Early fault warning mechanism based-on association rules in server clusters management system

Yumei Ning, Zhenguo Ding, Ping Zeng

School of computer science and technology, Xidian University, Xi’an 710071, Shaanxi, China

With the rapid development of Internet and business internal information technology, the problem of server cluster fault early warning becomes important in server clusters management. To solve the problem, a server cluster fault early warning system (SCFEWS) based on event association rules is presented. This system is mainly based on the fault event association tree. First obtain event relationship and association rules. Construct early warning events associated tree with system logs and association rules, and store it in binary tree linked list. Then by using the warning event filtering algorithm, redundancy fault early warning events are filtered out and only source event for early-warning notification are kept. Experiment shows the proposed algorithm can effectively improve the accuracy of fault location for server clusters management.