Research Of SaaS Software Aging Trend Prediction Methods Based On Chaos Theory

As the evolving of SaaS software technology, its application range is gradually expanding, and now more and more small and medium enterprises solve their core problem by this software service of hosted. For SaaS software service provider it is very important to provide satisfactory services for customer’s requesting. But as the running of SaaS software, the performance of system may decline and even downtime because of the defects of itself, and this phenomenon is known as software aging. Software aging is inevitable, and we can do is to find strategies to reduce the impact and losses. The prediction of software aging is very essential of software rejuvenation and is very important to SaaS.This thesis proposed a prediction method of SaaS software aging based on chaos theory after analysis the previous prediction methods. It also gives an overall process of this prediction method. The whole algorithm has two main parts, one is the demonstration of chaotic characters of SaaS software system and the other is prediction method of SaaS software based on RBFN_CHAOS. Firstly, phase space reconstruction method is used to get the information of SaaS software system, and the SaaS software characteristics can be getting through the chaotic attractor, such as correlation dimension, Lyapunov exponent, Kolmogorov entropy and saturated embedding dimension. A prediction method of RBFN_CHAOS proposed based on the analysis of RBF net prediction method and the chaotic characteristics of SaaS software system. Finally, this thesis proved the usefulness and suability of the above algorithm by simulation experiments.The results of experimental show that the prediction method of SaaS software aging trend based on chaos theory have a high predictive accuracy. And we also found many useful information which hidden within system and can’t be found before were clearly shown by the chaotic characteristics. The prediction method this thesis proposed is very important to determine the software aging time and resource allocation, so it has significance in this field.