A Research On The Accelerated Life Testing Of Software Aging

“Software aging” is a phenomenon caused by resource exhaustion, and characterized byprogressive software performance degradation. This phenomenon often occurs in server sidesoftwares, most of which are designed to provide7x24services, and causes performancedegradation or even server failures of these softwares. It has been reported that thisphenomenon is natural in software systems, and a method called “Software rejuvenation” hasbeen provided, which periodically restart the server, release the occupied resource, andreinitialize the server’s inner state, in order to mitigate the damage caused by software aging.In this paper, an experimental study of software aging caused by memory leaks is carriedout, in which an on-line bookstore application following the standard configuration of TPC-Wbenchmark is employed as the system under test (SUT). The experimental test bed use realwebsite for a bookstore on the server side and emulated clients which send requests to theserver following the specification of TPC-W benchmark. In order to reduce the time toapplication failures caused by software aging, the accelerated life testing (ALT) approach ischosen, to obtain the mean time to failure (MTTF) in a non-accelerated condition. Since thevalidation of MTTF using the experimental approach will take an inordinate long time, adiscrete-event simulation model is thus developed to validate the ALT experimental results. Atlast, the rejuvenation mechanism is introduced to the simulation program, and the availabilityof the system under a serial of tests at certain rejuvenation triggers is obtained.