Research And Implementation Of PC Cluster System Based On Linux

Nowadays, with computer technology has entered a high internet age, client/server model is largely used because of its particular advantages, such as simple, easily manage and easily maintain and so on, which due to the single server can't fit for demand of high performance compute. Therefore, people think of a method to build a high performance server with common PC, which is called cluster. Cluster is the best substitute of expensive super computer, because it's very cheap and convenient, moreover, it demands very low for software and hardware. Due to above reasons, we research and discuss deeply some primary technologies of cluster, and implement a master/slave cluster system based on Linux. Experiment demonstrates that the cluster not only has flexible capability of job management, but also has better capability of load balance; moreover, it is able to deal with client's jobs more efficiently, and return back the result to client in a short response time.This paper does the following works:1. Researches on classic cluster system in the world and primary technologies of current cluster system, such as the mode of job management, load balanceable arithmetic and job scheduling arithmetic.2. Analyzes deeply source code of kernel 2.6 of Linux Fedora core 4, include processes communication, system invoke functions and /proc virtual file system.3. Implements a PC cluster system based on Linux, which can't depend on any parallel computation environment, and is able to deal with client's jobs efficiently.