为了弥补现有的容器化PaaS平台监控方式在监控粒度和深度上的不足,研究并实现了一种智能监控方法.在该方法中,监控模块通过动态调整采样周期,使采样点集中于应用系统资源占用状况波动幅度较大的时段;对于波动幅度超过阈值的情况,采集应用系统调用级别的数据并加以分析;将统计分析结果及原始监测数据按时序整合并对外提供查询接口,进而为用户定位异常或发现性能瓶颈提供帮助.相较于现有的监控方案,该方法能更加细粒度地反映应用运行时状态的变化,同时动态的数据采集机制能有效节省计算和存储资源的开销.%To make up the deficiency of granularity and depth of existing containerized PaaS platform monitoring methods,an intelligent monitoring method was researched and implemented.In this method,the monitoring module dynamically adjusted the sampling period so that the sampling points were concentrated in the time period where the resource consumption occupancy of application system fluctuated greater;if the fluctuation amplitude exceeded the threshold,the system call data of application were collected and analyzed;the monitoring module integrated the statistical analysis results and the original monitoring data in time sequence and provides query interface,which could help the user located abnormalities or found performance bottlenecks.Compared with the existing monitoring scheme,the proposed method could reflect the change of application running state more finely.At the same time,dynamic data collection mechanism could effectively save the cost of computing and storage resources.
展开▼
