分布式并行服务器透明性及任务调度研究  

Research on Transparency and Task Scheduling of a Distributed Parallel Server

作　　者：杨峰[1] 刘心松[1] 左朝树[1] 唐续[1] 

机构地区：[1]电子科技大学计算机学院,成都610054

出　　处：《计算机研究与发展》2003年第9期1319-1325,共7页Journal of Computer Research and Development

基　　金：四川省科技攻关项目 ( 0 2GG0 0 6 0 18)

摘　　要：针对当前多服务器系统透明性和任务调度研究中存在的问题 ,提出一种分布式并行服务器的网络服务透明性实现机制和相应的任务调度算法。该透明性机制修改服务器结点的ARP地址解析协议以及客户端到服务器端的连接和数据请求处理 ,使得整个服务器系统对外界表现为惟一的VIP地址和VMAC地址 ;相应的任务调度算法则根据负载和阈值设置将服务器结点分成两个链 :有效服务器结点链和过载服务器结点链 ,然后由量值循环法对有效服务器结点链进行任务调度 在修改Linux内核网卡驱动程序和部分底层网络协议的基础上进行了实现 。It is well known that the effectiveness of a distributed parallel server is greatly dependent on the efficiency of the transparency implementation mechanism and the corresponding task scheduling algorithm For this reason, much work has been devoted to the development of new techniques But there still exist many problems such as load balancer, less efficiency and special purposeness The aim of this paper is to present a new transparency implement mechanism and the corresponding task scheduling algorithm for solving the problems The transparency mechanism makes the server system show only one VIP and one VMAC for clients by modifying the address resolution protocol of server node and the handling manner of request from client to server The task scheduling algorithm divides the server nodes into two lists according load and threshold, one is valid server node list, the other is overload server node list Then the valid server node list is scheduled by a weight round robin algorithm The transparency implementation mechanism and the corresponding task scheduling algorithm are proved efficient to improve the response characteristic and service capability of distributed parallel server

关 键 词：分布式并行服务器 透明性 任务调度 地址解析 阈值 

分 类 号：TP338[自动化与计算机技术—计算机系统结构] TP393[自动化与计算机技术—计算机科学与技术]