一个支持多优先级延迟限定的调度机制  

Delay-bounded Dispatching Mechanism for Supporting Multiple Priorities

作　　者：张永忠[1] 赵银亮[1] 李增智[1] 

机构地区：[1]西安交通大学计算机科学与技术系,陕西西安710049

出　　处：《系统仿真学报》2007年第5期978-982,共5页Journal of System Simulation

基　　金：863重点计划项目新型网络服务器系统(2004AA111110)

摘　　要：针对客户访问量动态变化的情况下,如何保证不同优先级客户的访问延迟限定,在Web集群服务器中,提出了一种新的基于概率的调度机制。在该机制中,依照排队论,公式化各优先级客户请求的选择概率,并通过折半搜索求出解;通过预测数据,使得选择概率能适应变化的客户访问量。模拟实验验证,在不同的客户访问量的情况下,基于概率的调度机制能有效地提供不同优先级客户的访问延迟限定。To guarantee different priorities＇ users to have pre-specified delay under varying the number of accessing users, a new probability-based dispatching mechanism in cluster-based web servers was proposed. In this mechanism, the choice probability of each priority users＇ requests was formulized according to queue theory, and was solved by bisearch; the choice probability of each priority＇s requests was adapted to the varying number of accessing users by predicting parameters. The simulation results show probability-based dispatching mechamism can succeed in ensuring users of the priority with pre-specified delay under the different number of accessing users.

关 键 词：预指定的延迟 WEB集群服务器 排队论 基于概率的调度机制 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]