Reliability analysis of web server cluster systems based on proportional hazards model
Chunyan Hou, Chen Chen, Jinsong Wang, Kai Shi · International Symposium on Software Reliability Engineering · 2015
With the universal application of web server clusters (WSCs), their reliability is drawing more and more attention from academia to industry. Many accelerated life testing (ALT) models have been proposed for hardware load-sharing systems (LSSs), which are not suitable for WSCs whose reliability is significantly dependent on system software. In contrast to hardware which runs all the time after a system launches, software performs only when it is called. This paper presents an approach for WSC reliability and degradation process analysis, which is modeled as a non-homogeneous Markov process (NHMH) composed of several non-homogeneous Poisson processes (NHPPs). The arrival rate of each NHPP corresponds to system software failure rate which is expressed using Cox's proportional hazards model (PHM) in terms of the cumulative and instantaneous load ofsoftware. The first refers to software cumulative execution time, and the latter denotes the rate at which user requests arrive. The result solved from NHMH is a time-varying reliability and degradation process over WSC lifetime. Finally, the evaluation experiment shows the potential of the approach.