基于自校正分散控制的Web服务器比例延迟保证

Apache Web服务器一般采用单变量模型,处理多优先级延迟保证时需多次建模,可扩展性不佳。为此提出一种Web服务器多输入多输出(MIMO)模型,并结合分散控制理论和自校正控制(STC)理论设计分散自校正控制器。该控制器动态调节处理不同优先级请求的工作线程数目,能保证较高优先级请求更快得到处理且维持不同优先级请求的平均延迟比为设定值,模型和控制器参数根据在线辨识结果实时更新。仿真表明,过载情况下即使并发客户连接数目急剧变化,闭环系统中的服务器仍能维持较好的比例延迟保证。

Proportional delay guarantee on Web server based on decentralized self-tuning control

Single-variable model is generally used for Apache Web server. It has poor scalability when it is used to handle multiple priorities delay guarantee. A Multi-Input Multi-Output (MIMO) system model for Web server was proposed with decentralized self-tuning controllers based on decentralized control theory and Self-Tuning Control (STC) theory. These controllers could dynamically adjust the number of worker threads, which handled different priority requests to ensure that the higher priority requests were processed faster and the average delay ratio of different priority was maintained the same as settings. The parameters of model and controllers were updated according to the result of online identification. The results of simulation tests indicate that proportional delay guarantee can be maintained on the Web server in the closed-loop system even if the number of concurrent client connections change abruptly under overload conditions.