A. Bobbio, S. Garg, M. Gribaudo, A. Horváth, M. Sereno, M. Telek
Software systems experience gradual performance degradation due to several reasons and different preventive and corrective techniques can be applied to restore their performance level. This paper presents a unified model to describe the behavior of long running software systems with performance degrading factors such as system aging and various recovery techniques such as rejuvenation, checkpointing, rollback recovery, restart and replication.
The proposed unified model is described as a fluid stochastic Petri net (FSPN). The FSPN formalism offers a descriptive language that allows a compact and precise description of the model behavior. Various analysis methods can be applied to obtain numerical results for the performance indices of interest. We illustrate the use of the model by means of a simple numerical example which captures rejuvenation, restart and replication.