Novel method to forecast software aging problems

Abstract

As software system becomes complex and large, some phenomena, performance degradation, error service, and even the abrupt system outage, have been found in the operating system, web server, and even in the security system. These phenomena can be called software ageing problems. In order to make software re-enter into a robust state, some prediction methods need be used to find an optimal time to do rejuvenation. Since the prediction method used now is linear or non-linear, a hybrid method is proposed in this work to improve the prediction accuracy. The proposed method contains three phases. Firstly, a linear method is applied to capture the linear characteristics of resource consumption series. Secondly, a non-linear method is used to fit non-linear features in the original data and residual data. Finally, the characteristic of the linear and non-linear part will be fitted by a non-linear model. In the experiments, the results show that the proposed method outperforms the other two single models.

Keywords

• software maintenance
• software reliability
• internet
• software fault tolerance
• software performance evaluation
• nonlinear method
• nonlinear features
• nonlinear part
• nonlinear model
• software aging problems
• software system
• performance degradation
• error service
• abrupt system outage
• operating system
• web server
• security system
• software ageing problems
• software re-enter
• prediction method
• hybrid method
• prediction accuracy
• linear method
• linear characteristics