Research And Implementation Of Cloud Monitoring System Based On Ganglia

With the emergence of cloud computing, the service quality of the cloud computing is gradually being concerned. The service quality is the standard of the consumer's satisfaction with the service provided by the service provider.Platform of cloud computing scale is huge, how to reduce all kinds of faults on the cloud platform, the minimum system resources to achieve maximum efforts to monitor the ensure cloud computing platform security operation to provide stable service is worthy of further study. In the actual production environment, administrators need to deploy effective monitoring system for cloud computing platform, to keep cloud platform running normally.Monitoring resource index of a series of CPU, memory and other informations is the most important part in cloud platform, we also should pay more attention to the cloud computing platform application running status.At the same time the alarm monitoring system is an important part, immediately inform cloud platform administrator to handle fault or exception.In this paper, the Ganglia is an open source project developed by Berkeley University, which is a distributed cluster monitoring system. There are Gmetad, Gmond, Gweb. We analyze the shortcomings of current Ganglia. First, Gmetad node using disk IO is too high when cluster had a lot of nodes. This is because Gemtad use RRD file storage which has shortcomings.Second, Ganglia has no application related monitoring. The monitoring system can monitor the hardware resources of the operating system, and meet the requirements of monitoring server nodes.The monitoring index data between service and hardware index is blank.Third, Ganglia lack of alarm mechanism.The alarm mechanism on the cloud platform is the most important to cloud administrators.The alarm mechanism can help administrators locate the problem quickly and maintain the stability of the platform.Finally, based on the Ganglia architecture, this paper modifies and improves the functions of monitoring system, improves layered architecture and presents feedback alarm model based on MDD. We modify Gmetad to achieve a data storage NoSQL database and improve the layered architecture; add new interface display due to change data storage; develop the application information monitoring module on the basis of Gmod; add the feedback module based on the multi valued decision diagram to realize the feedback alarm function. Modifing Gmetad module is used C language to achieve; front interface is display by HTML, PHP and JS technology etc. completed the front desk interface display; application information monitoring is realized through the Python and open source library Psutil etc., the feedback module is completed through the Python and open source asynchronous task queue Celery etc.. After the completion of the development, the system has a detailed unit functional testing, functional testing successfully passed, the monitoring system works well.