Survivability prediction of web system based on log statistics

Widely applied and quickly developed as the SOA theory has been, the instability of distributed Web services will lead to services composition failure. Currently, a hot research topic is that when does the system can make an appropriate adjustment of the system structure dynamically to ensure the system runs at the best performance while the runtime environment or requirement is changed. To address this problem, this paper proposes an approach to predicting the system survivability which bases on log statistic. The method gets the system usage model by monitoring Web log files, and then constructs value model and adopts quantitative model checking to forecast the system survivability to estimate whether the system is survivable in a certain period of time.