minMsgLen uint32, maxMsgLen uint32) {
-	if lenMsgLen == 1 || lenMsgLen == 2 || lenMsgLen == 4 {
-		p.lenMsgLen = lenMsgLen
-	}
-	if minMsgLen != 0 {
-		p.minMsgLen = minMsgLen
-	}
-	if maxMsgLen != 0 {
-		p.maxMsgLen = maxMsgLen
-	}
-
-	var max uint32
-	switch p.lenMsgLen {
+func (p *MsgParser) getMaxMsgLen(lenMsgLen int) uint32 {
+	switch p.LenMsgLen {
 	case 1:
-		max = math.MaxUint8
+		return math.MaxUint8
 	case 2:
-		max = math.MaxUint16
+		return math.MaxUint16
 	case 4:
-		max = math.MaxUint32
-	}
-	if p.minMsgLen > max {
-		p.minMsgLen = max
-	}
-	if p.maxMsgLen > max {
-		p.maxMsgLen = max
+		return math.MaxUint32
+	default:
+		panic("LenMsgLen value must be 1 or 2 or 4")
 	}
 }

-// It's dangerous to call the method on reading or writing
-func (p *MsgParser) SetByteOrder(littleEndian bool) {
-	p.littleEndian = littleEndian
+func (p *MsgParser) init(){
+	p.INetMempool = NewMemAreaPool()
 }

 // goroutine safe
 func (p *MsgParser) Read(conn *TCPConn) ([]byte, error) {
 	var b [4]byte
-	bufMsgLen := b[:p.lenMsgLen]
+	bufMsgLen := b[:p.LenMsgLen]

 	// read len
 	if _, err := io.ReadFull(conn, bufMsgLen); err != nil {
@@ -75,17 +49,17 @@ func (p *MsgParser) Read(conn *TCPConn) ([]byte, error) {

 	// parse len
 	var msgLen uint32
-	switch p.lenMsgLen {
+	switch p.LenMsgLen {
 	case 1:
 		msgLen = uint32(bufMsgLen[0])
 	case 2:
-		if p.littleEndian {
+		if p.LittleEndian {
 			msgLen = uint32(binary.LittleEndian.Uint16(bufMsgLen))
 		} else {
 			msgLen = uint32(binary.BigEndian.Uint16(bufMsgLen))
 		}
 	case 4:
-		if p.littleEndian {
+		if p.LittleEndian {
 			msgLen = binary.LittleEndian.Uint32(bufMsgLen)
 		} else {
 			msgLen = binary.BigEndian.Uint32(bufMsgLen)
@@ -93,9 +67,9 @@ func (p *MsgParser) Read(conn *TCPConn) ([]byte, error) {
 	}

 	// check len
-	if msgLen > p.maxMsgLen {
+	if msgLen > p.MaxMsgLen {
 		return nil, errors.New("message too long")
-	} else if msgLen < p.minMsgLen {
+	} else if msgLen < p.MinMsgLen {
 		return nil, errors.New("message too short")
 	}
 	
@@ -118,26 +92,26 @@ func (p *MsgParser) Write(conn *TCPConn, args ...[]byte) error {
 	}

 	// check len
-	if msgLen > p.maxMsgLen {
+	if msgLen > p.MaxMsgLen {
 		return errors.New("message too long")
-	} else if msgLen < p.minMsgLen {
+	} else if msgLen < p.MinMsgLen {
 		return errors.New("message too short")
 	}

 	//msg := make([]byte, uint32(p.lenMsgLen)+msgLen)
-	msg := p.MakeByteSlice(p.lenMsgLen+int(msgLen))
+	msg := p.MakeByteSlice(p.LenMsgLen+int(msgLen))
 	// write len
-	switch p.lenMsgLen {
+	switch p.LenMsgLen {
 	case 1:
 		msg[0] = byte(msgLen)
 	case 2:
-		if p.littleEndian {
+		if p.LittleEndian {
 			binary.LittleEndian.PutUint16(msg, uint16(msgLen))
 		} else {
 			binary.BigEndian.PutUint16(msg, uint16(msgLen))
 		}
 	case 4:
-		if p.littleEndian {
+		if p.LittleEndian {
 			binary.LittleEndian.PutUint32(msg, msgLen)
 		} else {
 			binary.BigEndian.PutUint32(msg, msgLen)
@@ -145,7 +119,7 @@ func (p *MsgParser) Write(conn *TCPConn, args ...[]byte) error {
 	}

 	// write data
-	l := p.lenMsgLen
+	l := p.LenMsgLen
 	for i := 0; i < len(args); i++ {
 		copy(msg[l:], args[i])
 		l += len(args[i])
--- a/network/tcp_server.go
+++ b/network/tcp_server.go
@@ -7,14 +7,16 @@ import (
 	"time"
 )

-const Default_ReadDeadline  = time.Second*30  //30s
-const Default_WriteDeadline = time.Second*30 //30s
-const Default_MaxConnNum = 3000
-const Default_PendingWriteNum = 10000
-const Default_LittleEndian = false
-const Default_MinMsgLen = 2
-const Default_MaxMsgLen = 65535
-
+const(
+	Default_ReadDeadline  = time.Second*30 //默认读超时30s
+	Default_WriteDeadline = time.Second*30 //默认写超时30s
+	Default_MaxConnNum = 1000000						 //默认最大连接数
+	Default_PendingWriteNum = 100000				     //单连接写消息Channel容量
+	Default_LittleEndian = false						 //默认大小端
+	Default_MinMsgLen = 2								 //最小消息长度2byte
+	Default_LenMsgLen = 2								 //包头字段长度占用2byte
+	Default_MaxMsgLen = 65535                            //最大消息长度
+)

 type TCPServer struct {
 	Addr            string
@@ -22,6 +24,7 @@ type TCPServer struct {
 	PendingWriteNum int
 	ReadDeadline    time.Duration
 	WriteDeadline 	time.Duration
+
 	NewAgent        func(*TCPConn) Agent
 	ln              net.Listener
 	conns           ConnSet
@@ -29,14 +32,7 @@ type TCPServer struct {
 	wgLn            sync.WaitGroup
 	wgConns         sync.WaitGroup

-
-	// msg parser
-	LenMsgLen    int
-	MinMsgLen    uint32
-	MaxMsgLen    uint32
-	LittleEndian bool
-	msgParser    *MsgParser
-	netMemPool   INetMempool
+	MsgParser
 }

 func (server *TCPServer) Start() {
@@ -54,14 +50,15 @@ func (server *TCPServer) init() {
 		server.MaxConnNum = Default_MaxConnNum
 		log.SRelease("invalid MaxConnNum, reset to ", server.MaxConnNum)
 	}
+
 	if server.PendingWriteNum <= 0 {
 		server.PendingWriteNum = Default_PendingWriteNum
 		log.SRelease("invalid PendingWriteNum, reset to ", server.PendingWriteNum)
 	}

-	if server.MinMsgLen <= 0 {
-		server.MinMsgLen = Default_MinMsgLen
-		log.SRelease("invalid MinMsgLen, reset to ", server.MinMsgLen)
+	if server.LenMsgLen <= 0 {
+		server.LenMsgLen = Default_LenMsgLen
+		log.SRelease("invalid LenMsgLen, reset to ", server.LenMsgLen)
 	}

 	if server.MaxMsgLen <= 0 {
@@ -69,10 +66,22 @@ func (server *TCPServer) init() {
 		log.SRelease("invalid MaxMsgLen, reset to ", server.MaxMsgLen)
 	}

+	maxMsgLen := server.MsgParser.getMaxMsgLen(server.LenMsgLen)
+	if server.MaxMsgLen > maxMsgLen {
+		server.MaxMsgLen = maxMsgLen
+		log.SRelease("invalid MaxMsgLen, reset to ", maxMsgLen)
+	}
+	
+	if server.MinMsgLen <= 0 {
+		server.MinMsgLen = Default_MinMsgLen
+		log.SRelease("invalid MinMsgLen, reset to ", server.MinMsgLen)
+	}
+
 	if server.WriteDeadline == 0 {
 		server.WriteDeadline = Default_WriteDeadline
 		log.SRelease("invalid WriteDeadline, reset to ", server.WriteDeadline.Seconds(),"s")
 	}
+
 	if server.ReadDeadline == 0 {
 		server.ReadDeadline = Default_ReadDeadline
 		log.SRelease("invalid ReadDeadline, reset to ", server.ReadDeadline.Seconds(),"s")
@@ -84,24 +93,15 @@ func (server *TCPServer) init() {

 	server.ln = ln
 	server.conns = make(ConnSet)
-
-	// msg parser
-	msgParser := NewMsgParser()
-	if msgParser.INetMempool == nil {
-		msgParser.INetMempool = NewMemAreaPool()
-	}
-
-	msgParser.SetMsgLen(server.LenMsgLen, server.MinMsgLen, server.MaxMsgLen)
-	msgParser.SetByteOrder(server.LittleEndian)
-	server.msgParser = msgParser
+	server.MsgParser.init()
 }

 func (server *TCPServer) SetNetMempool(mempool INetMempool){
-	server.msgParser.INetMempool = mempool
+	server.INetMempool = mempool
 }

 func (server *TCPServer) GetNetMempool() INetMempool{
-	return server.msgParser.INetMempool
+	return server.INetMempool
 }

 func (server *TCPServer) run() {
@@ -127,6 +127,7 @@ func (server *TCPServer) run() {
 			}
 			return
 		}
+
 		conn.(*net.TCPConn).SetNoDelay(true)
 		tempDelay = 0

@@ -137,16 +138,16 @@ func (server *TCPServer) run() {
 			log.SWarning("too many connections")
 			continue
 		}
+
 		server.conns[conn] = struct{}{}
 		server.mutexConns.Unlock()
-
 		server.wgConns.Add(1)

-		tcpConn := newTCPConn(conn, server.PendingWriteNum, server.msgParser,server.WriteDeadline)
+		tcpConn := newTCPConn(conn, server.PendingWriteNum, &server.MsgParser,server.WriteDeadline)
 		agent := server.NewAgent(tcpConn)
+
 		go func() {
 			agent.Run()
-
 			// cleanup
 			tcpConn.Close()
 			server.mutexConns.Lock()
--- a/node/node.go
+++ b/node/node.go
@@ -22,7 +22,6 @@ import (
 	"time"
 )

-var closeSig chan bool
 var sig chan os.Signal
 var nodeId int
 var preSetupService []service.IService //预安装
@@ -40,8 +39,6 @@ const(
 )

 func init() {
-
-	closeSig = make(chan bool, 1)
 	sig = make(chan os.Signal, 3)
 	signal.Notify(sig, syscall.SIGINT, syscall.SIGTERM, syscall.Signal(10))

@@ -155,21 +152,28 @@ func initNode(id int) {
 		return
 	}

-	//2.setup service
-	for _, s := range preSetupService {
-		//是否配置的service
-		if cluster.GetCluster().IsConfigService(s.GetName()) == false {
-			continue
+	//2.顺序安装服务
+	serviceOrder := cluster.GetCluster().GetLocalNodeInfo().ServiceList
+	for _,serviceName:= range serviceOrder{
+		bSetup := false
+		for _, s := range preSetupService {
+			if s.GetName() != serviceName {
+				continue
+			}
+			bSetup = true
+			pServiceCfg := cluster.GetCluster().GetServiceCfg(s.GetName())
+			s.Init(s, cluster.GetRpcClient, cluster.GetRpcServer, pServiceCfg)
+
+			service.Setup(s)
 		}

-		pServiceCfg := cluster.GetCluster().GetServiceCfg(s.GetName())
-		s.Init(s, cluster.GetRpcClient, cluster.GetRpcServer, pServiceCfg)
-
-		service.Setup(s)
+		if bSetup == false {
+			log.SFatal("Service name "+serviceName+" configuration error")
+		}
 	}

 	//3.service初始化
-	service.Init(closeSig)
+	service.Init()
 }

 func initLog() error {
@@ -274,8 +278,7 @@ func startNode(args interface{}) error {
 	}
 	cluster.GetCluster().Stop()
 	//7.退出
-	close(closeSig)
-	service.WaitStop()
+	service.StopAllService()

 	log.SRelease("Server is stop.")
 	return nil
@@ -292,9 +295,9 @@ func GetService(serviceName string) service.IService {
 	return service.GetService(serviceName)
 }

-func SetConfigDir(configDir string) {
-	configDir = configDir
-	cluster.SetConfigDir(configDir)
+func SetConfigDir(cfgDir string) {
+	configDir = cfgDir
+	cluster.SetConfigDir(cfgDir)
 }

 func GetConfigDir() string {
--- a/profiler/profiler.go
+++ b/profiler/profiler.go
@@ -193,9 +193,11 @@ func Report() {

 		record = prof.record
 		prof.record = list.New()
+		callNum := prof.callNum
+		totalCostTime := prof.totalCostTime
 		prof.stackLocker.RUnlock()

-		DefaultReportFunction(name,prof.callNum,prof.totalCostTime,record)
+		DefaultReportFunction(name,callNum,totalCostTime,record)
 	}
 }

--- a/rpc/client.go
+++ b/rpc/client.go
@@ -3,91 +3,63 @@ package rpc
 import (
 	"container/list"
 	"errors"
-	"fmt"
-	"github.com/duanhf2012/origin/log"
 	"github.com/duanhf2012/origin/network"
-	"math"
 	"reflect"
-	"runtime"
 	"strconv"
 	"sync"
 	"sync/atomic"
 	"time"
 )

-type Client struct {
-	clientSeq uint32
-	id        int
-	bSelfNode bool
-	network.TCPClient
-	conn *network.TCPConn
+const(
+	DefaultRpcConnNum           = 1
+	DefaultRpcLenMsgLen         = 4
+	DefaultRpcMinMsgLen         = 2
+	DefaultMaxCheckCallRpcCount = 1000
+	DefaultMaxPendingWriteNum 	 = 200000

+
+	DefaultConnectInterval = 2*time.Second
+	DefaultCheckRpcCallTimeoutInterval = 5*time.Second
+	DefaultRpcTimeout = 15*time.Second
+)
+
+var clientSeq uint32
+
+type IRealClient interface {
+	SetConn(conn *network.TCPConn)
+	Close(waitDone bool)
+
+	AsyncCall(rpcHandler IRpcHandler, serviceMethod string, callback reflect.Value, args interface{}, replyParam interface{}) error
+	Go(rpcHandler IRpcHandler, noReply bool, serviceMethod string, args interface{}, reply interface{}) *Call
+	RawGo(rpcHandler IRpcHandler,processor IRpcProcessor, noReply bool, rpcMethodId uint32, serviceMethod string, rawArgs []byte, reply interface{}) *Call
+	IsConnected() bool
+
+	Run()
+	OnClose()
+}
+
+type Client struct {
+	clientId 		uint32
+	nodeId        int
 	pendingLock          sync.RWMutex
 	startSeq             uint64
 	pending              map[uint64]*list.Element
 	pendingTimer         *list.List
 	callRpcTimeout       time.Duration
 	maxCheckCallRpcCount int
-	TriggerRpcEvent
+
+	IRealClient
 }

-const MaxCheckCallRpcCount = 1000
-const MaxPendingWriteNum = 200000
-const ConnectInterval = 2*time.Second
-var clientSeq uint32
-
 func (client *Client) NewClientAgent(conn *network.TCPConn) network.Agent {
-	client.conn = conn
-	client.ResetPending()
+	client.SetConn(conn)

 	return client
 }

-
-func (client *Client) Connect(id int, addr string, maxRpcParamLen uint32) error {
-	client.clientSeq = atomic.AddUint32(&clientSeq, 1)
-	client.id = id
-	client.Addr = addr
-	client.maxCheckCallRpcCount = MaxCheckCallRpcCount
-	client.callRpcTimeout = 15 * time.Second
-	client.ConnectInterval = ConnectInterval
-	client.PendingWriteNum = MaxPendingWriteNum
-	client.AutoReconnect = true
-
-	client.ConnNum = 1
-	client.LenMsgLen = 4
-	client.MinMsgLen = 2
-	client.ReadDeadline = Default_ReadWriteDeadline
-	client.WriteDeadline = Default_ReadWriteDeadline
-
-	if maxRpcParamLen > 0 {
-		client.MaxMsgLen = maxRpcParamLen
-	} else {
-		client.MaxMsgLen = math.MaxUint32
-	}
-
-	client.NewAgent = client.NewClientAgent
-	client.LittleEndian = LittleEndian
-	client.ResetPending()
-	go client.startCheckRpcCallTimer()
-	if addr == "" {
-		client.bSelfNode = true
-		return nil
-	}
-
-	client.Start()
-	return nil
-}
-
-func (client *Client) startCheckRpcCallTimer() {
-	for {
-		time.Sleep(5 * time.Second)
-		client.checkRpcCallTimeout()
-	}
-}
-
-func (client *Client) makeCallFail(call *Call) {
-	client.removePending(call.Seq)
+func (bc *Client) makeCallFail(call *Call) {
+	bc.removePending(call.Seq)
 	if call.callback != nil && call.callback.IsValid() {
 		call.rpcHandler.PushRpcResponse(call)
 	} else {
@@ -95,29 +67,38 @@ func (client *Client) makeCallFail(call *Call) {
 	}
 }

-func (client *Client) checkRpcCallTimeout() {
-	now := time.Now()
+func (bc *Client) checkRpcCallTimeout() {
+	for{
+		time.Sleep(DefaultCheckRpcCallTimeoutInterval)
+		now := time.Now()

-	for i := 0; i < client.maxCheckCallRpcCount; i++ {
-		client.pendingLock.Lock()
-		pElem := client.pendingTimer.Front()
-		if pElem == nil {
-			client.pendingLock.Unlock()
+		for i := 0; i < bc.maxCheckCallRpcCount; i++ {
+			bc.pendingLock.Lock()
+			if bc.pendingTimer == nil {
+				bc.pendingLock.Unlock()
+				break
+			}
+
+			pElem := bc.pendingTimer.Front()
+			if pElem == nil {
+				bc.pendingLock.Unlock()
+				break
+			}
+			pCall := pElem.Value.(*Call)
+			if now.Sub(pCall.callTime) > bc.callRpcTimeout {
+				strTimeout := strconv.FormatInt(int64(bc.callRpcTimeout/time.Second), 10)
+				pCall.Err = errors.New("RPC call takes more than " + strTimeout + " seconds")
+				bc.makeCallFail(pCall)
+				bc.pendingLock.Unlock()
+				continue
+			}
+			bc.pendingLock.Unlock()
 			break
 		}
-		pCall := pElem.Value.(*Call)
-		if now.Sub(pCall.callTime) > client.callRpcTimeout {
-			strTimeout := strconv.FormatInt(int64(client.callRpcTimeout/time.Second), 10)
-			pCall.Err = errors.New("RPC call takes more than " + strTimeout + " seconds")
-			client.makeCallFail(pCall)
-			client.pendingLock.Unlock()
-			continue
-		}
-		client.pendingLock.Unlock()
 	}
 }

-func (client *Client) ResetPending() {
+func (client *Client) InitPending() {
 	client.pendingLock.Lock()
 	if client.pending != nil {
 		for _, v := range client.pending {
@@ -131,235 +112,62 @@ func (client *Client) ResetPending() {
 	client.pendingLock.Unlock()
 }

-func (client *Client) AddPending(call *Call) {
-	client.pendingLock.Lock()
+
+func (bc *Client) AddPending(call *Call) {
+	bc.pendingLock.Lock()
 	call.callTime = time.Now()
-	elemTimer := client.pendingTimer.PushBack(call)
-	client.pending[call.Seq] = elemTimer //如果下面发送失败，将会一一直存在这里
-	client.pendingLock.Unlock()
+	elemTimer := bc.pendingTimer.PushBack(call)
+	bc.pending[call.Seq] = elemTimer //如果下面发送失败，将会一一直存在这里
+	bc.pendingLock.Unlock()
 }

-func (client *Client) RemovePending(seq uint64) *Call {
-	if seq == 0 {
+func (bc *Client) RemovePending(seq uint64) *Call {
+	if seq == 0  {
 		return nil
 	}
-	client.pendingLock.Lock()
-	call := client.removePending(seq)
-	client.pendingLock.Unlock()
+	bc.pendingLock.Lock()
+	call := bc.removePending(seq)
+	bc.pendingLock.Unlock()
 	return call
 }

-func (client *Client) removePending(seq uint64) *Call {
-	v, ok := client.pending[seq]
+func (bc *Client) removePending(seq uint64) *Call {
+	v, ok := bc.pending[seq]
 	if ok == false {
 		return nil
 	}
 	call := v.Value.(*Call)
-	client.pendingTimer.Remove(v)
-	delete(client.pending, seq)
+	bc.pendingTimer.Remove(v)
+	delete(bc.pending, seq)
 	return call
 }

-func (client *Client) FindPending(seq uint64) *Call {
+func (bc *Client) FindPending(seq uint64) *Call {
 	if seq == 0 {
 		return nil
 	}
 	
-	client.pendingLock.Lock()
-	v, ok := client.pending[seq]
+	bc.pendingLock.Lock()
+	v, ok := bc.pending[seq]
 	if ok == false {
-		client.pendingLock.Unlock()
+		bc.pendingLock.Unlock()
 		return nil
 	}

 	pCall := v.Value.(*Call)
-	client.pendingLock.Unlock()
+	bc.pendingLock.Unlock()

 	return pCall
 }

-func (client *Client) generateSeq() uint64 {
-	return atomic.AddUint64(&client.startSeq, 1)
+func (bc *Client) generateSeq() uint64 {
+	return atomic.AddUint64(&bc.startSeq, 1)
 }

-func (client *Client) AsyncCall(rpcHandler IRpcHandler, serviceMethod string, callback reflect.Value, args interface{}, replyParam interface{}) error {
-	processorType, processor := GetProcessorType(args)
-	InParam, herr := processor.Marshal(args)
-	if herr != nil {
-		return herr
-	}
-
-	seq := client.generateSeq()
-	request := MakeRpcRequest(processor, seq, 0, serviceMethod, false, InParam)
-	bytes, err := processor.Marshal(request.RpcRequestData)
-	ReleaseRpcRequest(request)
-	if err != nil {
-		return err
-	}
-
-	if client.conn == nil {
-		return errors.New("Rpc server is disconnect,call " + serviceMethod)
-	}
-
-	call := MakeCall()
-	call.Reply = replyParam
-	call.callback = &callback
-	call.rpcHandler = rpcHandler
-	call.ServiceMethod = serviceMethod
-	call.Seq = seq
-	client.AddPending(call)
-
-	err = client.conn.WriteMsg([]byte{uint8(processorType)}, bytes)
-	if err != nil {
-		client.RemovePending(call.Seq)
-		ReleaseCall(call)
-		return err
-	}
-
-	return nil
+func (client *Client) GetNodeId() int {
+	return client.nodeId
 }

-func (client *Client) RawGo(processor IRpcProcessor, noReply bool, rpcMethodId uint32, serviceMethod string, args []byte, reply interface{}) *Call {
-	call := MakeCall()
-	call.ServiceMethod = serviceMethod
-	call.Reply = reply
-	call.Seq = client.generateSeq()
-
-	request := MakeRpcRequest(processor, call.Seq, rpcMethodId, serviceMethod, noReply, args)
-	bytes, err := processor.Marshal(request.RpcRequestData)
-	ReleaseRpcRequest(request)
-	if err != nil {
-		call.Seq = 0
-		call.Err = err
-		return call
-	}
-
-	if client.conn == nil {
-		call.Seq = 0
-		call.Err = errors.New(serviceMethod + "  was called failed,rpc client is disconnect")
-		return call
-	}
-
-	if noReply == false {
-		client.AddPending(call)
-	}
-
-	err = client.conn.WriteMsg([]byte{uint8(processor.GetProcessorType())}, bytes)
-	if err != nil {
-		client.RemovePending(call.Seq)
-		call.Seq = 0
-		call.Err = err
-	}
-
-	return call
-}
-
-func (client *Client) Go(noReply bool, serviceMethod string, args interface{}, reply interface{}) *Call {
-	_, processor := GetProcessorType(args)
-	InParam, err := processor.Marshal(args)
-	if err != nil {
-		call := MakeCall()
-		call.Err = err
-		return call
-	}
-
-	return client.RawGo(processor, noReply, 0, serviceMethod, InParam, reply)
-}
-
-func (client *Client) Run() {
-	defer func() {
-		if r := recover(); r != nil {
-			buf := make([]byte, 4096)
-			l := runtime.Stack(buf, false)
-			errString := fmt.Sprint(r)
-			log.SError("core dump info[", errString, "]\n", string(buf[:l]))
-		}
-	}()
-
-	client.TriggerRpcEvent(true, client.GetClientSeq(), client.GetId())
-	for {
-		bytes, err := client.conn.ReadMsg()
-		if err != nil {
-			log.SError("rpcClient ", client.Addr, " ReadMsg error:", err.Error())
-			return
-		}
-
-		processor := GetProcessor(bytes[0])
-		if processor == nil {
-			client.conn.ReleaseReadMsg(bytes)
-			log.SError("rpcClient ", client.Addr, " ReadMsg head error:", err.Error())
-			return
-		}
-
-		//1.解析head
-		response := RpcResponse{}
-		response.RpcResponseData = processor.MakeRpcResponse(0, "", nil)
-
-		err = processor.Unmarshal(bytes[1:], response.RpcResponseData)
-		client.conn.ReleaseReadMsg(bytes)
-		if err != nil {
-			processor.ReleaseRpcResponse(response.RpcResponseData)
-			log.SError("rpcClient Unmarshal head error:", err.Error())
-			continue
-		}
-
-		v := client.RemovePending(response.RpcResponseData.GetSeq())
-		if v == nil {
-			log.SError("rpcClient cannot find seq ", response.RpcResponseData.GetSeq(), " in pending")
-		} else {
-			v.Err = nil
-			if len(response.RpcResponseData.GetReply()) > 0 {
-				err = processor.Unmarshal(response.RpcResponseData.GetReply(), v.Reply)
-				if err != nil {
-					log.SError("rpcClient Unmarshal body error:", err.Error())
-					v.Err = err
-				}
-			}
-
-			if response.RpcResponseData.GetErr() != nil {
-				v.Err = response.RpcResponseData.GetErr()
-			}
-
-			if v.callback != nil && v.callback.IsValid() {
-				v.rpcHandler.PushRpcResponse(v)
-			} else {
-				v.done <- v
-			}
-		}
-
-		processor.ReleaseRpcResponse(response.RpcResponseData)
-	}
-}
-
-func (client *Client) OnClose() {
-	client.TriggerRpcEvent(false, client.GetClientSeq(), client.GetId())
-}
-
-func (client *Client) IsConnected() bool {
-	return client.bSelfNode || (client.conn != nil && client.conn.IsConnected() == true)
-}
-
-func (client *Client) GetId() int {
-	return client.id
-}
-
-func (client *Client) Close(waitDone bool) {
-	client.TCPClient.Close(waitDone)
-
-	client.pendingLock.Lock()
-	for  {
-		pElem := client.pendingTimer.Front()
-		if pElem == nil {
-			break
-		}
-
-		pCall := pElem.Value.(*Call)
-		pCall.Err = errors.New("nodeid is disconnect ")
-		client.makeCallFail(pCall)
-	}
-	client.pendingLock.Unlock()
-}
-
-func (client *Client) GetClientSeq() uint32 {
-	return client.clientSeq
+func (client *Client) GetClientId() uint32 {
+	return client.clientId
 }
--- a/rpc/gogopbprocessor.go
+++ b/rpc/gogopbprocessor.go
@@ -3,6 +3,7 @@ package rpc
 import (
 	"github.com/duanhf2012/origin/util/sync"
 	"github.com/gogo/protobuf/proto"
+	"fmt"
 )

 type GoGoPBProcessor struct {
@@ -40,7 +41,10 @@ func (slf *GoGoPBProcessor) Marshal(v interface{}) ([]byte, error){
 }

 func (slf *GoGoPBProcessor) Unmarshal(data []byte, msg interface{}) error{
-	protoMsg := msg.(proto.Message)
+	protoMsg,ok := msg.(proto.Message)
+	if ok == false {
+		return fmt.Errorf("%+v is not of proto.Message type",msg)
+	}
 	return proto.Unmarshal(data, protoMsg)
 }

@@ -73,6 +77,15 @@ func (slf *GoGoPBProcessor)	GetProcessorType() RpcProcessorType{
 	return RpcProcessorGoGoPB
 }

+func (slf *GoGoPBProcessor) Clone(src interface{}) (interface{},error){
+	srcMsg,ok := src.(proto.Message)
+	if ok == false {
+		return nil,fmt.Errorf("param is not of proto.message type")
+	}
+
+	return proto.Clone(srcMsg),nil
+}
+
 func (slf *GoGoPBRpcRequestData) IsNoReply() bool{
 	return slf.GetNoReply()
 }
@@ -91,5 +104,3 @@ func (slf *GoGoPBRpcResponseData)		GetErr() *RpcError {



-
-
--- a/rpc/jsonprocessor.go
+++ b/rpc/jsonprocessor.go
@@ -3,6 +3,7 @@ package rpc
 import (
 	"github.com/duanhf2012/origin/util/sync"
 	jsoniter "github.com/json-iterator/go"
+	"reflect"
 )

 var json = jsoniter.ConfigCompatibleWithStandardLibrary
@@ -119,6 +120,22 @@ func (jsonRpcResponseData *JsonRpcResponseData)		GetReply() []byte{
 }


+func (jsonProcessor *JsonProcessor) Clone(src interface{}) (interface{},error){
+	dstValue := reflect.New(reflect.ValueOf(src).Type().Elem())
+	bytes,err := json.Marshal(src)
+	if err != nil {
+		return nil,err
+	}
+	
+	dst := dstValue.Interface()
+	err = json.Unmarshal(bytes,dst)
+	if err != nil {
+		return nil,err
+	}
+
+	return dst,nil
+}
+



--- a/rpc/lclient.go
+++ b/rpc/lclient.go
@@ -0,0 +1,133 @@
+package rpc
+
+import (
+	"errors"
+	"github.com/duanhf2012/origin/log"
+	"github.com/duanhf2012/origin/network"
+	"reflect"
+	"strings"
+	"sync/atomic"
+)
+
+//本结点的Client
+type LClient struct {
+	selfClient *Client
+}
+
+func (rc *LClient) Lock(){
+}
+
+func (rc *LClient) Unlock(){
+}
+
+func (lc *LClient) Run(){
+}
+
+func (lc *LClient) OnClose(){
+}
+
+func (lc *LClient) IsConnected() bool {
+	return true
+}
+
+func (lc *LClient) SetConn(conn *network.TCPConn){
+}
+
+func (lc *LClient) Close(waitDone bool){
+}
+
+func (lc *LClient) Go(rpcHandler IRpcHandler,noReply bool, serviceMethod string, args interface{}, reply interface{}) *Call {
+	pLocalRpcServer := rpcHandler.GetRpcServer()()
+	//判断是否是同一服务
+	findIndex := strings.Index(serviceMethod, ".")
+	if findIndex == -1 {
+		sErr := errors.New("Call serviceMethod " + serviceMethod + " is error!")
+		log.SError(sErr.Error())
+		call := MakeCall()
+		call.DoError(sErr)
+
+		return call
+	}
+
+	serviceName := serviceMethod[:findIndex]
+	if serviceName == rpcHandler.GetName() { //自己服务调用
+		//调用自己rpcHandler处理器
+		err := pLocalRpcServer.myselfRpcHandlerGo(lc.selfClient,serviceName, serviceMethod, args, requestHandlerNull,reply)
+		call := MakeCall()
+
+		if err != nil {
+			call.DoError(err)
+			return call
+		}
+
+		call.DoOK()
+		return call
+	}
+
+	//其他的rpcHandler的处理器
+	return pLocalRpcServer.selfNodeRpcHandlerGo(nil, lc.selfClient, noReply, serviceName, 0, serviceMethod, args, reply, nil)
+}
+
+
+func (rc *LClient) RawGo(rpcHandler IRpcHandler,processor IRpcProcessor, noReply bool, rpcMethodId uint32, serviceName string, rawArgs []byte, reply interface{}) *Call {
+	pLocalRpcServer := rpcHandler.GetRpcServer()()
+
+	call := MakeCall()
+	call.ServiceMethod = serviceName
+	call.Reply = reply
+
+	//服务自我调用
+	if serviceName == rpcHandler.GetName() {
+		err := pLocalRpcServer.myselfRpcHandlerGo(rc.selfClient,serviceName, serviceName, rawArgs, requestHandlerNull,nil)
+		call.Err = err
+		call.done <- call
+
+		return call
+	}
+
+	//其他的rpcHandler的处理器
+	return pLocalRpcServer.selfNodeRpcHandlerGo(processor,rc.selfClient, true, serviceName, rpcMethodId, serviceName, nil, nil, rawArgs)
+}
+
+
+func (lc *LClient) AsyncCall(rpcHandler IRpcHandler, serviceMethod string, callback reflect.Value, args interface{}, reply interface{}) error {
+	pLocalRpcServer := rpcHandler.GetRpcServer()()
+
+	//判断是否是同一服务
+	findIndex := strings.Index(serviceMethod, ".")
+	if findIndex == -1 {
+		err := errors.New("Call serviceMethod " + serviceMethod + " is error!")
+		callback.Call([]reflect.Value{reflect.ValueOf(reply), reflect.ValueOf(err)})
+		log.SError(err.Error())
+		return nil
+	}
+
+	serviceName := serviceMethod[:findIndex]
+	//调用自己rpcHandler处理器
+	if serviceName == rpcHandler.GetName() { //自己服务调用
+		return pLocalRpcServer.myselfRpcHandlerGo(lc.selfClient,serviceName, serviceMethod, args,callback ,reply)
+	}
+
+	//其他的rpcHandler的处理器
+	err := pLocalRpcServer.selfNodeRpcHandlerAsyncGo(lc.selfClient, rpcHandler, false, serviceName, serviceMethod, args, reply, callback)
+	if err != nil {
+		callback.Call([]reflect.Value{reflect.ValueOf(reply), reflect.ValueOf(err)})
+	}
+
+	return nil
+}
+
+func NewLClient(nodeId int) *Client{
+	client := &Client{}
+	client.clientId = atomic.AddUint32(&clientSeq, 1)
+	client.nodeId = nodeId
+	client.maxCheckCallRpcCount = DefaultMaxCheckCallRpcCount
+	client.callRpcTimeout = DefaultRpcTimeout
+
+	lClient := &LClient{}
+	lClient.selfClient = client
+	client.IRealClient = lClient
+	client.InitPending()
+	go client.checkRpcCallTimeout()
+	return client
+}
--- a/rpc/processor.go
+++ b/rpc/processor.go
@@ -1,6 +1,7 @@
 package rpc

 type IRpcProcessor interface {
+	Clone(src interface{}) (interface{},error)
 	Marshal(v interface{}) ([]byte, error) //b表示自定义缓冲区，可以填nil，由系统自动分配
 	Unmarshal(data []byte, v interface{}) error
 	MakeRpcRequest(seq uint64,rpcMethodId uint32,serviceMethod string,noReply bool,inParam []byte) IRpcRequestData
--- a/rpc/rank.pb.go
+++ b/rpc/rank.pb.go
--- a/rpc/rank.proto
+++ b/rpc/rank.proto
@@ -2,19 +2,48 @@ syntax = "proto3";
 package rpc;
 option go_package = ".;rpc";

-// RankData 排行数据
-message RankData {
-    uint64 Key                  = 1; //数据主建
-    repeated int64 SortData     = 2; //参与排行的数据
-    bytes Data                  = 3; //不参与排行的数据
+message SetSortAndExtendData{
+    bool IsSortData  = 1; //是否为排序字段,为true时，修改Sort字段，否则修改Extend数据
+    int32 Pos        = 2;   //排序位置
+    int64 Data       = 3;  //排序值
+}
+
+//自增值
+message IncreaseRankData {
+    uint64      RankId                      = 1; //排行榜的ID
+    uint64      Key                         = 2; //数据主建
+    repeated    ExtendIncData Extend        = 3;  //扩展数据
+    repeated    int64   IncreaseSortData    = 4;//自增排行数值
+    repeated    SetSortAndExtendData SetSortAndExtendData     = 5;//设置排序数据值
+    bool        ReturnRankData              = 6; //是否查找最新排名，否则不返回排行Rank字段
+
+    bool        InsertDataOnNonExistent    = 7;   //为true时:存在不进行更新，不存在则插入InitData与InitSortData数据。为false时：忽略不对InitData与InitSortData数据
+    bytes       InitData                   = 8;   //不参与排行的数据
+    repeated int64 InitSortData            = 9;   //参与排行的数据
+}
+
+message IncreaseRankDataRet{
+    RankPosData PosData         = 1;
+}
+
+//用于单独刷新排行榜数据
+message UpdateRankData {
+    uint64      RankId                      = 1; //排行榜的ID
+    uint64      Key                         = 2; //数据主建
+    bytes       Data                        = 3; //数据部分
+}
+
+message UpdateRankDataRet {
+    bool Ret = 1;
 }

 // RankPosData 排行数据——查询返回
 message RankPosData {
    uint64 Key                  = 1; //数据主建
-    uint64 Rank              = 2; //名次
+    uint64 Rank                 = 2; //名次
    repeated int64 SortData     = 3; //参与排行的数据
    bytes Data                  = 4; //不参与排行的数据
+    repeated int64 ExtendData   = 5; //扩展数据
 }

 // RankList 排行榜数据
@@ -31,6 +60,22 @@ message RankList {
 message UpsetRankData {
    uint64      RankId                 = 1; //排行榜的ID
    repeated    RankData RankDataList  = 2; //排行数据
+    bool        FindNewRank            = 3; //是否查找最新排名
+}
+
+message ExtendIncData {
+    int64 InitValue       = 1;
+    int64 IncreaseValue   = 2;
+}
+
+// RankData 排行数据
+message RankData {
+    uint64 Key                  = 1; //数据主建
+    repeated int64 SortData     = 2; //参与排行的数据
+
+    bytes Data                  = 3; //不参与排行的数据
+
+    repeated    ExtendIncData ExData  = 4;  //扩展增量数据
 }

 // DeleteByKey 删除排行榜数据
@@ -71,9 +116,15 @@ message RankDataList {
 	RankPosData KeyRank                   = 3; //附带的Key查询排行结果信息
 }

+message RankInfo{
+  uint64 Key = 1;
+  uint64 Rank = 2;
+}
+
 // RankResult
 message RankResult {
    int32 AddCount = 1;//新增数量
    int32 ModifyCount = 2; //修改数量
    int32 RemoveCount = 3;//删除数量
+    repeated RankInfo NewRank    = 4; //新的排名名次，只有UpsetRankData.FindNewRank为true时才生效
 }
--- a/rpc/rclient.go
+++ b/rpc/rclient.go
@@ -0,0 +1,268 @@
+package rpc
+
+import (
+	"errors"
+	"fmt"
+	"github.com/duanhf2012/origin/log"
+	"github.com/duanhf2012/origin/network"
+	"math"
+	"reflect"
+	"runtime"
+	"sync/atomic"
+)
+
+//跨结点连接的Client
+type RClient struct {
+	selfClient *Client
+	network.TCPClient
+	conn *network.TCPConn
+	TriggerRpcConnEvent
+}
+
+func (rc *RClient) IsConnected() bool {
+	rc.Lock()
+	defer rc.Unlock()
+
+	return rc.conn != nil && rc.conn.IsConnected() == true
+}
+
+func (rc *RClient) GetConn() *network.TCPConn{
+	rc.Lock()
+	conn := rc.conn
+	rc.Unlock()
+
+	return conn
+}
+
+func (rc *RClient) SetConn(conn *network.TCPConn){
+	rc.Lock()
+	rc.conn = conn
+	rc.Unlock()
+}
+
+func (rc *RClient) Go(rpcHandler IRpcHandler,noReply bool, serviceMethod string, args interface{}, reply interface{}) *Call {
+	_, processor := GetProcessorType(args)
+	InParam, err := processor.Marshal(args)
+	if err != nil {
+		log.SError(err.Error())
+		call := MakeCall()
+		call.DoError(err)
+		return call
+	}
+
+	return rc.RawGo(rpcHandler,processor, noReply, 0, serviceMethod, InParam, reply)
+}
+
+
+func (rc *RClient) RawGo(rpcHandler IRpcHandler,processor IRpcProcessor, noReply bool, rpcMethodId uint32, serviceMethod string, rawArgs []byte, reply interface{}) *Call {
+	call := MakeCall()
+	call.ServiceMethod = serviceMethod
+	call.Reply = reply
+	call.Seq = rc.selfClient.generateSeq()
+
+	request := MakeRpcRequest(processor, call.Seq, rpcMethodId, serviceMethod, noReply, rawArgs)
+	bytes, err := processor.Marshal(request.RpcRequestData)
+	ReleaseRpcRequest(request)
+
+	if err != nil {
+		call.Seq = 0
+		log.SError(err.Error())
+		call.DoError(err)
+		return call
+	}
+
+	conn := rc.GetConn()
+	if conn == nil || conn.IsConnected()==false {
+		call.Seq = 0
+		sErr := errors.New(serviceMethod + "  was called failed,rpc client is disconnect")
+		log.SError(sErr.Error())
+		call.DoError(sErr)
+		return call
+	}
+
+	if noReply == false {
+		rc.selfClient.AddPending(call)
+	}
+
+	err = conn.WriteMsg([]byte{uint8(processor.GetProcessorType())}, bytes)
+	if err != nil {
+		rc.selfClient.RemovePending(call.Seq)
+
+		log.SError(err.Error())
+
+		call.Seq = 0
+		call.DoError(err)
+	}
+
+	return call
+}
+
+
+func (rc *RClient) AsyncCall(rpcHandler IRpcHandler, serviceMethod string, callback reflect.Value, args interface{}, replyParam interface{}) error {
+	err := rc.asyncCall(rpcHandler, serviceMethod, callback, args, replyParam)
+	if err != nil {
+		callback.Call([]reflect.Value{reflect.ValueOf(replyParam), reflect.ValueOf(err)})
+	}
+
+	return nil
+}
+
+func (rc *RClient) asyncCall(rpcHandler IRpcHandler, serviceMethod string, callback reflect.Value, args interface{}, replyParam interface{}) error {
+	processorType, processor := GetProcessorType(args)
+	InParam, herr := processor.Marshal(args)
+	if herr != nil {
+		return herr
+	}
+
+	seq := rc.selfClient.generateSeq()
+	request := MakeRpcRequest(processor, seq, 0, serviceMethod, false, InParam)
+	bytes, err := processor.Marshal(request.RpcRequestData)
+	ReleaseRpcRequest(request)
+	if err != nil {
+		return err
+	}
+
+	conn := rc.GetConn()
+	if conn == nil || conn.IsConnected()==false {
+		return errors.New("Rpc server is disconnect,call " + serviceMethod)
+	}
+
+	call := MakeCall()
+	call.Reply = replyParam
+	call.callback = &callback
+	call.rpcHandler = rpcHandler
+	call.ServiceMethod = serviceMethod
+	call.Seq = seq
+	rc.selfClient.AddPending(call)
+
+	err = conn.WriteMsg([]byte{uint8(processorType)}, bytes)
+	if err != nil {
+		rc.selfClient.RemovePending(call.Seq)
+		ReleaseCall(call)
+		return err
+	}
+
+	return nil
+}
+
+func (rc *RClient) Run() {
+	defer func() {
+		if r := recover(); r != nil {
+			buf := make([]byte, 4096)
+			l := runtime.Stack(buf, false)
+			errString := fmt.Sprint(r)
+			log.SError("core dump info[", errString, "]\n", string(buf[:l]))
+		}
+	}()
+
+	rc.TriggerRpcConnEvent(true, rc.selfClient.GetClientId(), rc.selfClient.GetNodeId())
+	for {
+		bytes, err := rc.conn.ReadMsg()
+		if err != nil {
+			log.SError("rpcClient ", rc.Addr, " ReadMsg error:", err.Error())
+			return
+		}
+
+		processor := GetProcessor(bytes[0])
+		if processor == nil {
+			rc.conn.ReleaseReadMsg(bytes)
+			log.SError("rpcClient ", rc.Addr, " ReadMsg head error:", err.Error())
+			return
+		}
+
+		//1.解析head
+		response := RpcResponse{}
+		response.RpcResponseData = processor.MakeRpcResponse(0, "", nil)
+
+		err = processor.Unmarshal(bytes[1:], response.RpcResponseData)
+		rc.conn.ReleaseReadMsg(bytes)
+		if err != nil {
+			processor.ReleaseRpcResponse(response.RpcResponseData)
+			log.SError("rpcClient Unmarshal head error:", err.Error())
+			continue
+		}
+
+		v := rc.selfClient.RemovePending(response.RpcResponseData.GetSeq())
+		if v == nil {
+			log.SError("rpcClient cannot find seq ", response.RpcResponseData.GetSeq(), " in pending")
+		} else {
+			v.Err = nil
+			if len(response.RpcResponseData.GetReply()) > 0 {
+				err = processor.Unmarshal(response.RpcResponseData.GetReply(), v.Reply)
+				if err != nil {
+					log.SError("rpcClient Unmarshal body error:", err.Error())
+					v.Err = err
+				}
+			}
+
+			if response.RpcResponseData.GetErr() != nil {
+				v.Err = response.RpcResponseData.GetErr()
+			}
+
+			if v.callback != nil && v.callback.IsValid() {
+				v.rpcHandler.PushRpcResponse(v)
+			} else {
+				v.done <- v
+			}
+		}
+
+		processor.ReleaseRpcResponse(response.RpcResponseData)
+	}
+}
+
+func (rc *RClient) OnClose() {
+	rc.TriggerRpcConnEvent(false, rc.selfClient.GetClientId(), rc.selfClient.GetNodeId())
+}
+
+func NewRClient(nodeId int, addr string, maxRpcParamLen uint32,triggerRpcConnEvent TriggerRpcConnEvent) *Client{
+	client := &Client{}
+	client.clientId = atomic.AddUint32(&clientSeq, 1)
+	client.nodeId = nodeId
+	client.maxCheckCallRpcCount = DefaultMaxCheckCallRpcCount
+	client.callRpcTimeout = DefaultRpcTimeout
+
+	c:= &RClient{}
+	c.selfClient = client
+	c.Addr = addr
+	c.ConnectInterval = DefaultConnectInterval
+	c.PendingWriteNum = DefaultMaxPendingWriteNum
+	c.AutoReconnect = true
+	c.TriggerRpcConnEvent = triggerRpcConnEvent
+	c.ConnNum = DefaultRpcConnNum
+	c.LenMsgLen = DefaultRpcLenMsgLen
+	c.MinMsgLen = DefaultRpcMinMsgLen
+	c.ReadDeadline = Default_ReadWriteDeadline
+	c.WriteDeadline = Default_ReadWriteDeadline
+	c.LittleEndian = LittleEndian
+	c.NewAgent = client.NewClientAgent
+
+	if maxRpcParamLen > 0 {
+		c.MaxMsgLen = maxRpcParamLen
+	} else {
+		c.MaxMsgLen = math.MaxUint32
+	}
+	client.IRealClient = c
+	client.InitPending()
+	go client.checkRpcCallTimeout()
+	c.Start()
+	return client
+}
+
+
+func (rc *RClient) Close(waitDone bool) {
+	rc.TCPClient.Close(waitDone)
+
+	rc.selfClient.pendingLock.Lock()
+	for  {
+		pElem := rc.selfClient.pendingTimer.Front()
+		if pElem == nil {
+			break
+		}
+
+		pCall := pElem.Value.(*Call)
+		pCall.Err = errors.New("nodeid is disconnect ")
+		rc.selfClient.makeCallFail(pCall)
+	}
+	rc.selfClient.pendingLock.Unlock()
+}
+
--- a/rpc/rpc.go
+++ b/rpc/rpc.go
@@ -51,12 +51,6 @@ type IRpcResponseData interface {
 	GetReply() []byte
 }

-type IRawInputArgs interface {
-	GetRawData() []byte  //获取原始数据
-	DoFree()           	 //处理完成,回收内存
-	DoEscape()         	 //逃逸,GC自动回收
-}
-
 type RpcHandleFinder interface {
 	FindRpcHandler(serviceMethod string) IRpcHandler
 }
@@ -108,6 +102,15 @@ func (rpcResponse *RpcResponse) Clear() *RpcResponse{
 	return rpcResponse
 }

+func (call *Call) DoError(err error){
+	call.Err = err
+	call.done <- call
+}
+
+func (call *Call) DoOK(){
+	call.done <- call
+}
+
 func (call *Call) Clear() *Call{
 	call.Seq = 0
 	call.ServiceMethod = ""
--- a/rpc/rpchandler.go
+++ b/rpc/rpchandler.go
@@ -6,7 +6,6 @@ import (
 	"github.com/duanhf2012/origin/log"
 	"reflect"
 	"runtime"
-	"strconv"
 	"strings"
 	"unicode"
 	"unicode/utf8"
@@ -17,6 +16,7 @@ const maxClusterNode int = 128
 type FuncRpcClient func(nodeId int, serviceMethod string, client []*Client) (error, int)
 type FuncRpcServer func() *Server

+
 var nilError = reflect.Zero(reflect.TypeOf((*error)(nil)).Elem())

 type RpcError string
@@ -45,10 +45,7 @@ type RpcMethodInfo struct {
 	rpcProcessorType RpcProcessorType
 }

-type RawRpcCallBack interface {
-	Unmarshal(data []byte) (interface{}, error)
-	CB(data interface{})
-}
+type  RawRpcCallBack func(rawData []byte)

 type IRpcHandlerChannel interface {
 	PushRpcResponse(call *Call) error
@@ -67,7 +64,7 @@ type RpcHandler struct {
 	pClientList []*Client
 }

-type TriggerRpcEvent func(bConnect bool, clientSeq uint32, nodeId int)
+type TriggerRpcConnEvent func(bConnect bool, clientSeq uint32, nodeId int)
 type INodeListener interface {
 	OnNodeConnected(nodeId int)
 	OnNodeDisconnect(nodeId int)
@@ -92,10 +89,11 @@ type IRpcHandler interface {
 	AsyncCallNode(nodeId int, serviceMethod string, args interface{}, callback interface{}) error
 	CallNode(nodeId int, serviceMethod string, args interface{}, reply interface{}) error
 	GoNode(nodeId int, serviceMethod string, args interface{}) error
-	RawGoNode(rpcProcessorType RpcProcessorType, nodeId int, rpcMethodId uint32, serviceName string, rawArgs IRawInputArgs) error
+	RawGoNode(rpcProcessorType RpcProcessorType, nodeId int, rpcMethodId uint32, serviceName string, rawArgs []byte) error
 	CastGo(serviceMethod string, args interface{}) error
 	IsSingleCoroutine() bool
 	UnmarshalInParam(rpcProcessor IRpcProcessor, serviceMethod string, rawRpcMethodId uint32, inParam []byte) (interface{}, error)
+	GetRpcServer() FuncRpcServer
 }

 func reqHandlerNull(Returns interface{}, Err RpcError) {
@@ -140,7 +138,7 @@ func (handler *RpcHandler) isExportedOrBuiltinType(t reflect.Type) bool {

 func (handler *RpcHandler) suitableMethods(method reflect.Method) error {
 	//只有RPC_开头的才能被调用
-	if strings.Index(method.Name, "RPC_") != 0 {
+	if strings.Index(method.Name, "RPC_") != 0 && strings.Index(method.Name, "RPC") != 0 {
 		return nil
 	}

@@ -244,8 +242,13 @@ func (handler *RpcHandler) HandlerRpcRequest(request *RpcRequest) {
 			log.SError("RpcHandler cannot find request rpc id", rawRpcId)
 			return
 		}
+		rawData,ok := request.inParam.([]byte)
+		if ok == false {
+			log.SError("RpcHandler " + handler.rpcHandler.GetName()," cannot convert in param to []byte", rawRpcId)
+			return
+		}

-		v.CB(request.inParam)
+		v(rawData)
 		return
 	}

@@ -288,14 +291,16 @@ func (handler *RpcHandler) HandlerRpcRequest(request *RpcRequest) {
 		request.requestHandle(nil, RpcError(rErr))
 		return
 	}
+
+	requestHanle := request.requestHandle
 	returnValues := v.method.Func.Call(paramList)
 	errInter := returnValues[0].Interface()
 	if errInter != nil {
 		err = errInter.(error)
 	}

-	if request.requestHandle != nil && v.hasResponder == false {
-		request.requestHandle(oParam.Interface(), ConvertError(err))
+	if v.hasResponder == false && requestHanle != nil  {
+		requestHanle(oParam.Interface(), ConvertError(err))
 	}
 }

@@ -427,36 +432,8 @@ func (handler *RpcHandler) goRpc(processor IRpcProcessor, bCast bool, nodeId int
 	}

 	//2.rpcClient调用
-	//如果调用本结点服务
 	for i := 0; i < count; i++ {
-		if pClientList[i].bSelfNode == true {
-			pLocalRpcServer := handler.funcRpcServer()
-			//判断是否是同一服务
-			findIndex := strings.Index(serviceMethod, ".")
-			if findIndex == -1 {
-				sErr := errors.New("Call serviceMethod " + serviceMethod + " is error!")
-				log.SError(sErr.Error())
-				err = sErr
-
-				continue
-			}
-			serviceName := serviceMethod[:findIndex]
-			if serviceName == handler.rpcHandler.GetName() { //自己服务调用
-				//调用自己rpcHandler处理器
-				return pLocalRpcServer.myselfRpcHandlerGo(pClientList[i],serviceName, serviceMethod, args, requestHandlerNull,nil)
-			}
-			//其他的rpcHandler的处理器
-			pCall := pLocalRpcServer.selfNodeRpcHandlerGo(processor, pClientList[i], true, serviceName, 0, serviceMethod, args, nil, nil)
-			if pCall.Err != nil {
-				err = pCall.Err
-			}
-			pClientList[i].RemovePending(pCall.Seq)
-			ReleaseCall(pCall)
-			continue
-		}
-
-		//跨node调用
-		pCall := pClientList[i].Go(true, serviceMethod, args, nil)
+		pCall := pClientList[i].Go(handler.rpcHandler,true, serviceMethod, args, nil)
 		if pCall.Err != nil {
 			err = pCall.Err
 		}
@@ -482,38 +459,9 @@ func (handler *RpcHandler) callRpc(nodeId int, serviceMethod string, args interf
 		return errors.New("cannot call more then 1 node")
 	}

-	//2.rpcClient调用
-	//如果调用本结点服务
 	pClient := pClientList[0]
-	if pClient.bSelfNode == true {
-		pLocalRpcServer := handler.funcRpcServer()
-		//判断是否是同一服务
-		findIndex := strings.Index(serviceMethod, ".")
-		if findIndex == -1 {
-			err := errors.New("Call serviceMethod " + serviceMethod + "is error!")
-			log.SError(err.Error())
-			return err
-		}
-		serviceName := serviceMethod[:findIndex]
-		if serviceName == handler.rpcHandler.GetName() { //自己服务调用
-			//调用自己rpcHandler处理器
-			return pLocalRpcServer.myselfRpcHandlerGo(pClient,serviceName, serviceMethod, args,requestHandlerNull, reply)
-		}
-		//其他的rpcHandler的处理器
-		pCall := pLocalRpcServer.selfNodeRpcHandlerGo(nil, pClient, false, serviceName, 0, serviceMethod, args, reply, nil)
-		err = pCall.Done().Err
-		pClient.RemovePending(pCall.Seq)
-		ReleaseCall(pCall)
-		return err
-	}
+	pCall := pClient.Go(handler.rpcHandler,false, serviceMethod, args, reply)

-	//跨node调用
-	pCall := pClient.Go(false, serviceMethod, args, reply)
-	if pCall.Err != nil {
-		err = pCall.Err
-		ReleaseCall(pCall)
-		return err
-	}
 	err = pCall.Done().Err
 	pClient.RemovePending(pCall.Seq)
 	ReleaseCall(pCall)
@@ -541,12 +489,15 @@ func (handler *RpcHandler) asyncCallRpc(nodeId int, serviceMethod string, args i
 	}

 	reply := reflect.New(fVal.Type().In(0).Elem()).Interface()
-	var pClientList [maxClusterNode]*Client
+	var pClientList [2]*Client
 	err, count := handler.funcRpcClient(nodeId, serviceMethod, pClientList[:])
 	if count == 0 || err != nil {
-		strNodeId := strconv.Itoa(nodeId)
 		if err == nil {
-			err = errors.New("cannot find rpcClient from nodeId " + strNodeId + " " + serviceMethod)
+			if nodeId  > 0 {
+				err = fmt.Errorf("cannot find %s from nodeId %d",serviceMethod,nodeId)
+			}else {
+				err = fmt.Errorf("No %s service found in the origin network",serviceMethod)
+			}
 		}
 		fVal.Call([]reflect.Value{reflect.ValueOf(reply), reflect.ValueOf(err)})
 		log.SError("Call serviceMethod is error:", err.Error())
@@ -563,35 +514,9 @@ func (handler *RpcHandler) asyncCallRpc(nodeId int, serviceMethod string, args i
 	//2.rpcClient调用
 	//如果调用本结点服务
 	pClient := pClientList[0]
-	if pClient.bSelfNode == true {
-		pLocalRpcServer := handler.funcRpcServer()
-		//判断是否是同一服务
-		findIndex := strings.Index(serviceMethod, ".")
-		if findIndex == -1 {
-			err := errors.New("Call serviceMethod " + serviceMethod + " is error!")
-			fVal.Call([]reflect.Value{reflect.ValueOf(reply), reflect.ValueOf(err)})
-			log.SError(err.Error())
-			return nil
-		}
-		serviceName := serviceMethod[:findIndex]
-		//调用自己rpcHandler处理器
-		if serviceName == handler.rpcHandler.GetName() { //自己服务调用
-			return pLocalRpcServer.myselfRpcHandlerGo(pClient,serviceName, serviceMethod, args,fVal ,reply)
-		}
+	pClient.AsyncCall(handler.rpcHandler, serviceMethod, fVal, args, reply)

-		//其他的rpcHandler的处理器
-		err = pLocalRpcServer.selfNodeRpcHandlerAsyncGo(pClient, handler, false, serviceName, serviceMethod, args, reply, fVal)
-		if err != nil {
-			fVal.Call([]reflect.Value{reflect.ValueOf(reply), reflect.ValueOf(err)})
-		}
-		return nil
-	}
-
-	//跨node调用
-	err = pClient.AsyncCall(handler, serviceMethod, fVal, args, reply)
-	if err != nil {
-		fVal.Call([]reflect.Value{reflect.ValueOf(reply), reflect.ValueOf(err)})
-	}
+	
 	return nil
 }

@@ -631,16 +556,14 @@ func (handler *RpcHandler) CastGo(serviceMethod string, args interface{}) error
 	return handler.goRpc(nil, true, 0, serviceMethod, args)
 }

-func (handler *RpcHandler) RawGoNode(rpcProcessorType RpcProcessorType, nodeId int, rpcMethodId uint32, serviceName string, rawArgs IRawInputArgs) error {
+func (handler *RpcHandler) RawGoNode(rpcProcessorType RpcProcessorType, nodeId int, rpcMethodId uint32, serviceName string, rawArgs []byte) error {
 	processor := GetProcessor(uint8(rpcProcessorType))
 	err, count := handler.funcRpcClient(nodeId, serviceName, handler.pClientList)
 	if count == 0 || err != nil {
-		//args.DoGc()
 		log.SError("Call serviceMethod is error:", err.Error())
 		return err
 	}
 	if count > 1 {
-		//args.DoGc()
 		err := errors.New("cannot call more then 1 node")
 		log.SError(err.Error())
 		return err
@@ -649,32 +572,12 @@ func (handler *RpcHandler) RawGoNode(rpcProcessorType RpcProcessorType, nodeId i
 	//2.rpcClient调用
 	//如果调用本结点服务
 	for i := 0; i < count; i++ {
-		if handler.pClientList[i].bSelfNode == true {
-			pLocalRpcServer := handler.funcRpcServer()
-			//调用自己rpcHandler处理器
-			if serviceName == handler.rpcHandler.GetName() { //自己服务调用
-				err := pLocalRpcServer.myselfRpcHandlerGo(handler.pClientList[i],serviceName, serviceName, rawArgs.GetRawData(), requestHandlerNull,nil)
-				//args.DoGc()
-				return err
-			}
-
-			//其他的rpcHandler的处理器
-			pCall := pLocalRpcServer.selfNodeRpcHandlerGo(processor, handler.pClientList[i], true, serviceName, rpcMethodId, serviceName, nil, nil, rawArgs.GetRawData())
-			rawArgs.DoEscape()
-			if pCall.Err != nil {
-				err = pCall.Err
-			}
-			handler.pClientList[i].RemovePending(pCall.Seq)
-			ReleaseCall(pCall)
-			continue
-		}
-
 		//跨node调用
-		pCall := handler.pClientList[i].RawGo(processor, true, rpcMethodId, serviceName, rawArgs.GetRawData(), nil)
-		rawArgs.DoFree()
+		pCall := handler.pClientList[i].RawGo(handler.rpcHandler,processor, true, rpcMethodId, serviceName, rawArgs, nil)
 		if pCall.Err != nil {
 			err = pCall.Err
 		}
+
 		handler.pClientList[i].RemovePending(pCall.Seq)
 		ReleaseCall(pCall)
 	}
@@ -688,23 +591,7 @@ func (handler *RpcHandler) RegRawRpc(rpcMethodId uint32, rawRpcCB RawRpcCallBack

 func (handler *RpcHandler) UnmarshalInParam(rpcProcessor IRpcProcessor, serviceMethod string, rawRpcMethodId uint32, inParam []byte) (interface{}, error) {
 	if rawRpcMethodId > 0 {
-		v, ok := handler.mapRawFunctions[rawRpcMethodId]
-		if ok == false {
-			strRawRpcMethodId := strconv.FormatUint(uint64(rawRpcMethodId), 10)
-			err := errors.New("RpcHandler cannot find request rpc id " + strRawRpcMethodId)
-			log.SError(err.Error())
-			return nil, err
-		}
-
-		msg, err := v.Unmarshal(inParam)
-		if err != nil {
-			strRawRpcMethodId := strconv.FormatUint(uint64(rawRpcMethodId), 10)
-			err := errors.New("RpcHandler cannot Unmarshal rpc id " + strRawRpcMethodId)
-			log.SError(err.Error())
-			return nil, err
-		}
-
-		return msg, err
+		return inParam,nil
 	}

 	v, ok := handler.mapFunctions[serviceMethod]
@@ -717,3 +604,8 @@ func (handler *RpcHandler) UnmarshalInParam(rpcProcessor IRpcProcessor, serviceM
 	err = rpcProcessor.Unmarshal(inParam, param)
 	return param, err
 }
+
+
+func (handler *RpcHandler) GetRpcServer() FuncRpcServer{
+	return handler.funcRpcServer
+}
--- a/rpc/server.go
+++ b/rpc/server.go
@@ -19,7 +19,6 @@ const (
 	RpcProcessorGoGoPB RpcProcessorType = 1
 )

-//var processor IRpcProcessor = &JsonProcessor{}
 var arrayProcessor = []IRpcProcessor{&JsonProcessor{}, &GoGoPBProcessor{}}
 var arrayProcessorLen uint8 = 2
 var LittleEndian bool
@@ -72,7 +71,6 @@ func (server *Server) Start(listenAddr string, maxRpcParamLen uint32) {
 	}

 	server.rpcServer.Addr = ":" + splitAddr[1]
-	server.rpcServer.LenMsgLen = 4 //uint16
 	server.rpcServer.MinMsgLen = 2
 	if maxRpcParamLen > 0 {
 		server.rpcServer.MaxMsgLen = maxRpcParamLen
@@ -86,6 +84,8 @@ func (server *Server) Start(listenAddr string, maxRpcParamLen uint32) {
 	server.rpcServer.LittleEndian = LittleEndian
 	server.rpcServer.WriteDeadline = Default_ReadWriteDeadline
 	server.rpcServer.ReadDeadline = Default_ReadWriteDeadline
+	server.rpcServer.LenMsgLen = DefaultRpcLenMsgLen
+
 	server.rpcServer.Start()
 }

@@ -148,7 +148,6 @@ func (agent *RpcAgent) Run() {
 				ReleaseRpcRequest(req)
 				continue
 			} else {
-				//will close tcpconn
 				ReleaseRpcRequest(req)
 				break
 			}
@@ -245,48 +244,39 @@ func (server *Server) myselfRpcHandlerGo(client *Client,handlerName string, serv
 		log.SError(err.Error())
 		return err
 	}
-
-
-
+	
 	return rpcHandler.CallMethod(client,serviceMethod, args,callBack, reply)
 }

+
+
 func (server *Server) selfNodeRpcHandlerGo(processor IRpcProcessor, client *Client, noReply bool, handlerName string, rpcMethodId uint32, serviceMethod string, args interface{}, reply interface{}, rawArgs []byte) *Call {
 	pCall := MakeCall()
 	pCall.Seq = client.generateSeq()

 	rpcHandler := server.rpcHandleFinder.FindRpcHandler(handlerName)
 	if rpcHandler == nil {
+		err := errors.New("service method " + serviceMethod + " not config!")
+		log.SError(err.Error())
 		pCall.Seq = 0
-		pCall.Err = errors.New("service method " + serviceMethod + " not config!")
-		pCall.done <- pCall
-		log.SError(pCall.Err.Error())
+		pCall.DoError(err)

 		return pCall
 	}

 	var iParam interface{}
-
-
 	if processor == nil {
 		_, processor = GetProcessorType(args)
 	}

 	if args != nil {
-		inParamValue := reflect.New(reflect.ValueOf(args).Type().Elem())
-		//args
-		//复制输入参数
-		iParam = inParamValue.Interface()
-		bytes,err := processor.Marshal(args)
-		if err == nil {
-			err = processor.Unmarshal(bytes,iParam)
-		}
-
+		var err error
+		iParam,err = processor.Clone(args)
 		if err != nil {
+			sErr := errors.New("RpcHandler " + handlerName + "."+serviceMethod+" deep copy inParam is error:" + err.Error())
+			log.SError(sErr.Error())
 			pCall.Seq = 0
-			pCall.Err = errors.New("RpcHandler " + handlerName + "."+serviceMethod+" deep copy inParam is error:" + err.Error())
-			pCall.done <- pCall
-			log.SError(pCall.Err.Error())
+			pCall.DoError(sErr)

 			return pCall
 		}
@@ -299,9 +289,10 @@ func (server *Server) selfNodeRpcHandlerGo(processor IRpcProcessor, client *Clie
 		var err error
 		req.inParam, err = rpcHandler.UnmarshalInParam(processor, serviceMethod, rpcMethodId, rawArgs)
 		if err != nil {
+			log.SError(err.Error())
+			pCall.Seq = 0
+			pCall.DoError(err)
 			ReleaseRpcRequest(req)
-			pCall.Err = err
-			pCall.done <- pCall
 			return pCall
 		}
 	}
@@ -313,38 +304,40 @@ func (server *Server) selfNodeRpcHandlerGo(processor IRpcProcessor, client *Clie
 			if reply != nil && Returns != reply && Returns != nil {
 				byteReturns, err := req.rpcProcessor.Marshal(Returns)
 				if err != nil {
-					log.SError("returns data cannot be marshal ", callSeq)
-					ReleaseRpcRequest(req)
-				}
-
-				err = req.rpcProcessor.Unmarshal(byteReturns, reply)
-				if err != nil {
-					log.SError("returns data cannot be Unmarshal ", callSeq)
-					ReleaseRpcRequest(req)
+					Err = ConvertError(err)
+					log.SError("returns data cannot be marshal,callSeq is ", callSeq," error is ",err.Error())
+				}else{
+					err = req.rpcProcessor.Unmarshal(byteReturns, reply)
+					if err != nil {
+						Err = ConvertError(err)
+						log.SError("returns data cannot be Unmarshal,callSeq is ", callSeq," error is ",err.Error())
+					}
 				}
 			}

+			ReleaseRpcRequest(req)
 			v := client.RemovePending(callSeq)
 			if v == nil {
 				log.SError("rpcClient cannot find seq ",callSeq, " in pending")
-				ReleaseRpcRequest(req)
+
 				return
 			}
+
 			if len(Err) == 0 {
 				v.Err = nil
+				v.DoOK()
 			} else {
-				v.Err = Err
+				log.SError(Err.Error())
+				v.DoError(Err)
 			}
-			v.done <- v
-			ReleaseRpcRequest(req)
 		}
 	}

 	err := rpcHandler.PushRpcRequest(req)
 	if err != nil {
+		log.SError(err.Error())
+		pCall.DoError(err)
 		ReleaseRpcRequest(req)
-		pCall.Err = err
-		pCall.done <- pCall
 	}

 	return pCall
@@ -359,15 +352,7 @@ func (server *Server) selfNodeRpcHandlerAsyncGo(client *Client, callerRpcHandler
 	}

 	_, processor := GetProcessorType(args)
-	inParamValue := reflect.New(reflect.ValueOf(args).Type().Elem())
-	//args
-	//复制输入参数
-	iParam := inParamValue.Interface()
-	bytes,err := processor.Marshal(args)
-	if err == nil {
-		err = processor.Unmarshal(bytes,iParam)
-	}
-
+	iParam,err := processor.Clone(args)
 	if err != nil {
 		errM := errors.New("RpcHandler " + handlerName + "."+serviceMethod+" deep copy inParam is error:" + err.Error())
 		log.SError(errM.Error())
--- a/service/module.go
+++ b/service/module.go
@@ -10,11 +10,13 @@ import (
 	"github.com/duanhf2012/origin/log"
 	rpcHandle "github.com/duanhf2012/origin/rpc"
 	"github.com/duanhf2012/origin/util/timer"
+	"github.com/duanhf2012/origin/concurrent"
 )

 const InitModuleId = 1e9

 type IModule interface {
+	concurrent.IConcurrent
 	SetModuleId(moduleId uint32) bool
 	GetModuleId() uint32
 	AddModule(module IModule) (uint32, error)
@@ -56,6 +58,7 @@ type Module struct {

 	//事件管道
 	eventHandler event.IEventHandler
+	concurrent.IConcurrent
 }

 func (m *Module) SetModuleId(moduleId uint32) bool {
@@ -105,6 +108,7 @@ func (m *Module) AddModule(module IModule) (uint32, error) {
 	pAddModule.moduleName = reflect.Indirect(reflect.ValueOf(module)).Type().Name()
 	pAddModule.eventHandler = event.NewEventHandler()
 	pAddModule.eventHandler.Init(m.eventHandler.GetEventProcessor())
+	pAddModule.IConcurrent = m.IConcurrent
 	err := module.OnInit()
 	if err != nil {
 		return 0, err
@@ -273,6 +277,11 @@ func (m *Module) SafeNewTicker(tickerId *uint64, d time.Duration, AdditionData i
 }

 func (m *Module) CancelTimerId(timerId *uint64) bool {
+	if timerId==nil || *timerId == 0 {
+		log.SWarning("timerId is invalid")
+		return false
+	}
+
 	if m.mapActiveIdTimer == nil {
 		log.SError("mapActiveIdTimer is nil")
 		return false
@@ -280,7 +289,7 @@ func (m *Module) CancelTimerId(timerId *uint64) bool {

 	t, ok := m.mapActiveIdTimer[*timerId]
 	if ok == false {
-		log.SError("cannot find timer id ", timerId)
+		log.SStack("cannot find timer id ", timerId)
 		return false
 	}

--- a/service/service.go
+++ b/service/service.go
@@ -7,22 +7,22 @@ import (
 	"github.com/duanhf2012/origin/log"
 	"github.com/duanhf2012/origin/profiler"
 	"github.com/duanhf2012/origin/rpc"
-	originSync "github.com/duanhf2012/origin/util/sync"
 	"github.com/duanhf2012/origin/util/timer"
 	"reflect"
 	"runtime"
 	"strconv"
 	"sync"
 	"sync/atomic"
+	"github.com/duanhf2012/origin/concurrent"
 )

-
-var closeSig chan bool
 var timerDispatcherLen = 100000
+var maxServiceEventChannelNum = 2000000

 type IService interface {
+	concurrent.IConcurrent
 	Init(iService IService,getClientFun rpc.FuncRpcClient,getServerFun rpc.FuncRpcServer,serviceCfg interface{})
-	Wait()
+	Stop()
 	Start()

 	OnSetup(iService IService)
@@ -42,14 +42,9 @@ type IService interface {
 	OpenProfiler()
 }

-// eventPool的内存池,缓存Event
-var maxServiceEventChannel = 2000000
-var eventPool = originSync.NewPoolEx(make(chan originSync.IPoolData, maxServiceEventChannel), func() originSync.IPoolData {
-	return &event.Event{}
-})
-
 type Service struct {
 	Module
+
 	rpcHandler rpc.RpcHandler           //rpc
 	name           string    //service name
 	wg             sync.WaitGroup
@@ -61,6 +56,7 @@ type Service struct {
 	nodeEventLister rpc.INodeListener
 	discoveryServiceLister rpc.IDiscoveryServiceListener
 	chanEvent chan event.IEvent
+	closeSig chan struct{}
 }

 // RpcConnEvent Node结点连接事件
@@ -77,10 +73,7 @@ type DiscoveryServiceEvent struct{
 }

 func SetMaxServiceChannel(maxEventChannel int){
-	maxServiceEventChannel = maxEventChannel
-	eventPool = originSync.NewPoolEx(make(chan originSync.IPoolData, maxServiceEventChannel), func() originSync.IPoolData {
-		return &event.Event{}
-	})
+	maxServiceEventChannelNum = maxEventChannel
 }

 func (rpcEventData *DiscoveryServiceEvent)  GetEventType() event.EventType{
@@ -105,9 +98,10 @@ func (s *Service) OpenProfiler()  {
 }

 func (s *Service) Init(iService IService,getClientFun rpc.FuncRpcClient,getServerFun rpc.FuncRpcServer,serviceCfg interface{}) {
+	s.closeSig = make(chan struct{})
 	s.dispatcher =timer.NewDispatcher(timerDispatcherLen)
 	if s.chanEvent == nil {
-		s.chanEvent = make(chan event.IEvent,maxServiceEventChannel)
+		s.chanEvent = make(chan event.IEvent,maxServiceEventChannelNum)
 	}

 	s.rpcHandler.InitRpcHandler(iService.(rpc.IRpcHandler),getClientFun,getServerFun,iService.(rpc.IRpcHandlerChannel))
@@ -123,29 +117,42 @@ func (s *Service) Init(iService IService,getClientFun rpc.FuncRpcClient,getServe
 	s.eventProcessor.Init(s)
 	s.eventHandler =  event.NewEventHandler()
 	s.eventHandler.Init(s.eventProcessor)
+	s.Module.IConcurrent = &concurrent.Concurrent{}
 }

-
 func (s *Service) Start() {
 	s.startStatus = true
+	var waitRun sync.WaitGroup
+
 	for i:=int32(0);i< s.goroutineNum;i++{
 		s.wg.Add(1)
+		waitRun.Add(1)
 		go func(){
+			log.SRelease(s.GetName()," service is running",)
+			waitRun.Done()
 			s.Run()
 		}()
 	}
+
+	waitRun.Wait()
 }

 func (s *Service) Run() {
-	log.SDebug("Start running Service ", s.GetName())
 	defer s.wg.Done()
 	var bStop = false
+	
+	concurrent := s.IConcurrent.(*concurrent.Concurrent)
+	concurrentCBChannel := concurrent.GetCallBackChannel()
+
 	s.self.(IService).OnStart()
 	for{
 		var analyzer *profiler.Analyzer
 		select {
-		case <- closeSig:
+		case <- s.closeSig:
 			bStop = true
+			concurrent.Close()
+		case cb:=<-concurrentCBChannel:
+			concurrent.DoCallback(cb)
 		case ev := <- s.chanEvent:
 			switch ev.GetEventType() {
 			case event.ServiceRpcRequestEvent:
@@ -168,7 +175,7 @@ func (s *Service) Run() {
 					analyzer.Pop()
 					analyzer = nil
 				}
-				eventPool.Put(cEvent)
+				event.DeleteEvent(cEvent)
 			case event.ServiceRpcResponseEvent:
 				cEvent,ok := ev.(*event.Event)
 				if ok == false {
@@ -188,7 +195,7 @@ func (s *Service) Run() {
 					analyzer.Pop()
 					analyzer = nil
 				}
-				eventPool.Put(cEvent)
+				event.DeleteEvent(cEvent)
 			default:
 				if s.profiler!=nil {
 					analyzer = s.profiler.Push("[SEvent]"+strconv.Itoa(int(ev.GetEventType())))
@@ -238,8 +245,8 @@ func (s *Service) Release(){
 			log.SError("core dump info[",errString,"]\n",string(buf[:l]))
 		}
 	}()
+	
 	s.self.OnRelease()
-	log.SDebug("Release Service ", s.GetName())
 }

 func (s *Service) OnRelease(){
@@ -249,8 +256,11 @@ func (s *Service) OnInit() error {
 	return nil
 }

-func (s *Service) Wait(){
+func (s *Service) Stop(){
+	log.SRelease("stop ",s.GetName()," service ")
+	close(s.closeSig)
 	s.wg.Wait()
+	log.SRelease(s.GetName()," service has been stopped")
 }

 func (s *Service) GetServiceCfg()interface{}{
@@ -320,9 +330,8 @@ func (s *Service) UnRegDiscoverListener(rpcLister rpc.INodeListener) {
 	UnRegDiscoveryServiceEventFun(s.GetName())
 }

-
 func (s *Service) PushRpcRequest(rpcRequest *rpc.RpcRequest) error{
-	ev := eventPool.Get().(*event.Event)
+	ev := event.NewEvent()
 	ev.Type = event.ServiceRpcRequestEvent
 	ev.Data = rpcRequest

@@ -330,7 +339,7 @@ func (s *Service) PushRpcRequest(rpcRequest *rpc.RpcRequest) error{
 }

 func (s *Service) PushRpcResponse(call *rpc.Call) error{
-	ev := eventPool.Get().(*event.Event)
+	ev := event.NewEvent()
 	ev.Type = event.ServiceRpcResponseEvent
 	ev.Data = call

@@ -342,7 +351,7 @@ func (s *Service) PushEvent(ev event.IEvent) error{
 }

 func (s *Service) pushEvent(ev event.IEvent) error{
-	if len(s.chanEvent) >= maxServiceEventChannel {
+	if len(s.chanEvent) >= maxServiceEventChannelNum {
 		err := errors.New("The event channel in the service is full")
 		log.SError(err.Error())
 		return err
--- a/service/servicemgr.go
+++ b/service/servicemgr.go
@@ -19,9 +19,7 @@ func init(){
 	setupServiceList = []IService{}
 }

-func Init(chanCloseSig chan bool) {
-	closeSig=chanCloseSig
-
+func Init() {
 	for _,s := range setupServiceList {
 		err := s.OnInit()
 		if err != nil {
@@ -57,8 +55,8 @@ func Start(){
 	}
 }

-func WaitStop(){
+func StopAllService(){
 	for i := len(setupServiceList) - 1; i >= 0; i-- {
-		setupServiceList[i].Wait()
+		setupServiceList[i].Stop()
 	}
 }
--- a/sysservice/messagequeueservice/MongoPersist.go
+++ b/sysservice/messagequeueservice/MongoPersist.go
@@ -1,18 +1,49 @@
 package messagequeueservice

 import (
+	"errors"
 	"fmt"
 	"github.com/duanhf2012/origin/log"
 	"github.com/duanhf2012/origin/service"
 	"github.com/duanhf2012/origin/sysmodule/mongodbmodule"
 	"go.mongodb.org/mongo-driver/bson"
 	"go.mongodb.org/mongo-driver/mongo/options"
-	"sunserver/common/util"
 	"time"
 )

 const MaxDays = 180

+type DataType interface {
+	int | uint | int64 | uint64 | float32 | float64 | int32 | uint32 | int16 | uint16
+}
+
+func convertToNumber[DType DataType](val interface{}) (error, DType) {
+	switch val.(type) {
+	case int64:
+		return nil, DType(val.(int64))
+	case int:
+		return nil, DType(val.(int))
+	case uint:
+		return nil, DType(val.(uint))
+	case uint64:
+		return nil, DType(val.(uint64))
+	case float32:
+		return nil, DType(val.(float32))
+	case float64:
+		return nil, DType(val.(float64))
+	case int32:
+		return nil, DType(val.(int32))
+	case uint32:
+		return nil, DType(val.(uint32))
+	case int16:
+		return nil, DType(val.(int16))
+	case uint16:
+		return nil, DType(val.(uint16))
+	}
+
+	return errors.New("unsupported type"), 0
+}
+
 type MongoPersist struct {
 	service.Module
 	mongo mongodbmodule.MongoModule
@@ -363,7 +394,7 @@ func (mp *MongoPersist) GetIndex(topicData *TopicData) uint64 {

 	for _, e := range document {
 		if e.Key == "_id" {
-			errC, seq := util.ConvertToNumber[uint64](e.Value)
+			errC, seq := convertToNumber[uint64](e.Value)
 			if errC != nil {
 				log.Error("value is error:%s,%+v, ", errC.Error(), e.Value)
 			}
--- a/sysservice/rankservice/MongodbPersist.go
+++ b/sysservice/rankservice/MongodbPersist.go
@@ -6,9 +6,9 @@ import (
 	"github.com/duanhf2012/origin/rpc"
 	"github.com/duanhf2012/origin/service"
 	"github.com/duanhf2012/origin/sysmodule/mongodbmodule"
-	"github.com/duanhf2012/origin/util/coroutine"
 	"go.mongodb.org/mongo-driver/bson"
 	"go.mongodb.org/mongo-driver/mongo/options"
+	"runtime"
 	"sync"
 	"sync/atomic"
 	"time"
@@ -18,10 +18,11 @@ const batchRemoveNum  = 128  //一切删除的最大数量

 // RankDataDB 排行表数据
 type RankDataDB struct {
-	Id                 uint64                      `bson:"_id,omitempty"`
-	RefreshTime        int64                       `bson:"RefreshTime,omitempty"`
-	SortData             []int64  				   `bson:"SortData,omitempty"`
-	Data                 []byte   				   `bson:"Data,omitempty"`
+	Id                 uint64                      `bson:"_id"`
+	RefreshTime        int64                       `bson:"RefreshTime"`
+	SortData             []int64  				   `bson:"SortData"`
+	Data                 []byte   				   `bson:"Data"`
+	ExData               []int64                    `bson:"ExData"`
 }

 // MongoPersist持久化Module
@@ -70,7 +71,9 @@ func (mp *MongoPersist) OnInit() error {
 	}

 	//开启协程
-	coroutine.GoRecover(mp.persistCoroutine,-1)
+	mp.waitGroup.Add(1)
+	go mp.persistCoroutine()
+
 	return nil
 }

@@ -186,6 +189,9 @@ func (mp *MongoPersist) loadFromDB(rankId uint64,rankCollectName string) error{
 		rankData.Data = rankDataDB.Data
 		rankData.Key = rankDataDB.Id
 		rankData.SortData = rankDataDB.SortData
+		for _,eData := range rankDataDB.ExData{
+			rankData.ExData = append(rankData.ExData,&rpc.ExtendIncData{InitValue:eData})
+		}

 		//更新到排行榜
 		rankSkip.UpsetRank(&rankData,rankDataDB.RefreshTime,true)
@@ -256,7 +262,6 @@ func (mp *MongoPersist) JugeTimeoutSave() bool{
 }

 func (mp *MongoPersist)  persistCoroutine(){
-	mp.waitGroup.Add(1)
 	defer mp.waitGroup.Done()
 	for atomic.LoadInt32(&mp.stop)==0 || mp.hasPersistData(){
 		//间隔时间sleep
@@ -287,6 +292,15 @@ func (mp *MongoPersist) hasPersistData() bool{
 }

 func (mp *MongoPersist) saveToDB(){
+	defer func() {
+		if r := recover(); r != nil {
+			buf := make([]byte, 4096)
+			l := runtime.Stack(buf, false)
+			errString := fmt.Sprint(r)
+			log.SError(" Core dump info[", errString, "]\n", string(buf[:l]))
+		}
+	}()
+
 	//1.copy数据
 	mp.Lock()
 	mapRemoveRankData := mp.mapRemoveRankData
@@ -343,7 +357,7 @@ func (mp *MongoPersist) removeRankData(rankId uint64,keys []uint64) bool {

 func (mp *MongoPersist) upsertToDB(collectName string,rankData *RankData) error{
 	condition := bson.D{{"_id", rankData.Key}}
-	upsert := bson.M{"_id":rankData.Key,"RefreshTime": rankData.refreshTimestamp, "SortData": rankData.SortData, "Data": rankData.Data}
+	upsert := bson.M{"_id":rankData.Key,"RefreshTime": rankData.refreshTimestamp, "SortData": rankData.SortData, "Data": rankData.Data,"ExData":rankData.ExData}
 	update := bson.M{"$set": upsert}

 	s := mp.mongo.TakeSession()
--- a/sysservice/rankservice/RankData.go
+++ b/sysservice/rankservice/RankData.go
@@ -14,7 +14,11 @@ var RankDataPool = sync.NewPoolEx(make(chan sync.IPoolData, 10240), func() sync.
 })

 type RankData struct {
-	*rpc.RankData
+	Key                  uint64
+	SortData             []int64
+	Data                 []byte
+	ExData               []int64
+
 	refreshTimestamp int64 //刷新时间
 	//bRelease bool
 	ref         bool
@@ -27,7 +31,14 @@ func NewRankData(isDec bool, data *rpc.RankData,refreshTimestamp int64) *RankDat
 	if isDec {
 		ret.compareFunc = ret.desCompare
 	}
-	ret.RankData = data
+	ret.Key = data.Key
+	ret.SortData = data.SortData
+	ret.Data = data.Data
+
+	for _,d := range data.ExData{
+		ret.ExData = append(ret.ExData,d.InitValue+d.IncreaseValue)
+	}
+
 	ret.refreshTimestamp = refreshTimestamp

 	return ret
--- a/sysservice/rankservice/RankService.go
+++ b/sysservice/rankservice/RankService.go
@@ -2,13 +2,15 @@ package rankservice

 import (
 	"fmt"
+	"time"
+
 	"github.com/duanhf2012/origin/log"
 	"github.com/duanhf2012/origin/rpc"
 	"github.com/duanhf2012/origin/service"
-	"time"
 )

 const PreMapRankSkipLen = 10
+
 type RankService struct {
 	service.Service

@@ -61,11 +63,11 @@ func (rs *RankService) RPC_ManualAddRankSkip(addInfo *rpc.AddRankList, addResult
 			continue
 		}

-		newSkip := NewRankSkip(addRankListData.RankId,addRankListData.RankName,addRankListData.IsDec, transformLevel(addRankListData.SkipListLevel), addRankListData.MaxRank,time.Duration(addRankListData.ExpireMs)*time.Millisecond)
+		newSkip := NewRankSkip(addRankListData.RankId, addRankListData.RankName, addRankListData.IsDec, transformLevel(addRankListData.SkipListLevel), addRankListData.MaxRank, time.Duration(addRankListData.ExpireMs)*time.Millisecond)
 		newSkip.SetupRankModule(rs.rankModule)

 		rs.mapRankSkip[addRankListData.RankId] = newSkip
-		rs.rankModule.OnSetupRank(true,newSkip)
+		rs.rankModule.OnSetupRank(true, newSkip)
 	}

 	addResult.AddCount = 1
@@ -82,6 +84,52 @@ func (rs *RankService) RPC_UpsetRank(upsetInfo *rpc.UpsetRankData, upsetResult *
 	addCount, updateCount := rankSkip.UpsetRankList(upsetInfo.RankDataList)
 	upsetResult.AddCount = addCount
 	upsetResult.ModifyCount = updateCount
+
+	if upsetInfo.FindNewRank == true {
+		for _, rdata := range upsetInfo.RankDataList {
+			_, rank := rankSkip.GetRankNodeData(rdata.Key)
+			upsetResult.NewRank = append(upsetResult.NewRank, &rpc.RankInfo{Key: rdata.Key, Rank: rank})
+		}
+	}
+
+	return nil
+}
+
+// RPC_IncreaseRankData 增量更新排行扩展数据
+func (rs *RankService) RPC_IncreaseRankData(changeRankData *rpc.IncreaseRankData, changeRankDataRet *rpc.IncreaseRankDataRet) error {
+	rankSkip, ok := rs.mapRankSkip[changeRankData.RankId]
+	if ok == false || rankSkip == nil {
+		return fmt.Errorf("RPC_ChangeRankData[", changeRankData.RankId, "] no this rank id")
+	}
+
+	ret := rankSkip.ChangeExtendData(changeRankData)
+	if ret == false {
+		return fmt.Errorf("RPC_ChangeRankData[", changeRankData.RankId, "] no this key ", changeRankData.Key)
+	}
+
+	if changeRankData.ReturnRankData == true {
+		rankData, rank := rankSkip.GetRankNodeData(changeRankData.Key)
+		changeRankDataRet.PosData = &rpc.RankPosData{}
+		changeRankDataRet.PosData.Rank = rank
+
+		changeRankDataRet.PosData.Key = rankData.Key
+		changeRankDataRet.PosData.Data = rankData.Data
+		changeRankDataRet.PosData.SortData = rankData.SortData
+		changeRankDataRet.PosData.ExtendData = rankData.ExData
+	}
+
+	return nil
+}
+
+// RPC_UpsetRank 更新排行榜
+func (rs *RankService) RPC_UpdateRankData(updateRankData *rpc.UpdateRankData, updateRankDataRet *rpc.UpdateRankDataRet) error {
+	rankSkip, ok := rs.mapRankSkip[updateRankData.RankId]
+	if ok == false || rankSkip == nil {
+		updateRankDataRet.Ret = false
+		return nil
+	}
+
+	updateRankDataRet.Ret = rankSkip.UpdateRankData(updateRankData)
 	return nil
 }

@@ -114,6 +162,7 @@ func (rs *RankService) RPC_FindRankDataByKey(findInfo *rpc.FindRankDataByKey, fi
 		findResult.Key = findRankData.Key
 		findResult.SortData = findRankData.SortData
 		findResult.Rank = rank
+		findResult.ExtendData = findRankData.ExData
 	}
 	return nil
 }
@@ -131,6 +180,7 @@ func (rs *RankService) RPC_FindRankDataByRank(findInfo *rpc.FindRankDataByRank,
 		findResult.Key = findRankData.Key
 		findResult.SortData = findRankData.SortData
 		findResult.Rank = rankPos
+		findResult.ExtendData = findRankData.ExData
 	}
 	return nil
 }
@@ -139,7 +189,7 @@ func (rs *RankService) RPC_FindRankDataByRank(findInfo *rpc.FindRankDataByRank,
 func (rs *RankService) RPC_FindRankDataList(findInfo *rpc.FindRankDataList, findResult *rpc.RankDataList) error {
 	rankObj, ok := rs.mapRankSkip[findInfo.RankId]
 	if ok == false || rankObj == nil {
-		err := fmt.Errorf("not config rank %d",findInfo.RankId)
+		err := fmt.Errorf("not config rank %d", findInfo.RankId)
 		log.SError(err.Error())
 		return err
 	}
@@ -151,7 +201,7 @@ func (rs *RankService) RPC_FindRankDataList(findInfo *rpc.FindRankDataList, find
 	}

 	//查询附带的key
-	if findInfo.Key!= 0 {
+	if findInfo.Key != 0 {
 		findRankData, rank := rankObj.GetRankNodeData(findInfo.Key)
 		if findRankData != nil {
 			findResult.KeyRank = &rpc.RankPosData{}
@@ -159,6 +209,7 @@ func (rs *RankService) RPC_FindRankDataList(findInfo *rpc.FindRankDataList, find
 			findResult.KeyRank.Key = findRankData.Key
 			findResult.KeyRank.SortData = findRankData.SortData
 			findResult.KeyRank.Rank = rank
+			findResult.KeyRank.ExtendData = findRankData.ExData
 		}
 	}

@@ -193,12 +244,12 @@ func (rs *RankService) dealCfg() error {
 		}

 		rankId, okId := mapCfg["RankID"].(float64)
-		if okId == false || uint64(rankId)==0 {
+		if okId == false || uint64(rankId) == 0 {
 			return fmt.Errorf("RankService SortCfg data must has RankID[number]")
 		}

 		rankName, okId := mapCfg["RankName"].(string)
-		if okId == false || len(rankName)==0 {
+		if okId == false || len(rankName) == 0 {
 			return fmt.Errorf("RankService SortCfg data must has RankName[string]")
 		}

@@ -207,11 +258,10 @@ func (rs *RankService) dealCfg() error {
 		maxRank, _ := mapCfg["MaxRank"].(float64)
 		expireMs, _ := mapCfg["ExpireMs"].(float64)

-
-		newSkip := NewRankSkip(uint64(rankId),rankName,isDec, transformLevel(int32(level)), uint64(maxRank),time.Duration(expireMs)*time.Millisecond)
+		newSkip := NewRankSkip(uint64(rankId), rankName, isDec, transformLevel(int32(level)), uint64(maxRank), time.Duration(expireMs)*time.Millisecond)
 		newSkip.SetupRankModule(rs.rankModule)
 		rs.mapRankSkip[uint64(rankId)] = newSkip
-		err := rs.rankModule.OnSetupRank(false,newSkip)
+		err := rs.rankModule.OnSetupRank(false, newSkip)
 		if err != nil {
 			return err
 		}
@@ -219,5 +269,3 @@ func (rs *RankService) dealCfg() error {

 	return nil
 }
-
-
--- a/sysservice/rankservice/RankSkip.go
+++ b/sysservice/rankservice/RankSkip.go
@@ -2,20 +2,21 @@ package rankservice

 import (
 	"fmt"
+	"time"
+
 	"github.com/duanhf2012/origin/rpc"
 	"github.com/duanhf2012/origin/util/algorithms/skip"
-	"time"
 )

 type RankSkip struct {
-	rankId      uint64               //排行榜ID
-	rankName    string               //排行榜名称
-	isDes       bool                 //是否为降序 true：降序 false：升序
-	skipList    *skip.SkipList       //跳表
-	mapRankData map[uint64]*RankData //排行数据map
-	maxLen      uint64               //排行数据长度
-	expireMs    time.Duration        //有效时间
-	rankModule  IRankModule
+	rankId         uint64               //排行榜ID
+	rankName       string               //排行榜名称
+	isDes          bool                 //是否为降序 true：降序 false：升序
+	skipList       *skip.SkipList       //跳表
+	mapRankData    map[uint64]*RankData //排行数据map
+	maxLen         uint64               //排行数据长度
+	expireMs       time.Duration        //有效时间
+	rankModule     IRankModule
 	rankDataExpire rankDataHeap
 }

@@ -28,7 +29,7 @@ const (
 )

 // NewRankSkip 创建排行榜
-func NewRankSkip(rankId uint64,rankName string,isDes bool, level interface{}, maxLen uint64,expireMs time.Duration) *RankSkip {
+func NewRankSkip(rankId uint64, rankName string, isDes bool, level interface{}, maxLen uint64, expireMs time.Duration) *RankSkip {
 	rs := &RankSkip{}

 	rs.rankId = rankId
@@ -38,17 +39,17 @@ func NewRankSkip(rankId uint64,rankName string,isDes bool, level interface{}, ma
 	rs.mapRankData = make(map[uint64]*RankData, 10240)
 	rs.maxLen = maxLen
 	rs.expireMs = expireMs
-	rs.rankDataExpire.Init(int32(maxLen),expireMs)
+	rs.rankDataExpire.Init(int32(maxLen), expireMs)

 	return rs
 }

-func (rs *RankSkip) pickExpireKey(){
+func (rs *RankSkip) pickExpireKey() {
 	if rs.expireMs == 0 {
 		return
 	}

-	for i:=1;i<=MaxPickExpireNum;i++{
+	for i := 1; i <= MaxPickExpireNum; i++ {
 		key := rs.rankDataExpire.PopExpireKey()
 		if key == 0 {
 			return
@@ -79,46 +80,211 @@ func (rs *RankSkip) GetRankLen() uint64 {

 func (rs *RankSkip) UpsetRankList(upsetRankData []*rpc.RankData) (addCount int32, modifyCount int32) {
 	for _, upsetData := range upsetRankData {
-		 changeType := rs.UpsetRank(upsetData,time.Now().UnixNano(),false)
-		 if changeType == RankDataAdd{
-			 addCount+=1
-		 } else if changeType == RankDataUpdate{
-			 modifyCount+=1
-		 }
+		changeType := rs.UpsetRank(upsetData, time.Now().UnixNano(), false)
+		if changeType == RankDataAdd {
+			addCount += 1
+		} else if changeType == RankDataUpdate {
+			modifyCount += 1
+		}
 	}

 	rs.pickExpireKey()
 	return
 }

+func (rs *RankSkip) InsertDataOnNonExistent(changeRankData *rpc.IncreaseRankData) bool {
+	if changeRankData.InsertDataOnNonExistent == false {
+		return false
+	}
+
+	var upsetData rpc.RankData
+	upsetData.Key = changeRankData.Key
+	upsetData.Data = changeRankData.InitData
+	upsetData.SortData = changeRankData.InitSortData
+
+	for i := 0; i < len(changeRankData.IncreaseSortData) && i < len(upsetData.SortData); i++ {
+		upsetData.SortData[i] += changeRankData.IncreaseSortData[i]
+	}
+
+	for _, val := range changeRankData.Extend {
+		upsetData.ExData = append(upsetData.ExData, &rpc.ExtendIncData{InitValue: val.InitValue, IncreaseValue: val.IncreaseValue})
+	}
+
+	//强制设计指定值
+	for _, setData := range changeRankData.SetSortAndExtendData {
+		if setData.IsSortData == true {
+			if int(setData.Pos) >= len(upsetData.SortData) {
+				return false
+			}
+			upsetData.SortData[setData.Pos] = setData.Data
+		} else {
+			if int(setData.Pos) < len(upsetData.ExData) {
+				upsetData.ExData[setData.Pos].IncreaseValue = 0
+				upsetData.ExData[setData.Pos].InitValue = setData.Data
+			}
+		}
+	}
+
+	refreshTimestamp := time.Now().UnixNano()
+	newRankData := NewRankData(rs.isDes, &upsetData, refreshTimestamp)
+	rs.skipList.Insert(newRankData)
+	rs.mapRankData[upsetData.Key] = newRankData
+
+	//刷新有效期和存档数据
+	rs.rankDataExpire.PushOrRefreshExpireKey(upsetData.Key, refreshTimestamp)
+	rs.rankModule.OnChangeRankData(rs, newRankData)
+
+	return true
+}
+
+func (rs *RankSkip) UpdateRankData(updateRankData *rpc.UpdateRankData) bool {
+	rankNode, ok := rs.mapRankData[updateRankData.Key]
+	if ok == false {
+		return false
+	}
+
+	rankNode.Data = updateRankData.Data
+	rs.rankDataExpire.PushOrRefreshExpireKey(updateRankData.Key, time.Now().UnixNano())
+	rs.rankModule.OnChangeRankData(rs, rankNode)
+	return true
+}
+
+func (rs *RankSkip) ChangeExtendData(changeRankData *rpc.IncreaseRankData) bool {
+	rankNode, ok := rs.mapRankData[changeRankData.Key]
+	if ok == false {
+		return rs.InsertDataOnNonExistent(changeRankData)
+	}
+
+	//先判断是不是有修改
+	bChange := false
+	for i := 0; i < len(changeRankData.IncreaseSortData) && i < len(rankNode.SortData); i++ {
+		if changeRankData.IncreaseSortData[i] != 0 {
+			bChange = true
+		}
+	}
+
+	if bChange == false {
+		for _, setSortAndExtendData := range changeRankData.SetSortAndExtendData {
+			if setSortAndExtendData.IsSortData == true {
+				bChange = true
+			}
+		}
+	}
+
+	//如果有改变，删除原有的数据,重新刷新到跳表
+	rankData := rankNode
+	refreshTimestamp := time.Now().UnixNano()
+	if bChange == true {
+		//copy数据
+		var upsetData rpc.RankData
+		upsetData.Key = rankNode.Key
+		upsetData.Data = rankNode.Data
+		upsetData.SortData = append(upsetData.SortData, rankNode.SortData...)
+
+		for i := 0; i < len(changeRankData.IncreaseSortData) && i < len(upsetData.SortData); i++ {
+			if changeRankData.IncreaseSortData[i] != 0 {
+				upsetData.SortData[i] += changeRankData.IncreaseSortData[i]
+			}
+		}
+
+		for _, setData := range changeRankData.SetSortAndExtendData {
+			if setData.IsSortData == true {
+				if int(setData.Pos) < len(upsetData.SortData) {
+					upsetData.SortData[setData.Pos] = setData.Data
+				}
+			}
+		}
+
+		rankData = NewRankData(rs.isDes, &upsetData, refreshTimestamp)
+		rankData.ExData = append(rankData.ExData, rankNode.ExData...)
+
+		//从排行榜中删除
+		rs.skipList.Delete(rankNode)
+		ReleaseRankData(rankNode)
+
+		rs.skipList.Insert(rankData)
+		rs.mapRankData[upsetData.Key] = rankData
+	}
+
+	//增长扩展参数
+	for i := 0; i < len(changeRankData.Extend); i++ {
+		if i < len(rankData.ExData) {
+			//直接增长
+			rankData.ExData[i] += changeRankData.Extend[i].IncreaseValue
+		} else {
+			//如果不存在的扩展位置,append补充，并按IncreaseValue增长
+			rankData.ExData = append(rankData.ExData, changeRankData.Extend[i].InitValue+changeRankData.Extend[i].IncreaseValue)
+		}
+	}
+
+	//设置固定值
+	for _, setData := range changeRankData.SetSortAndExtendData {
+		if setData.IsSortData == false {
+			if int(setData.Pos) < len(rankData.ExData) {
+				rankData.ExData[setData.Pos] = setData.Data
+			}
+		}
+	}
+
+	rs.rankDataExpire.PushOrRefreshExpireKey(rankData.Key, refreshTimestamp)
+	rs.rankModule.OnChangeRankData(rs, rankData)
+
+	return true
+}
+
 // UpsetRank 更新玩家排行数据,返回变化后的数据及变化类型
-func (rs *RankSkip) UpsetRank(upsetData *rpc.RankData,refreshTimestamp int64,fromLoad bool)  RankDataChangeType {
+func (rs *RankSkip) UpsetRank(upsetData *rpc.RankData, refreshTimestamp int64, fromLoad bool) RankDataChangeType {
 	rankNode, ok := rs.mapRankData[upsetData.Key]
 	if ok == true {
+		//增长扩展数据
+		for i := 0; i < len(upsetData.ExData); i++ {
+			if i < len(rankNode.ExData) {
+				//直接增长
+				rankNode.ExData[i] += upsetData.ExData[i].IncreaseValue
+			} else {
+				//如果不存在的扩展位置,append补充，并按IncreaseValue增长
+				rankNode.ExData = append(rankNode.ExData, upsetData.ExData[i].InitValue+upsetData.ExData[i].IncreaseValue)
+			}
+		}
+
 		//找到的情况对比排名数据是否有变化,无变化进行data更新,有变化则进行删除更新
 		if compareIsEqual(rankNode.SortData, upsetData.SortData) {
 			rankNode.Data = upsetData.GetData()
 			rankNode.refreshTimestamp = refreshTimestamp

 			if fromLoad == false {
-				rs.rankModule.OnChangeRankData(rs,rankNode)
+				rs.rankModule.OnChangeRankData(rs, rankNode)
 			}
-			rs.rankDataExpire.PushOrRefreshExpireKey(upsetData.Key,refreshTimestamp)
+			rs.rankDataExpire.PushOrRefreshExpireKey(upsetData.Key, refreshTimestamp)
 			return RankDataUpdate
 		}

 		if upsetData.Data == nil {
 			upsetData.Data = rankNode.Data
 		}
+
+		//设置额外数据
+		for idx, exValue := range rankNode.ExData {
+			currentIncreaseValue := int64(0)
+			if idx < len(upsetData.ExData) {
+				currentIncreaseValue = upsetData.ExData[idx].IncreaseValue
+			}
+
+			upsetData.ExData = append(upsetData.ExData, &rpc.ExtendIncData{
+				InitValue:     exValue,
+				IncreaseValue: currentIncreaseValue,
+			})
+		}
+
 		rs.skipList.Delete(rankNode)
 		ReleaseRankData(rankNode)

-		newRankData := NewRankData(rs.isDes, upsetData,refreshTimestamp)
+		newRankData := NewRankData(rs.isDes, upsetData, refreshTimestamp)
 		rs.skipList.Insert(newRankData)
 		rs.mapRankData[upsetData.Key] = newRankData

 		//刷新有效期
-		rs.rankDataExpire.PushOrRefreshExpireKey(upsetData.Key,refreshTimestamp)
+		rs.rankDataExpire.PushOrRefreshExpireKey(upsetData.Key, refreshTimestamp)

 		if fromLoad == false {
 			rs.rankModule.OnChangeRankData(rs, newRankData)
@@ -127,10 +293,11 @@ func (rs *RankSkip) UpsetRank(upsetData *rpc.RankData,refreshTimestamp int64,fro
 	}

 	if rs.checkInsertAndReplace(upsetData) {
-		newRankData := NewRankData(rs.isDes, upsetData,refreshTimestamp)
+		newRankData := NewRankData(rs.isDes, upsetData, refreshTimestamp)
+
 		rs.skipList.Insert(newRankData)
 		rs.mapRankData[upsetData.Key] = newRankData
-		rs.rankDataExpire.PushOrRefreshExpireKey(upsetData.Key,refreshTimestamp)
+		rs.rankDataExpire.PushOrRefreshExpireKey(upsetData.Key, refreshTimestamp)

 		if fromLoad == false {
 			rs.rankModule.OnEnterRank(rs, newRankData)
@@ -152,7 +319,7 @@ func (rs *RankSkip) DeleteRankData(delKeys []uint64) int32 {
 			continue
 		}

-		removeRankData+=1
+		removeRankData += 1
 		rs.skipList.Delete(rankData)
 		delete(rs.mapRankData, rankData.Key)
 		rs.rankDataExpire.RemoveExpireKey(rankData.Key)
@@ -172,13 +339,13 @@ func (rs *RankSkip) GetRankNodeData(findKey uint64) (*RankData, uint64) {

 	rs.pickExpireKey()
 	_, index := rs.skipList.GetWithPosition(rankNode)
-	return rankNode, index+1
+	return rankNode, index + 1
 }

 // GetRankNodeDataByPos 获取,返回排名节点与名次
 func (rs *RankSkip) GetRankNodeDataByRank(rank uint64) (*RankData, uint64) {
 	rs.pickExpireKey()
-	rankNode := rs.skipList.ByPosition(rank-1)
+	rankNode := rs.skipList.ByPosition(rank - 1)
 	if rankNode == nil {
 		return nil, 0
 	}
@@ -189,12 +356,12 @@ func (rs *RankSkip) GetRankNodeDataByRank(rank uint64) (*RankData, uint64) {
 // GetRankKeyPrevToLimit 获取key前count名的数据
 func (rs *RankSkip) GetRankKeyPrevToLimit(findKey, count uint64, result *rpc.RankDataList) error {
 	if rs.GetRankLen() <= 0 {
-		return fmt.Errorf("rank[", rs.rankId, "] no data")
+		return fmt.Errorf("rank[%d] no data", rs.rankId)
 	}

 	findData, ok := rs.mapRankData[findKey]
 	if ok == false {
-		return fmt.Errorf("rank[", rs.rankId, "] no data")
+		return fmt.Errorf("rank[%d] no data", rs.rankId)
 	}

 	_, rankPos := rs.skipList.GetWithPosition(findData)
@@ -203,10 +370,11 @@ func (rs *RankSkip) GetRankKeyPrevToLimit(findKey, count uint64, result *rpc.Ran
 	for iter.Prev() && iterCount < count {
 		rankData := iter.Value().(*RankData)
 		result.RankPosDataList = append(result.RankPosDataList, &rpc.RankPosData{
-			Key:      rankData.Key,
-			Rank:  rankPos - iterCount+1,
-			SortData: rankData.SortData,
-			Data:     rankData.Data,
+			Key:        rankData.Key,
+			Rank:       rankPos - iterCount + 1,
+			SortData:   rankData.SortData,
+			Data:       rankData.Data,
+			ExtendData: rankData.ExData,
 		})
 		iterCount++
 	}
@@ -217,12 +385,12 @@ func (rs *RankSkip) GetRankKeyPrevToLimit(findKey, count uint64, result *rpc.Ran
 // GetRankKeyPrevToLimit 获取key前count名的数据
 func (rs *RankSkip) GetRankKeyNextToLimit(findKey, count uint64, result *rpc.RankDataList) error {
 	if rs.GetRankLen() <= 0 {
-		return fmt.Errorf("rank[", rs.rankId, "] no data")
+		return fmt.Errorf("rank[%d] no data", rs.rankId)
 	}

 	findData, ok := rs.mapRankData[findKey]
 	if ok == false {
-		return fmt.Errorf("rank[", rs.rankId, "] no data")
+		return fmt.Errorf("rank[%d] no data", rs.rankId)
 	}

 	_, rankPos := rs.skipList.GetWithPosition(findData)
@@ -231,10 +399,11 @@ func (rs *RankSkip) GetRankKeyNextToLimit(findKey, count uint64, result *rpc.Ran
 	for iter.Next() && iterCount < count {
 		rankData := iter.Value().(*RankData)
 		result.RankPosDataList = append(result.RankPosDataList, &rpc.RankPosData{
-			Key:      rankData.Key,
-			Rank:  rankPos + iterCount+1,
-			SortData: rankData.SortData,
-			Data:     rankData.Data,
+			Key:        rankData.Key,
+			Rank:       rankPos + iterCount + 1,
+			SortData:   rankData.SortData,
+			Data:       rankData.Data,
+			ExtendData: rankData.ExData,
 		})
 		iterCount++
 	}
@@ -259,10 +428,11 @@ func (rs *RankSkip) GetRankDataFromToLimit(startPos, count uint64, result *rpc.R
 	for iter.Next() && iterCount < count {
 		rankData := iter.Value().(*RankData)
 		result.RankPosDataList = append(result.RankPosDataList, &rpc.RankPosData{
-			Key:      rankData.Key,
-			Rank:  iterCount + startPos+1,
-			SortData: rankData.SortData,
-			Data:     rankData.Data,
+			Key:        rankData.Key,
+			Rank:       iterCount + startPos + 1,
+			SortData:   rankData.SortData,
+			Data:       rankData.Data,
+			ExtendData: rankData.ExData,
 		})
 		iterCount++
 	}
@@ -301,4 +471,3 @@ func (rs *RankSkip) checkInsertAndReplace(upsetData *rpc.RankData) bool {
 	ReleaseRankData(lastRankData)
 	return true
 }
-
--- a/sysservice/tcpservice/tcpservice.go
+++ b/sysservice/tcpservice/tcpservice.go
@@ -90,6 +90,10 @@ func (tcpService *TcpService) OnInit() error{
 	if ok == true {
 		tcpService.tcpServer.LittleEndian = LittleEndian.(bool)
 	}
+	LenMsgLen,ok := tcpCfg["LenMsgLen"]
+	if ok == true {
+		tcpService.tcpServer.LenMsgLen = int(LenMsgLen.(float64))
+	}
 	MinMsgLen,ok := tcpCfg["MinMsgLen"]
 	if ok == true {
 		tcpService.tcpServer.MinMsgLen = uint32(MinMsgLen.(float64))
--- a/util/queue/deque.go
+++ b/util/queue/deque.go
@@ -0,0 +1,413 @@
+package queue
+
+// minCapacity is the smallest capacity that deque may have. Must be power of 2
+// for bitwise modulus: x % n == x & (n - 1).
+const minCapacity = 16
+
+// Deque represents a single instance of the deque data structure. A Deque
+// instance contains items of the type sepcified by the type argument.
+type Deque[T any] struct {
+	buf    []T
+	head   int
+	tail   int
+	count  int
+	minCap int
+}
+
+// New creates a new Deque, optionally setting the current and minimum capacity
+// when non-zero values are given for these. The Deque instance returns
+// operates on items of the type specified by the type argument. For example,
+// to create a Deque that contains strings,
+//
+//   stringDeque := deque.New[string]()
+//
+// To create a Deque with capacity to store 2048 ints without resizing, and
+// that will not resize below space for 32 items when removing items:
+//   d := deque.New[int](2048, 32)
+//
+// To create a Deque that has not yet allocated memory, but after it does will
+// never resize to have space for less than 64 items:
+//   d := deque.New[int](0, 64)
+//
+// Any size values supplied here are rounded up to the nearest power of 2.
+func New[T any](size ...int) *Deque[T] {
+	var capacity, minimum int
+	if len(size) >= 1 {
+		capacity = size[0]
+		if len(size) >= 2 {
+			minimum = size[1]
+		}
+	}
+
+	minCap := minCapacity
+	for minCap < minimum {
+		minCap <<= 1
+	}
+
+	var buf []T
+	if capacity != 0 {
+		bufSize := minCap
+		for bufSize < capacity {
+			bufSize <<= 1
+		}
+		buf = make([]T, bufSize)
+	}
+
+	return &Deque[T]{
+		buf:    buf,
+		minCap: minCap,
+	}
+}
+
+// Cap returns the current capacity of the Deque. If q is nil, q.Cap() is zero.
+func (q *Deque[T]) Cap() int {
+	if q == nil {
+		return 0
+	}
+	return len(q.buf)
+}
+
+// Len returns the number of elements currently stored in the queue. If q is
+// nil, q.Len() is zero.
+func (q *Deque[T]) Len() int {
+	if q == nil {
+		return 0
+	}
+	return q.count
+}
+
+// PushBack appends an element to the back of the queue. Implements FIFO when
+// elements are removed with PopFront(), and LIFO when elements are removed
+// with PopBack().
+func (q *Deque[T]) PushBack(elem T) {
+	q.growIfFull()
+
+	q.buf[q.tail] = elem
+	// Calculate new tail position.
+	q.tail = q.next(q.tail)
+	q.count++
+}
+
+// PushFront prepends an element to the front of the queue.
+func (q *Deque[T]) PushFront(elem T) {
+	q.growIfFull()
+
+	// Calculate new head position.
+	q.head = q.prev(q.head)
+	q.buf[q.head] = elem
+	q.count++
+}
+
+// PopFront removes and returns the element from the front of the queue.
+// Implements FIFO when used with PushBack(). If the queue is empty, the call
+// panics.
+func (q *Deque[T]) PopFront() T {
+	if q.count <= 0 {
+		panic("deque: PopFront() called on empty queue")
+	}
+	ret := q.buf[q.head]
+	var zero T
+	q.buf[q.head] = zero
+	// Calculate new head position.
+	q.head = q.next(q.head)
+	q.count--
+
+	q.shrinkIfExcess()
+	return ret
+}
+
+// PopBack removes and returns the element from the back of the queue.
+// Implements LIFO when used with PushBack(). If the queue is empty, the call
+// panics.
+func (q *Deque[T]) PopBack() T {
+	if q.count <= 0 {
+		panic("deque: PopBack() called on empty queue")
+	}
+
+	// Calculate new tail position
+	q.tail = q.prev(q.tail)
+
+	// Remove value at tail.
+	ret := q.buf[q.tail]
+	var zero T
+	q.buf[q.tail] = zero
+	q.count--
+
+	q.shrinkIfExcess()
+	return ret
+}
+
+// Front returns the element at the front of the queue. This is the element
+// that would be returned by PopFront(). This call panics if the queue is
+// empty.
+func (q *Deque[T]) Front() T {
+	if q.count <= 0 {
+		panic("deque: Front() called when empty")
+	}
+	return q.buf[q.head]
+}
+
+// Back returns the element at the back of the queue. This is the element that
+// would be returned by PopBack(). This call panics if the queue is empty.
+func (q *Deque[T]) Back() T {
+	if q.count <= 0 {
+		panic("deque: Back() called when empty")
+	}
+	return q.buf[q.prev(q.tail)]
+}
+
+// At returns the element at index i in the queue without removing the element
+// from the queue. This method accepts only non-negative index values. At(0)
+// refers to the first element and is the same as Front(). At(Len()-1) refers
+// to the last element and is the same as Back(). If the index is invalid, the
+// call panics.
+//
+// The purpose of At is to allow Deque to serve as a more general purpose
+// circular buffer, where items are only added to and removed from the ends of
+// the deque, but may be read from any place within the deque. Consider the
+// case of a fixed-size circular log buffer: A new entry is pushed onto one end
+// and when full the oldest is popped from the other end. All the log entries
+// in the buffer must be readable without altering the buffer contents.
+func (q *Deque[T]) At(i int) T {
+	if i < 0 || i >= q.count {
+		panic("deque: At() called with index out of range")
+	}
+	// bitwise modulus
+	return q.buf[(q.head+i)&(len(q.buf)-1)]
+}
+
+// Set puts the element at index i in the queue. Set shares the same purpose
+// than At() but perform the opposite operation. The index i is the same index
+// defined by At(). If the index is invalid, the call panics.
+func (q *Deque[T]) Set(i int, elem T) {
+	if i < 0 || i >= q.count {
+		panic("deque: Set() called with index out of range")
+	}
+	// bitwise modulus
+	q.buf[(q.head+i)&(len(q.buf)-1)] = elem
+}
+
+// Clear removes all elements from the queue, but retains the current capacity.
+// This is useful when repeatedly reusing the queue at high frequency to avoid
+// GC during reuse. The queue will not be resized smaller as long as items are
+// only added. Only when items are removed is the queue subject to getting
+// resized smaller.
+func (q *Deque[T]) Clear() {
+	// bitwise modulus
+	modBits := len(q.buf) - 1
+	var zero T
+	for h := q.head; h != q.tail; h = (h + 1) & modBits {
+		q.buf[h] = zero
+	}
+	q.head = 0
+	q.tail = 0
+	q.count = 0
+}
+
+// Rotate rotates the deque n steps front-to-back. If n is negative, rotates
+// back-to-front. Having Deque provide Rotate() avoids resizing that could
+// happen if implementing rotation using only Pop and Push methods. If q.Len()
+// is one or less, or q is nil, then Rotate does nothing.
+func (q *Deque[T]) Rotate(n int) {
+	if q.Len() <= 1 {
+		return
+	}
+	// Rotating a multiple of q.count is same as no rotation.
+	n %= q.count
+	if n == 0 {
+		return
+	}
+
+	modBits := len(q.buf) - 1
+	// If no empty space in buffer, only move head and tail indexes.
+	if q.head == q.tail {
+		// Calculate new head and tail using bitwise modulus.
+		q.head = (q.head + n) & modBits
+		q.tail = q.head
+		return
+	}
+
+	var zero T
+
+	if n < 0 {
+		// Rotate back to front.
+		for ; n < 0; n++ {
+			// Calculate new head and tail using bitwise modulus.
+			q.head = (q.head - 1) & modBits
+			q.tail = (q.tail - 1) & modBits
+			// Put tail value at head and remove value at tail.
+			q.buf[q.head] = q.buf[q.tail]
+			q.buf[q.tail] = zero
+		}
+		return
+	}
+
+	// Rotate front to back.
+	for ; n > 0; n-- {
+		// Put head value at tail and remove value at head.
+		q.buf[q.tail] = q.buf[q.head]
+		q.buf[q.head] = zero
+		// Calculate new head and tail using bitwise modulus.
+		q.head = (q.head + 1) & modBits
+		q.tail = (q.tail + 1) & modBits
+	}
+}
+
+// Index returns the index into the Deque of the first item satisfying f(item),
+// or -1 if none do. If q is nil, then -1 is always returned. Search is linear
+// starting with index 0.
+func (q *Deque[T]) Index(f func(T) bool) int {
+	if q.Len() > 0 {
+		modBits := len(q.buf) - 1
+		for i := 0; i < q.count; i++ {
+			if f(q.buf[(q.head+i)&modBits]) {
+				return i
+			}
+		}
+	}
+	return -1
+}
+
+// RIndex is the same as Index, but searches from Back to Front. The index
+// returned is from Front to Back, where index 0 is the index of the item
+// returned by Front().
+func (q *Deque[T]) RIndex(f func(T) bool) int {
+	if q.Len() > 0 {
+		modBits := len(q.buf) - 1
+		for i := q.count - 1; i >= 0; i-- {
+			if f(q.buf[(q.head+i)&modBits]) {
+				return i
+			}
+		}
+	}
+	return -1
+}
+
+// Insert is used to insert an element into the middle of the queue, before the
+// element at the specified index. Insert(0,e) is the same as PushFront(e) and
+// Insert(Len(),e) is the same as PushBack(e). Accepts only non-negative index
+// values, and panics if index is out of range.
+//
+// Important: Deque is optimized for O(1) operations at the ends of the queue,
+// not for operations in the the middle. Complexity of this function is
+// constant plus linear in the lesser of the distances between the index and
+// either of the ends of the queue.
+func (q *Deque[T]) Insert(at int, item T) {
+	if at < 0 || at > q.count {
+		panic("deque: Insert() called with index out of range")
+	}
+	if at*2 < q.count {
+		q.PushFront(item)
+		front := q.head
+		for i := 0; i < at; i++ {
+			next := q.next(front)
+			q.buf[front], q.buf[next] = q.buf[next], q.buf[front]
+			front = next
+		}
+		return
+	}
+	swaps := q.count - at
+	q.PushBack(item)
+	back := q.prev(q.tail)
+	for i := 0; i < swaps; i++ {
+		prev := q.prev(back)
+		q.buf[back], q.buf[prev] = q.buf[prev], q.buf[back]
+		back = prev
+	}
+}
+
+// Remove removes and returns an element from the middle of the queue, at the
+// specified index. Remove(0) is the same as PopFront() and Remove(Len()-1) is
+// the same as PopBack(). Accepts only non-negative index values, and panics if
+// index is out of range.
+//
+// Important: Deque is optimized for O(1) operations at the ends of the queue,
+// not for operations in the the middle. Complexity of this function is
+// constant plus linear in the lesser of the distances between the index and
+// either of the ends of the queue.
+func (q *Deque[T]) Remove(at int) T {
+	if at < 0 || at >= q.Len() {
+		panic("deque: Remove() called with index out of range")
+	}
+
+	rm := (q.head + at) & (len(q.buf) - 1)
+	if at*2 < q.count {
+		for i := 0; i < at; i++ {
+			prev := q.prev(rm)
+			q.buf[prev], q.buf[rm] = q.buf[rm], q.buf[prev]
+			rm = prev
+		}
+		return q.PopFront()
+	}
+	swaps := q.count - at - 1
+	for i := 0; i < swaps; i++ {
+		next := q.next(rm)
+		q.buf[rm], q.buf[next] = q.buf[next], q.buf[rm]
+		rm = next
+	}
+	return q.PopBack()
+}
+
+// SetMinCapacity sets a minimum capacity of 2^minCapacityExp. If the value of
+// the minimum capacity is less than or equal to the minimum allowed, then
+// capacity is set to the minimum allowed. This may be called at anytime to set
+// a new minimum capacity.
+//
+// Setting a larger minimum capacity may be used to prevent resizing when the
+// number of stored items changes frequently across a wide range.
+func (q *Deque[T]) SetMinCapacity(minCapacityExp uint) {
+	if 1<<minCapacityExp > minCapacity {
+		q.minCap = 1 << minCapacityExp
+	} else {
+		q.minCap = minCapacity
+	}
+}
+
+// prev returns the previous buffer position wrapping around buffer.
+func (q *Deque[T]) prev(i int) int {
+	return (i - 1) & (len(q.buf) - 1) // bitwise modulus
+}
+
+// next returns the next buffer position wrapping around buffer.
+func (q *Deque[T]) next(i int) int {
+	return (i + 1) & (len(q.buf) - 1) // bitwise modulus
+}
+
+// growIfFull resizes up if the buffer is full.
+func (q *Deque[T]) growIfFull() {
+	if q.count != len(q.buf) {
+		return
+	}
+	if len(q.buf) == 0 {
+		if q.minCap == 0 {
+			q.minCap = minCapacity
+		}
+		q.buf = make([]T, q.minCap)
+		return
+	}
+	q.resize()
+}
+
+// shrinkIfExcess resize down if the buffer 1/4 full.
+func (q *Deque[T]) shrinkIfExcess() {
+	if len(q.buf) > q.minCap && (q.count<<2) == len(q.buf) {
+		q.resize()
+	}
+}
+
+// resize resizes the deque to fit exactly twice its current contents. This is
+// used to grow the queue when it is full, and also to shrink it when it is
+// only a quarter full.
+func (q *Deque[T]) resize() {
+	newBuf := make([]T, q.count<<1)
+	if q.tail > q.head {
+		copy(newBuf, q.buf[q.head:q.tail])
+	} else {
+		n := copy(newBuf, q.buf[q.head:])
+		copy(newBuf[n:], q.buf[:q.tail])
+	}
+
+	q.head = 0
+	q.tail = q.count
+	q.buf = newBuf
+}
--- a/util/queue/deque_test.go
+++ b/util/queue/deque_test.go
@@ -0,0 +1,836 @@
+package queue
+
+import (
+	"fmt"
+	"testing"
+	"unicode"
+)
+
+func TestEmpty(t *testing.T) {
+	q := New[string]()
+	if q.Len() != 0 {
+		t.Error("q.Len() =", q.Len(), "expect 0")
+	}
+	if q.Cap() != 0 {
+		t.Error("expected q.Cap() == 0")
+	}
+	idx := q.Index(func(item string) bool {
+		return true
+	})
+	if idx != -1 {
+		t.Error("should return -1 index for nil deque")
+	}
+	idx = q.RIndex(func(item string) bool {
+		return true
+	})
+	if idx != -1 {
+		t.Error("should return -1 index for nil deque")
+	}
+}
+
+func TestNil(t *testing.T) {
+	var q *Deque[int]
+	if q.Len() != 0 {
+		t.Error("expected q.Len() == 0")
+	}
+	if q.Cap() != 0 {
+		t.Error("expected q.Cap() == 0")
+	}
+	q.Rotate(5)
+	idx := q.Index(func(item int) bool {
+		return true
+	})
+	if idx != -1 {
+		t.Error("should return -1 index for nil deque")
+	}
+	idx = q.RIndex(func(item int) bool {
+		return true
+	})
+	if idx != -1 {
+		t.Error("should return -1 index for nil deque")
+	}
+}
+
+func TestFrontBack(t *testing.T) {
+	var q Deque[string]
+	q.PushBack("foo")
+	q.PushBack("bar")
+	q.PushBack("baz")
+	if q.Front() != "foo" {
+		t.Error("wrong value at front of queue")
+	}
+	if q.Back() != "baz" {
+		t.Error("wrong value at back of queue")
+	}
+
+	if q.PopFront() != "foo" {
+		t.Error("wrong value removed from front of queue")
+	}
+	if q.Front() != "bar" {
+		t.Error("wrong value remaining at front of queue")
+	}
+	if q.Back() != "baz" {
+		t.Error("wrong value remaining at back of queue")
+	}
+
+	if q.PopBack() != "baz" {
+		t.Error("wrong value removed from back of queue")
+	}
+	if q.Front() != "bar" {
+		t.Error("wrong value remaining at front of queue")
+	}
+	if q.Back() != "bar" {
+		t.Error("wrong value remaining at back of queue")
+	}
+}
+
+func TestGrowShrinkBack(t *testing.T) {
+	var q Deque[int]
+	size := minCapacity * 2
+
+	for i := 0; i < size; i++ {
+		if q.Len() != i {
+			t.Error("q.Len() =", q.Len(), "expected", i)
+		}
+		q.PushBack(i)
+	}
+	bufLen := len(q.buf)
+
+	// Remove from back.
+	for i := size; i > 0; i-- {
+		if q.Len() != i {
+			t.Error("q.Len() =", q.Len(), "expected", i)
+		}
+		x := q.PopBack()
+		if x != i-1 {
+			t.Error("q.PopBack() =", x, "expected", i-1)
+		}
+	}
+	if q.Len() != 0 {
+		t.Error("q.Len() =", q.Len(), "expected 0")
+	}
+	if len(q.buf) == bufLen {
+		t.Error("queue buffer did not shrink")
+	}
+}
+
+func TestGrowShrinkFront(t *testing.T) {
+	var q Deque[int]
+	size := minCapacity * 2
+
+	for i := 0; i < size; i++ {
+		if q.Len() != i {
+			t.Error("q.Len() =", q.Len(), "expected", i)
+		}
+		q.PushBack(i)
+	}
+	bufLen := len(q.buf)
+
+	// Remove from Front
+	for i := 0; i < size; i++ {
+		if q.Len() != size-i {
+			t.Error("q.Len() =", q.Len(), "expected", minCapacity*2-i)
+		}
+		x := q.PopFront()
+		if x != i {
+			t.Error("q.PopBack() =", x, "expected", i)
+		}
+	}
+	if q.Len() != 0 {
+		t.Error("q.Len() =", q.Len(), "expected 0")
+	}
+	if len(q.buf) == bufLen {
+		t.Error("queue buffer did not shrink")
+	}
+}
+
+func TestSimple(t *testing.T) {
+	var q Deque[int]
+
+	for i := 0; i < minCapacity; i++ {
+		q.PushBack(i)
+	}
+	if q.Front() != 0 {
+		t.Fatalf("expected 0 at front, got %d", q.Front())
+	}
+	if q.Back() != minCapacity-1 {
+		t.Fatalf("expected %d at back, got %d", minCapacity-1, q.Back())
+	}
+
+	for i := 0; i < minCapacity; i++ {
+		if q.Front() != i {
+			t.Error("peek", i, "had value", q.Front())
+		}
+		x := q.PopFront()
+		if x != i {
+			t.Error("remove", i, "had value", x)
+		}
+	}
+
+	q.Clear()
+	for i := 0; i < minCapacity; i++ {
+		q.PushFront(i)
+	}
+	for i := minCapacity - 1; i >= 0; i-- {
+		x := q.PopFront()
+		if x != i {
+			t.Error("remove", i, "had value", x)
+		}
+	}
+}
+
+func TestBufferWrap(t *testing.T) {
+	var q Deque[int]
+
+	for i := 0; i < minCapacity; i++ {
+		q.PushBack(i)
+	}
+
+	for i := 0; i < 3; i++ {
+		q.PopFront()
+		q.PushBack(minCapacity + i)
+	}
+
+	for i := 0; i < minCapacity; i++ {
+		if q.Front() != i+3 {
+			t.Error("peek", i, "had value", q.Front())
+		}
+		q.PopFront()
+	}
+}
+
+func TestBufferWrapReverse(t *testing.T) {
+	var q Deque[int]
+
+	for i := 0; i < minCapacity; i++ {
+		q.PushFront(i)
+	}
+	for i := 0; i < 3; i++ {
+		q.PopBack()
+		q.PushFront(minCapacity + i)
+	}
+
+	for i := 0; i < minCapacity; i++ {
+		if q.Back() != i+3 {
+			t.Error("peek", i, "had value", q.Front())
+		}
+		q.PopBack()
+	}
+}
+
+func TestLen(t *testing.T) {
+	var q Deque[int]
+
+	if q.Len() != 0 {
+		t.Error("empty queue length not 0")
+	}
+
+	for i := 0; i < 1000; i++ {
+		q.PushBack(i)
+		if q.Len() != i+1 {
+			t.Error("adding: queue with", i, "elements has length", q.Len())
+		}
+	}
+	for i := 0; i < 1000; i++ {
+		q.PopFront()
+		if q.Len() != 1000-i-1 {
+			t.Error("removing: queue with", 1000-i-i, "elements has length", q.Len())
+		}
+	}
+}
+
+func TestBack(t *testing.T) {
+	var q Deque[int]
+
+	for i := 0; i < minCapacity+5; i++ {
+		q.PushBack(i)
+		if q.Back() != i {
+			t.Errorf("Back returned wrong value")
+		}
+	}
+}
+
+func TestNew(t *testing.T) {
+	minCap := 64
+	q := New[string](0, minCap)
+	if q.Cap() != 0 {
+		t.Fatal("should not have allowcated mem yet")
+	}
+	q.PushBack("foo")
+	q.PopFront()
+	if q.Len() != 0 {
+		t.Fatal("Len() should return 0")
+	}
+	if q.Cap() != minCap {
+		t.Fatalf("worng capactiy expected %d, got %d", minCap, q.Cap())
+	}
+
+	curCap := 128
+	q = New[string](curCap, minCap)
+	if q.Cap() != curCap {
+		t.Fatalf("Cap() should return %d, got %d", curCap, q.Cap())
+	}
+	if q.Len() != 0 {
+		t.Fatalf("Len() should return 0")
+	}
+	q.PushBack("foo")
+	if q.Cap() != curCap {
+		t.Fatalf("Cap() should return %d, got %d", curCap, q.Cap())
+	}
+}
+
+func checkRotate(t *testing.T, size int) {
+	var q Deque[int]
+	for i := 0; i < size; i++ {
+		q.PushBack(i)
+	}
+
+	for i := 0; i < q.Len(); i++ {
+		x := i
+		for n := 0; n < q.Len(); n++ {
+			if q.At(n) != x {
+				t.Fatalf("a[%d] != %d after rotate and copy", n, x)
+			}
+			x++
+			if x == q.Len() {
+				x = 0
+			}
+		}
+		q.Rotate(1)
+		if q.Back() != i {
+			t.Fatal("wrong value during rotation")
+		}
+	}
+	for i := q.Len() - 1; i >= 0; i-- {
+		q.Rotate(-1)
+		if q.Front() != i {
+			t.Fatal("wrong value during reverse rotation")
+		}
+	}
+}
+
+func TestRotate(t *testing.T) {
+	checkRotate(t, 10)
+	checkRotate(t, minCapacity)
+	checkRotate(t, minCapacity+minCapacity/2)
+
+	var q Deque[int]
+	for i := 0; i < 10; i++ {
+		q.PushBack(i)
+	}
+	q.Rotate(11)
+	if q.Front() != 1 {
+		t.Error("rotating 11 places should have been same as one")
+	}
+	q.Rotate(-21)
+	if q.Front() != 0 {
+		t.Error("rotating -21 places should have been same as one -1")
+	}
+	q.Rotate(q.Len())
+	if q.Front() != 0 {
+		t.Error("should not have rotated")
+	}
+	q.Clear()
+	q.PushBack(0)
+	q.Rotate(13)
+	if q.Front() != 0 {
+		t.Error("should not have rotated")
+	}
+}
+
+func TestAt(t *testing.T) {
+	var q Deque[int]
+
+	for i := 0; i < 1000; i++ {
+		q.PushBack(i)
+	}
+
+	// Front to back.
+	for j := 0; j < q.Len(); j++ {
+		if q.At(j) != j {
+			t.Errorf("index %d doesn't contain %d", j, j)
+		}
+	}
+
+	// Back to front
+	for j := 1; j <= q.Len(); j++ {
+		if q.At(q.Len()-j) != q.Len()-j {
+			t.Errorf("index %d doesn't contain %d", q.Len()-j, q.Len()-j)
+		}
+	}
+}
+
+func TestSet(t *testing.T) {
+	var q Deque[int]
+
+	for i := 0; i < 1000; i++ {
+		q.PushBack(i)
+		q.Set(i, i+50)
+	}
+
+	// Front to back.
+	for j := 0; j < q.Len(); j++ {
+		if q.At(j) != j+50 {
+			t.Errorf("index %d doesn't contain %d", j, j+50)
+		}
+	}
+}
+
+func TestClear(t *testing.T) {
+	var q Deque[int]
+
+	for i := 0; i < 100; i++ {
+		q.PushBack(i)
+	}
+	if q.Len() != 100 {
+		t.Error("push: queue with 100 elements has length", q.Len())
+	}
+	cap := len(q.buf)
+	q.Clear()
+	if q.Len() != 0 {
+		t.Error("empty queue length not 0 after clear")
+	}
+	if len(q.buf) != cap {
+		t.Error("queue capacity changed after clear")
+	}
+
+	// Check that there are no remaining references after Clear()
+	for i := 0; i < len(q.buf); i++ {
+		if q.buf[i] != 0 {
+			t.Error("queue has non-nil deleted elements after Clear()")
+			break
+		}
+	}
+}
+
+func TestIndex(t *testing.T) {
+	var q Deque[rune]
+	for _, x := range "Hello, 世界" {
+		q.PushBack(x)
+	}
+	idx := q.Index(func(item rune) bool {
+		c := item
+		return unicode.Is(unicode.Han, c)
+	})
+	if idx != 7 {
+		t.Fatal("Expected index 7, got", idx)
+	}
+	idx = q.Index(func(item rune) bool {
+		c := item
+		return c == 'H'
+	})
+	if idx != 0 {
+		t.Fatal("Expected index 0, got", idx)
+	}
+	idx = q.Index(func(item rune) bool {
+		return false
+	})
+	if idx != -1 {
+		t.Fatal("Expected index -1, got", idx)
+	}
+}
+
+func TestRIndex(t *testing.T) {
+	var q Deque[rune]
+	for _, x := range "Hello, 世界" {
+		q.PushBack(x)
+	}
+	idx := q.RIndex(func(item rune) bool {
+		c := item
+		return unicode.Is(unicode.Han, c)
+	})
+	if idx != 8 {
+		t.Fatal("Expected index 8, got", idx)
+	}
+	idx = q.RIndex(func(item rune) bool {
+		c := item
+		return c == 'H'
+	})
+	if idx != 0 {
+		t.Fatal("Expected index 0, got", idx)
+	}
+	idx = q.RIndex(func(item rune) bool {
+		return false
+	})
+	if idx != -1 {
+		t.Fatal("Expected index -1, got", idx)
+	}
+}
+
+func TestInsert(t *testing.T) {
+	q := new(Deque[rune])
+	for _, x := range "ABCDEFG" {
+		q.PushBack(x)
+	}
+	q.Insert(4, 'x') // ABCDxEFG
+	if q.At(4) != 'x' {
+		t.Error("expected x at position 4, got", q.At(4))
+	}
+
+	q.Insert(2, 'y') // AByCDxEFG
+	if q.At(2) != 'y' {
+		t.Error("expected y at position 2")
+	}
+	if q.At(5) != 'x' {
+		t.Error("expected x at position 5")
+	}
+
+	q.Insert(0, 'b') // bAByCDxEFG
+	if q.Front() != 'b' {
+		t.Error("expected b inserted at front, got", q.Front())
+	}
+
+	q.Insert(q.Len(), 'e') // bAByCDxEFGe
+
+	for i, x := range "bAByCDxEFGe" {
+		if q.PopFront() != x {
+			t.Error("expected", x, "at position", i)
+		}
+	}
+
+	qs := New[string](16)
+
+	for i := 0; i < qs.Cap(); i++ {
+		qs.PushBack(fmt.Sprint(i))
+	}
+	// deque: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
+	// buffer: [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+	for i := 0; i < qs.Cap()/2; i++ {
+		qs.PopFront()
+	}
+	// deque: 8 9 10 11 12 13 14 15
+	// buffer: [_,_,_,_,_,_,_,_,8,9,10,11,12,13,14,15]
+	for i := 0; i < qs.Cap()/4; i++ {
+		qs.PushBack(fmt.Sprint(qs.Cap() + i))
+	}
+	// deque: 8 9 10 11 12 13 14 15 16 17 18 19
+	// buffer: [16,17,18,19,_,_,_,_,8,9,10,11,12,13,14,15]
+
+	at := qs.Len() - 2
+	qs.Insert(at, "x")
+	// deque: 8 9 10 11 12 13 14 15 16 17 x 18 19
+	// buffer: [16,17,x,18,19,_,_,_,8,9,10,11,12,13,14,15]
+	if qs.At(at) != "x" {
+		t.Error("expected x at position", at)
+	}
+	if qs.At(at) != "x" {
+		t.Error("expected x at position", at)
+	}
+
+	qs.Insert(2, "y")
+	// deque: 8 9 y 10 11 12 13 14 15 16 17 x 18 19
+	// buffer: [16,17,x,18,19,_,_,8,9,y,10,11,12,13,14,15]
+	if qs.At(2) != "y" {
+		t.Error("expected y at position 2")
+	}
+	if qs.At(at+1) != "x" {
+		t.Error("expected x at position 5")
+	}
+
+	qs.Insert(0, "b")
+	// deque: b 8 9 y 10 11 12 13 14 15 16 17 x 18 19
+	// buffer: [16,17,x,18,19,_,b,8,9,y,10,11,12,13,14,15]
+	if qs.Front() != "b" {
+		t.Error("expected b inserted at front, got", qs.Front())
+	}
+
+	qs.Insert(qs.Len(), "e")
+	if qs.Cap() != qs.Len() {
+		t.Fatal("Expected full buffer")
+	}
+	// deque: b 8 9 y 10 11 12 13 14 15 16 17 x 18 19 e
+	// buffer: [16,17,x,18,19,e,b,8,9,y,10,11,12,13,14,15]
+	for i, x := range []string{"16", "17", "x", "18", "19", "e", "b", "8", "9", "y", "10", "11", "12", "13", "14", "15"} {
+		if qs.buf[i] != x {
+			t.Error("expected", x, "at buffer position", i)
+		}
+	}
+	for i, x := range []string{"b", "8", "9", "y", "10", "11", "12", "13", "14", "15", "16", "17", "x", "18", "19", "e"} {
+		if qs.Front() != x {
+			t.Error("expected", x, "at position", i, "got", qs.Front())
+		}
+		qs.PopFront()
+	}
+}
+
+func TestRemove(t *testing.T) {
+	q := new(Deque[rune])
+	for _, x := range "ABCDEFG" {
+		q.PushBack(x)
+	}
+
+	if q.Remove(4) != 'E' { // ABCDFG
+		t.Error("expected E from position 4")
+	}
+
+	if q.Remove(2) != 'C' { // ABDFG
+		t.Error("expected C at position 2")
+	}
+	if q.Back() != 'G' {
+		t.Error("expected G at back")
+	}
+
+	if q.Remove(0) != 'A' { // BDFG
+		t.Error("expected to remove A from front")
+	}
+	if q.Front() != 'B' {
+		t.Error("expected G at back")
+	}
+
+	if q.Remove(q.Len()-1) != 'G' { // BDF
+		t.Error("expected to remove G from back")
+	}
+	if q.Back() != 'F' {
+		t.Error("expected F at back")
+	}
+
+	if q.Len() != 3 {
+		t.Error("wrong length")
+	}
+}
+
+func TestFrontBackOutOfRangePanics(t *testing.T) {
+	const msg = "should panic when peeking empty queue"
+	var q Deque[int]
+	assertPanics(t, msg, func() {
+		q.Front()
+	})
+	assertPanics(t, msg, func() {
+		q.Back()
+	})
+
+	q.PushBack(1)
+	q.PopFront()
+
+	assertPanics(t, msg, func() {
+		q.Front()
+	})
+	assertPanics(t, msg, func() {
+		q.Back()
+	})
+}
+
+func TestPopFrontOutOfRangePanics(t *testing.T) {
+	var q Deque[int]
+
+	assertPanics(t, "should panic when removing empty queue", func() {
+		q.PopFront()
+	})
+
+	q.PushBack(1)
+	q.PopFront()
+
+	assertPanics(t, "should panic when removing emptied queue", func() {
+		q.PopFront()
+	})
+}
+
+func TestPopBackOutOfRangePanics(t *testing.T) {
+	var q Deque[int]
+
+	assertPanics(t, "should panic when removing empty queue", func() {
+		q.PopBack()
+	})
+
+	q.PushBack(1)
+	q.PopBack()
+
+	assertPanics(t, "should panic when removing emptied queue", func() {
+		q.PopBack()
+	})
+}
+
+func TestAtOutOfRangePanics(t *testing.T) {
+	var q Deque[int]
+
+	q.PushBack(1)
+	q.PushBack(2)
+	q.PushBack(3)
+
+	assertPanics(t, "should panic when negative index", func() {
+		q.At(-4)
+	})
+
+	assertPanics(t, "should panic when index greater than length", func() {
+		q.At(4)
+	})
+}
+
+func TestSetOutOfRangePanics(t *testing.T) {
+	var q Deque[int]
+
+	q.PushBack(1)
+	q.PushBack(2)
+	q.PushBack(3)
+
+	assertPanics(t, "should panic when negative index", func() {
+		q.Set(-4, 1)
+	})
+
+	assertPanics(t, "should panic when index greater than length", func() {
+		q.Set(4, 1)
+	})
+}
+
+func TestInsertOutOfRangePanics(t *testing.T) {
+	q := new(Deque[string])
+
+	assertPanics(t, "should panic when inserting out of range", func() {
+		q.Insert(1, "X")
+	})
+
+	q.PushBack("A")
+
+	assertPanics(t, "should panic when inserting at negative index", func() {
+		q.Insert(-1, "Y")
+	})
+
+	assertPanics(t, "should panic when inserting out of range", func() {
+		q.Insert(2, "B")
+	})
+}
+
+func TestRemoveOutOfRangePanics(t *testing.T) {
+	q := new(Deque[string])
+
+	assertPanics(t, "should panic when removing from empty queue", func() {
+		q.Remove(0)
+	})
+
+	q.PushBack("A")
+
+	assertPanics(t, "should panic when removing at negative index", func() {
+		q.Remove(-1)
+	})
+
+	assertPanics(t, "should panic when removing out of range", func() {
+		q.Remove(1)
+	})
+}
+
+func TestSetMinCapacity(t *testing.T) {
+	var q Deque[string]
+	exp := uint(8)
+	q.SetMinCapacity(exp)
+	q.PushBack("A")
+	if q.minCap != 1<<exp {
+		t.Fatal("wrong minimum capacity")
+	}
+	if len(q.buf) != 1<<exp {
+		t.Fatal("wrong buffer size")
+	}
+	q.PopBack()
+	if q.minCap != 1<<exp {
+		t.Fatal("wrong minimum capacity")
+	}
+	if len(q.buf) != 1<<exp {
+		t.Fatal("wrong buffer size")
+	}
+	q.SetMinCapacity(0)
+	if q.minCap != minCapacity {
+		t.Fatal("wrong minimum capacity")
+	}
+}
+
+func assertPanics(t *testing.T, name string, f func()) {
+	defer func() {
+		if r := recover(); r == nil {
+			t.Errorf("%s: didn't panic as expected", name)
+		}
+	}()
+
+	f()
+}
+
+func BenchmarkPushFront(b *testing.B) {
+	var q Deque[int]
+	for i := 0; i < b.N; i++ {
+		q.PushFront(i)
+	}
+}
+
+func BenchmarkPushBack(b *testing.B) {
+	var q Deque[int]
+	for i := 0; i < b.N; i++ {
+		q.PushBack(i)
+	}
+}
+
+func BenchmarkSerial(b *testing.B) {
+	var q Deque[int]
+	for i := 0; i < b.N; i++ {
+		q.PushBack(i)
+	}
+	for i := 0; i < b.N; i++ {
+		q.PopFront()
+	}
+}
+
+func BenchmarkSerialReverse(b *testing.B) {
+	var q Deque[int]
+	for i := 0; i < b.N; i++ {
+		q.PushFront(i)
+	}
+	for i := 0; i < b.N; i++ {
+		q.PopBack()
+	}
+}
+
+func BenchmarkRotate(b *testing.B) {
+	q := new(Deque[int])
+	for i := 0; i < b.N; i++ {
+		q.PushBack(i)
+	}
+	b.ResetTimer()
+	// N complete rotations on length N - 1.
+	for i := 0; i < b.N; i++ {
+		q.Rotate(b.N - 1)
+	}
+}
+
+func BenchmarkInsert(b *testing.B) {
+	q := new(Deque[int])
+	for i := 0; i < b.N; i++ {
+		q.PushBack(i)
+	}
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		q.Insert(q.Len()/2, -i)
+	}
+}
+
+func BenchmarkRemove(b *testing.B) {
+	q := new(Deque[int])
+	for i := 0; i < b.N; i++ {
+		q.PushBack(i)
+	}
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		q.Remove(q.Len() / 2)
+	}
+}
+
+func BenchmarkYoyo(b *testing.B) {
+	var q Deque[int]
+	for i := 0; i < b.N; i++ {
+		for j := 0; j < 65536; j++ {
+			q.PushBack(j)
+		}
+		for j := 0; j < 65536; j++ {
+			q.PopFront()
+		}
+	}
+}
+
+func BenchmarkYoyoFixed(b *testing.B) {
+	var q Deque[int]
+	q.SetMinCapacity(16)
+	for i := 0; i < b.N; i++ {
+		for j := 0; j < 65536; j++ {
+			q.PushBack(j)
+		}
+		for j := 0; j < 65536; j++ {
+			q.PopFront()
+		}
+	}
+}
--- a/util/queue/priorityqueue.go
+++ b/util/queue/priorityqueue.go
@@ -69,6 +69,13 @@ func (pq *PriorityQueue) Pop() *Item {
 	return heap.Pop(&pq.priorityQueueSlice).(*Item)
 }

+func (pq *PriorityQueue) GetHighest() *Item{
+	if len(pq.priorityQueueSlice)>0 {
+		return pq.priorityQueueSlice[0]
+	}
+	
+	return nil
+}
 func (pq *PriorityQueue) Len() int {
 	return len(pq.priorityQueueSlice)
 }
Author	SHA1	Message	Date
duanhf2012	a61979e985	优化消息队列服务持久化	2023-05-17 11:26:29 +08:00
duanhf2012	6de25d1c6d	优化rankservice错误返回	2023-05-09 14:34:33 +08:00
duanhf2012	b392617d6e	优化性能监控与rankservice持久化	2023-05-09 14:06:17 +08:00
duanhf2012	92fdb7860c	优化本地node中的服务rpc	2023-05-04 17:53:42 +08:00
duanhf2012	f78d0d58be	优化rpc与rankservice持久化	2023-05-04 17:35:40 +08:00
duanhf2012	5675681ab1	优化concurrent与rpc模块	2023-05-04 14:21:29 +08:00
duanhf2012	ddeaaf7d77	优化concurrent模块	2023-04-11 10:29:06 +08:00
duanhf2012	1174b47475	IService接口新增扩展IConcurrent	2023-04-04 16:36:05 +08:00
duanhf2012	18fff3b567	优化concurrent模块，新增返回值控制是否回调	2023-03-31 15:12:27 +08:00
duanhf2012	7ab6c88f9c	整理优化rpc	2023-03-23 10:06:41 +08:00
duanhf2012	6b64de06a2	优化增加TcpService的包长度字段配置	2023-03-22 14:59:22 +08:00
duanhf2012	95b153f8cf	优化network包长度字段自动计算	2023-03-20 15:20:04 +08:00
duanhf2012	f3ff09b90f	优化rpc调用错误日志限制配置的服务必需安装优化结点断开连接时删除结点	2023-03-17 12:09:00 +08:00
duanhf2012	f9738fb9d0	Merge branch 'master' of https://github.com/duanhf2012/origin	2023-03-06 15:57:33 +08:00
duanhf2012	91e773aa8c	补充说明RPC函数名的规范支持RPCFunctionName	2023-03-06 15:57:23 +08:00
origin	c9b96404f4	Merge pull request #873 from duanhf2012/dependabot/go_modules/golang.org/x/crypto-0.1.0 Bump golang.org/x/crypto from 0.0.0-20201216223049-8b5274cf687f to 0.1.0	2023-03-06 15:45:40 +08:00
duanhf2012	aaae63a674	新增支持RPC函数命名RPCXXX格式	2023-03-06 15:41:51 +08:00
duanhf2012	47dc21aee1	优化rpc返回参数与请求参数不一致时报错	2023-03-06 11:47:23 +08:00
dependabot[bot]	4d09532801	Bump golang.org/x/crypto from 0.0.0-20201216223049-8b5274cf687f to 0.1.0 Bumps [golang.org/x/crypto](https://github.com/golang/crypto) from 0.0.0-20201216223049-8b5274cf687f to 0.1.0. - [Release notes](https://github.com/golang/crypto/releases) - [Commits](https://github.com/golang/crypto/commits/v0.1.0) --- updated-dependencies: - dependency-name: golang.org/x/crypto dependency-type: indirect ... Signed-off-by: dependabot[bot] <support@github.com>	2023-03-06 02:26:33 +00:00
origin	d3ad7fc898	Merge pull request #871 from duanhf2012/dependabot/go_modules/golang.org/x/text-0.3.8 Bump golang.org/x/text from 0.3.6 to 0.3.8	2023-03-06 10:08:56 +08:00
dependabot[bot]	ba2b0568b2	Bump golang.org/x/text from 0.3.6 to 0.3.8 Bumps [golang.org/x/text](https://github.com/golang/text) from 0.3.6 to 0.3.8. - [Release notes](https://github.com/golang/text/releases) - [Commits](https://github.com/golang/text/compare/v0.3.6...v0.3.8) --- updated-dependencies: - dependency-name: golang.org/x/text dependency-type: indirect ... Signed-off-by: dependabot[bot] <support@github.com>	2023-02-23 07:30:17 +00:00
duanhf2012	5a3600bd62	补充说明服务的安装启动与停止顺序	2023-02-22 15:43:56 +08:00
duanhf2012	4783d05e75	优化并发模块	2023-02-22 15:13:51 +08:00
duanhf2012	8cc1b1afcb	补充并发函数调用说明	2023-02-22 10:49:01 +08:00
duanhf2012	53d9392901	新增《并发函数调用》说明	2023-02-22 10:45:17 +08:00
duanhf2012	8111b12da5	新增异步函数执行功能	2023-02-22 09:53:50 +08:00
duanhf2012	0ebbe0e31d	优化服务的启停顺序	2023-02-16 15:59:07 +08:00
duanhf2012	e326e342f2	新增排行榜扩展数据以及数据的修改	2023-02-13 17:04:36 +08:00
duanhf2012	a7c6b45764	优化本结点与跨结点Rpc结构&简化原始Rpc接口	2023-01-31 13:50:41 +08:00
duanhf2012	541abd93b4	Merge branch 'master' of https://github.com/duanhf2012/origin	2023-01-29 10:16:23 +08:00
duanhf2012	60064cbba6	优化网络模块	2022-12-27 14:20:18 +08:00