Research Of Monitoring System For Data Center

With the rapid development of cloud computing over the recent years,the degree of social information has been further improved,many efforts have been paid to cloud computing in various industries to satisfy the demand of information service,along with it came the expanding scale of enterprise data center,the number of devices is getting more and more and also the environmental conditions of the data center is becoming more complex than before,hence,it is of the very essence to provide a real-time monitoring for various resources in the data center to ensure the reliability of services,however,along with the rapid expanding scale of data center is much more resources,which makes the monitoring tasks become more and more complex,the traditional monitoring method is gradually hard to meet the expanding scale of modern data center.In order to meet the monitoring requirements brought by the expanding scale of data center,this paper designed a monitoring system by analyzing the characteristics of data center resources,moreover,a new approach for condition data acquisition named dynamic data acquisition is provided to extend the data acquisition approaches in the design and implementation of the system,which makes the monitoring task to support dynamic data collection to reduce task load and volume of monitoring data,and ultimately to reduce the workload of monitoring task as well as provide a more flexible method for data collection.A detailed investigation of related monitoring technologies and main monitoring system of data center is conducted firstly by this paper,and then analyzes the functional characteristics and shortcomings of data center monitoring system,what followed is the two aspects of the research which is to reduce the workload of monitoring task and improve the scalability of monitoring system on the basis of investigation.Dynamic data collection algorithm is introduced to solve the first problem,the possibility of abnormal data can be predicted according to the fluctuation of monitoring data,which is as the basis of data collection frequency adjustment.The scalability of monitoring system is mainly studied from the aspect of data transmission and data storage,distributed cluster monitoring system ganglia is chosen as the base module,the distributed message queue kafka is chosen as the data transmission module for message forwarding and the document-oriented database mongodb is chosen as data storagemodule to achieve horizontal expansion.A possible problem in the monitoring process of the data center is further analyzed on the basis of previous research,which is to provide a unified information acquisition interface for different virtual machines,a unified information acquisition interface for virtual machines is implemented by a extended MIB which is used to define common monitoring metrics for different virtualization platform and libvirt which is a library used to retrieve the information of different virtual machines.Finally,the monitoring system is experimented from the aspects of function and performance,the functional experiment is mainly consists of two parts: the display of monitoring data view and the unified information acquisition for different virtual machine,the performance experiment is mainly consists of three parts: dynamic data collection algorithm,the throughput of message queue and database,the experiment results show that the monitoring system platform provide a lower monitoring costs and a better scalability.