A Simulation Analysis On The Performance Of Software Aging Under Virtual Environment

“Software Aging” refers to a phenomenon of software performance degradation or even server failures caused by resource depletion after the system software runs for a long time.Software reliability and availability may degrade due to its aging problems.In order to solve such problems,in this paper,we propose an active rejuvenation strategy to recover software performance with the help of response time,a critical indicator for reflecting software performance.Recently,Alberto et al.firstly found real world software aging problems in an online business system of AT&T.After long period of analysis,they finally figure out the cause for such phenomenon—high level of workloads made the kernel of system become degradation.Towards this,they established a mathematical model,named M/M/c queue model,for such process.They also proposed three rejuvenation algorithm based on different techniques:(1)The static rejuvenation algorithm with averaging(SRAA);(2)The sampling acceleration rejuvenation algorithm with averaging(SARAA);(3)The central limit theorembased rejuvenation algorithm(CLTA).Because of the effect of workload flood blowing up will also enlarge the response time,the execution of each algorithm are based on the average response time of several continuous requests.In order to make such model more practical,we propose a new model named MMPP/ M/ c based on our previous analysis on traditional M/ M/ c model.Furthermore we present a new rejuvenation algorithm called the empirical distributionbased rejuvenation algorithm(EDRA).For comparing the efficiency of our algorithm to traditional ones,we designed experimental case studies to make contrasts between our algorithm and traditional algorithms.Firstly we use discrete simulate model to design simulation program.Then use this program to simulate normal condition and aging condition separately.We apply the rejuvenation algorithm into experiments to balance system's workloads and,after long period of experiment,we collect experimental indicators and compare the performance and packet lost rate between our algorithm and traditional ones by using the indicators we collected in experiments.The result shows that rejuvenation algorithm can balance the workload well when the system is under high pressures.Moreover,different algorithms are suitable for different types of aging rejuvenation strategies,however if we take suitable configuration of rejuvenation algorithm strategy,it is more effective than other situations.