Studies about Convex Hull Support Vector Machine in the Server Performance Alarm

Abstract

With the wide application of computer network, the server is considered as the core of the whole network application. Therefore, the server performance monitoring is becoming more and more important. Through the monitoring about the CPU, memory, and numbers of processes and other parameters of the server, it can judge the server performance is good or bad. However, most alarm systems are based on the simple threshold value method, and cannot accurately reflect the server performance. Whether the server performance is abnormal belongs to pattern recognition problem, thus, the paper gives a new method that the support vector machine is applied to analyze the server performance, and introduces a convex hull theory to support vector machines to improve efficiency, in order to divide the server performance parameters quickly and accurately.