Cacti And Lsf Cluster-based Monitoring And Management System Design And Realization

Cluster monitor and management as one of the hot issues in cloud computing times, has aroused great attention from numbers of enterprises around the world, it can effectively solve the enterprises with the problems to monitor and manage the clusters on the way to achieve cloud computing. Cluster monitor and management system is designed to help users to monitor, report the high performance computing work load conditions and to analyse the relevant data for further. The article around specific application requirements, design and develop a enterprise oriented cluster monitor and management system, studies some key technologies for realizing the system. The main works are as follows:1. System analysis and design. To solve the specific problems of cluster real-time monitor and management caused by the cloud computing platform (LSF), around the applied requirements of LSF, combins with the real-time mapping functional characteristics of Cacti, connects to LSF and Cacti through the "Grid Data Pollers" and "Advocate", the article desings the overall framework of cluster monitor and management system for enterprise administrator, and analyzes each layer of the framework.2. Analysis and research in the structure of Cacti and LSF. The powerful and extensible Cacti graphical architecture is used as the central part in the system, as well as the LSF system is used as the bottom srtuctue, therefore the cluster monitoring and management system could realize its fouction for montitoring and management. Cacti is to collect data by SNMP and then use RRDtool to draw the graphics. The key framework of LSF, which is based on LSF cluster and key process, is to complete monitoring and management of the queue, task and assignments in LSF cluster, through a series of key process.3. Designing and implementation of Grid Data Poller and Advocate. Cluster monitoring and management system can get the only one data module Grid Data Pollers from LSF. It has the function of real-time using LSF API, and has the ability of communication with Cacti. When Grid Data Pollers gets the data, it could transform the data into the special format for RRDtool, making RRDtool generate the real-time charts. Advocate is the process to send the Cacti command to cluster to perform.4. Research in data storage and query optimization technology. The data gained by Grid Data Pollers need to store in the database, because the amount of data is very large, so we should make sure the various performance of the database when we design it. we should avoid the trouble cases that the system slower its running speed when the database products a large number of records.5. Implementation of the platform. SME-informatization-oriented SaaS operation and management support platform, the first phase of "Xiangyun Services", has been implemented using open source software development tools and has been put into trial operation.