Research And Implementation Of Monitor System Based On Cloud Platform

In recent years, Cloud computing and other Internet-based services have been developed rapidly, the storm of Cloud Computing has swept across the world. In the meantime, Cloud Computing also have been developed greatly in China, the domestic Internet companies started to pay a lot attention to it.Monitoring System plays an important role in hardening system stability, improving quality of service and reducing operation cost. Monitoring System is an essential part of Cloud Computing Platform. It is the most important part in the operation process even in the entire product life cycle. It can provide early warning ahead of time, store detailed data for tracing problems. Therefore research and implementation on monitoring system have great significance.Nowadays there are many open source monitoring system, such as zabbix、cacti、 collected and so on and so forth. However they both have different disadvantages on different aspects:do not support distribution, horizontal scalability is awful; when monitored machines and monitored items grow rappidly, system performance also decreases rapidly; monitored items are nothing much, do not support the monitoring of certain service and application; do not provide early warning mechanism, etc.In this paper, based on the study of current mutiple open source monitoring system, we design and implement an distributed monitoring system which is more general and high-available. Based on time series rule-based real value prediction technology, we realize traffic peak early warning mechanism;We optimize the push&pull model;divide monitored items into basic monitored items and special monitored items, basic monitored items are configuration free and are automatically collected; using consistent hashing algorithm to distribute data and get distributed load balance property; set standard data format, support plug-in extension, can monitor lots of applications and services; with regard to monitor data, we adopt "double write" strategy to make data high-available, etc.Finally, this paper provides a demo of monitoring mysql and demonstrates basic functions. Then we take a performance test, our monitoring system still has good performance in high concurrency environment. Through the simulation experiment, prove the feasibility of time series rule-based real value prediction technology on traffic prediction.