Giraph源码分析(二)——启动Master/Worker服务


本人原创,转载请注明出处! 本人QQ:530422429,欢迎大家指正、讨论。

欢迎访问: 西北工业大学 - 大数据与知识管理研究室 (Northwestern Polytechnical University - BigData and Knowledge Management Lab),链接:http://wowbigdata.cn/,http://wowbigdata.net.cn/,http://wowbigdata.com.cn。


1. org.apache.giraph.bsp.CentralizedService 接口
功能:Basic service interface shared by both CentralizedServiceMaster and CentralizedServiceWorker.

2. org.apache.giraph.bsp.CentralizedServiceMaster 接口
功能:At most, there will be one active master at a time, but many threads can be trying to be the active master.

3. org.apache.giraph.bsp.CentralizedServiceWorker 接口
功能:All workers should have access to this centralized service to execute the following methods.
4. org.apache.giraph.bsp.BspService 抽象类
功能:Zookeeper-based implementation of CentralizedService.
5. org.apache.giraph.master.BspServiceMaster 类
功能:ZooKeeper-based implementation of CentralizedServiceMaster.
6. org.apache.giraph.worker.BspServiceWorker 类
功能:ZooKeeper-based implementation of CentralizedServiceWorker.

    BspServiceWorker类有WorkerClient和WorkerServer实例,分别作为IPC通信的客户端和服务器端,通过Netty来发送数据。WorkerClient实例实际为NettyWorkerClient对象,WorkerServert实例实际为NettyWorkerServer对象。

    NettyWorkerClient implements WorkerClient接口,NettyWorkerServer implements WorkerServer接口。

    NettyWorkerServer类的构造方法中创建一个NettyServer对象,用于底层的IPC的通信,还有一个ServerData对象,作为数据实体。ServerData中包含该Worker的partitionStore、edgeStore、incomingMessageStore、currentMessageStore、聚集值等。

    NettyWorkerClient类的构造方法中创建一个NettyClient对象,用于底层的IPC的通信,作为客户端。

7. org.apache.giraph.worker.InputSplitsCallable 抽象类,继承 Callable接口。

    

   功能:用于加载顶点或边 输入splits,每个线程都有一个WorkerClientRequestProcessor实例(实为 NettyWorkerClientRequestProcessor对象),负责向远端的worker发送数据。

   

    NettyWorkerClientRequestProcessor对象用于发送的WorkerClient对象就是BspServiceWorker里面的WorkerClient对象。

    VertexInputSplitsCallable类中的readInputSplit()方法用来从split中读取顶点的信息,然后调用NettyWorkerClientRequestProcessor对象的sendVertexRequest()方法把顶点发送到它所属的Partition上。

8. org.apache.giraph.graph.ComputeCallable 类,继承Callable接口。在该对象中完成“计算-通信-同步”的过程。每个线程都有一个WorkerClientRequestProcessor实例(实为 NettyWorkerClientRequestProcessor对象),负责向远端的worker发送数据。

 本人原创,转载请注明出处! 本人QQ:530422429,欢迎大家指正、讨论。 

相关内容