Research On Association Rules And The Application Of Server Intelligent Management

As the most critical equipment in the services of network application, the security and efficient operation of server is critical. At present, there are several ways to manage server: artificial management, monitoring software management, KVM management, proprietary tools for management et al. All of them have some shortcomings such as limited ability to monitor, lack of independence and the level of intelligence is low et al. With the growing popularity and development of Web services, making the server load and number increase, Meanwhile, it's more and more difficult to manage server by the affects of environment, business requirements and other conditions, can't achieve server centralized management and difficult to ensure that it can provide the long-term, stable services. When a server fails, it could not provide normal network services during the time between administrators arrive in the room and the trouble is cleared. Therefore, it is essential to establish a higher independence, scalability, intelligent server management system. We can improve the monitoring capacity and independence of the server management system by IPMI. The IPMI-based server management platform can work unaffected, even if the monitored server was shut down or downtime, the system still can achieve the monitoring server, and the exception is saved to the running log. At the same time, there are some useful information included in the server running log, they can provide decision support service for server intelligent management. But due to lack of powerful analytical tool, we can't access that valuable information. The purpose of this paper is to use association rules mining technology combined with the IPMI specification, established a server management module by the IPMI specification to collect health information server event log as training data, then the association rule mining is used to access the operation rules, and improving the administrator's ability to control the server.Specifically, we carried out and completed the following work:(1) We analyzed and summarized data mining association rules mining especially, and then an improved algorithm Apriori-M is proposed which combined with matrix to improve the efficiency of mining frequent item lower in Apriori algorithm. The algorithm simply scan the database twice to get all the correct set of frequent items, improved the efficiency of the connectivity and the pruning by the character of matrix. And also solved the question of bottleneck in 2-itemsets. Experimental results show that the capability of the improved algorithm is more efficient than Apriori.(2) In the aspect of researching on server management platform, we expand server remote management system based on IPMI for the current status and problems, make it to have KVM over IP and remote access health information functions in the server management platform. There are many problems in using XML at the transmission efficiency, compatibility and the storage space, which is not suitable for embedded real-time management system. In order to store the health information of server, we propose a framework basing on JSON specification to acquire and analyze the server dynamic management data. This framework improves the efficiency of data exchange, achieving cross-platform database access, saving flash space of the embedded systems, and improving the compatibility.(3) We introduced the association rules mining into the field of server intelligent management, use the previous server log to develop a knowledge base. Matching the data which is obtained from the real-time monitor server running status with the knowledge library's information. Thus forecast the potential failure and the risk, enhancing server's safe operation time, and providing higher efficiency of the network service.