Statistical method based computer fault prediction model for large computer centers

Computer laboratories in large computer centers play an important role for the computing education in higher education institutions. Due to heavy usage, there are often hardware or software problems on these computers. Thus computer laboratories management and maintenance is a very important and long-term task. For a computer laboratory, the computers´ monthly used time follows the normal distribution, which is verified by the chi-square goodness-of-fit hypothesis test. According to probability distributions based on the normal distribution, a statistical method based computer fault prediction model for public computer laboratories is presented. Computers´ health state is divided into four classes: the Normal state, the Concerning state, the Warning state, and the Fault state. For different levels of computer fault, different methods are used to maintain it. The experiment and application was conducted in a university computer center with hundreds of computers in 9 computer laboratories. The experiment and application results show that it is an effective way for finding fault computers. It can be used to guide computer maintenance. Computers maintenance workload is greatly reduced by using this fault prediction model.