基于任务树的分布式计算平台的设计与实现

针对当前网络中PC资源利用率低和大型任务的难解性，设计并实现了一个通用的基于任务树的分布式计算平台TT-DCP（Distributed Computing Platform based on Task Tree）．TT—DCP旨在利用网络中计算机的空闲时间求解可分布运行的大型任务，主要包括服务器端程序、客户端程序、后台数据库三部分．服务器端程序主要负责利用任务树描述大型任务、对树的操作以及处理客户端返回的任务结果；客户端程序采用主动式策略获取子任务，并下载任务到本地执行、将结果返回服务器；后台数据库保存任务树．任务分布化过程中，提出采用任务树结构保存任务数据，使用回溯法构建任务树并保存在后台数据库中．同时，设计并实现了平台中服务器端与客户端的应用层通信协议UPDP（Unprocessed Packet Download Protocol）和PPUP（Processed Packet Upload Protocol）．本文进一步讨论了如何利用TT—DCP求解Mesh网络连通性问题，并给出了处理结果．TT—DCP可以用来求解生物计算中的多序列比对、motif查找等问题．

Design and Implementation of a Distributed Computing Platform Based on Task Tree

To solve large complex problems and use the residual computing power of PCs in the network, a general-purpose distributed computing platform based on task tree (TT-DCP) is designed and implemented in this paper. The goal of TT-DCP is to utilize the unused computing power on clients to solve part of large complex problems. It includes a server program, a client program, and a background database. The server program describes a large task with a task tree, operates the tree, and processes the results returned by the clients; the client program downloads subtasks from server positively, executes them, and sends the result back to the server; the database stores the task tree. This paper brought the conception of task tree in the task decomposition process, and the task tree was built by trace-back algorithm in the database. Meanwhile, application layer protocols UPDP and PPUP were designed and implemented for the communication between the server and the clientd. The paper discussed how to employ TT-DCP to solve the mesh network connectivity problem. TT-DCP can be used to solve computational biology problems like multiple sequence alignment, motif finding, etc.