基于软件再生理论的分布式自适应性能监控系统设计

Design of distributed and self-adaptive performance monitoring system based on software rejuvenation

软件再生理论认为,计算系统运行过程中的系统资源损耗是影响系统性能的主要因素。设计一个性能监控系统,通过采集和分析资源使用情况,适时释放被损耗的资源可以有效保证系统的持续高性能。监控系统采用C/S模式以减轻监控端的负载,保证监控端的轻量级,同时实现对监控端的异步监控;基于自组织映射网络对数据的分析,实现对监控端监控参数的自适应调节;提供多种数学模型对系统性能变化进行分析和预测;设计了简单有效的决策方法支持系统的重启控制;最后通过实验证明自适应采集策略有效减少了数据采集和传输量,保证了监控端的轻量级、低负载,尽可能地降低了监控系统本身对被监控系统的影响。

According to the theory of software rejuvenation,the system resources wastage is the major factor of the degradation of computing system.It is useful for the maintenance of system performance to design a performance monitoring system.The monitoring system releases the resources wasted at the appropriate time by collecting and analyzing the run-time system resources data.C/S mode was used in the monitoring system to reduce the load of monitored system,to ensure the lightweight of monitor-side,to achieve asynchronous monitoring of monitored system.The self-adaptive adjustment of monitoring system parameters was implemented based on self-organization map net.Several models were provided to analyze and forecast the performance of system.A simple decision-making method was designed to support the control of reboot.At last the experiments demonstrate that the self-adaptive collection strategy is effective to reduce the data collected and transmitted,to ensure the lightweight and low-load of monitor-side,to minimize the impact of monitoring system on monitored system.