The Design And Implementation Of SLA Monitoring System Based On SNMP Collection Platform

A Service Level Agreement is a formal negotiated agreement between ISP and its customers, in order to prescribe the specified Quality of Service, in which both expected behavior and QoS parameters of service are included. When validating SLA, all parameters should be measured to insure the contract is kept. Some performance parameters are measurable through SNMP method. With the advantage of simple design, easy implementation, SNMP has been widely applied and become factual industrial standard of network management. Eastern China (North) Regional Network of CERNET provides CERNET connection and internet service for high schools and other education and research institutions in Jiangsu, Shandong and Anhui Province. Its maintenance and management are very important. As the scale of network is larger and lager, the number of users increases sharply, the keystone of network management has changed from validity management to user management, service management and receivable QoS management. The former management software, because of earlier development, some deficiencies of functions, and some bugs found during long-term usage, hasn't been able to fit current network management. Taking requirements of regional network's maintenance as background, the thesis researches the method of monitoring and evaluating network's QoS based on a SNMP collection platform, designs and implements a system evaluating QoS of regional network..The system mainly focuses on the availability of backbone links when evaluating QoS of regional network. In order to describe availability of links, according to data collection, the thesis chooses some suitable performance parameters including link's utilization rate, error rate, discard rate, average and variance of each parameter, as well as network connectivity, and confirms the parameters'calculate method and data collection period. Furthermore the thesis gives out a model of evaluating network's SLA fulfillment. Considering shortcomings of collecting network delay through SNMP, the system integrates some tools to measure performance accessing outer network including response time to outer server and available bandwidth of specified path.SNMP collection platform is system's foundation, takes charge of collecting data from routers on regional network. It adopts traditional manager/agent model, uses multithread to deal with multi-task. In order to improve efficiency of platform, the system improves data collection. It predigests the method of fetching table variables to reduce collected data and request/response course.The system uses file system to store historical data. File system is easy to export and extend, but difficult to search. In order not to reduce the efficiency of searching original data, it takes collection time as directory, link's address as file name when storing. For improving the efficiency of making period report, the system stores some period data with different compression ratio, avoiding temporary data taking too much disk space.Fault is one of important factors effecting network availability. The thesis researches fault detection method without fault information in trap. The fault reason can be separated into five types: router fault, configure fault, interface fault, link fault and system fault. When fault occurs, the system inspects the fault equipment, conclude the reason.The whole framework and the design and implementation of each model are presented, as well as test in real network environment. In the end the thesis summarizes and expects future work.