Research On Software Aging Detection Method Based On SPC

Software aging is a phenomenon defined as the continuing degradation of software systems during runtime,being particularly noticeable in long-running applications.The elimination of such failures during software development is system reliability and the key to robustness.One of the main causes of software aging is memory leaks,which is caused by a software error in the use of application memory management.It will lead to memory resource exhaustion and accumulation in the system running continuously for a long time,and eventually lead to system performance degradation or system failure and other adverse consequences.The detection of software aging prevent the occurrence of software aging in advance,and the system operator therefore obtain sufficient time to deal with the fault,so the software aging detection has certain practical significance.The traditional trend detection method based on trend detection itself ensure the existence of a memory leak,because the size of the process heap may increase,not only because of the existence of a memory leak,but also because of the memory usage pattern of the application.Because of trend detection alone does not ensure the existence of memory leaks,the paper presents a method based on SPC(Statistical Process Control)technology to detect software aging phenomena.Through stress testing and trend detection,the time of software aging detection is shortened and the accuracy of software aging detection is improved.The method compares the tested software version with the previous stable version in terms of behavioral changes in resource measurement based on software difference analysis.The key tool used is the divergence graph,which represents the time-varying difference between the two signals.The experimental research in the paper focuses on detecting software aging caused by memory leaks,and evaluates the applicability of detecting software aging caused by memory leaks under actual workloads.At the same time,RSS(resident set size)and HUS(heap usage)are used as important performance indicators for software aging detection.Finally,in order to verify the effectiveness of the method proposed in the paper,the paper based on the Squid software platform for experiments,the experimental results show that the EWMA-CUSUM control chart based on SPC has the advantages of short software aging detection time and high detection accuracy compared with the linear regression technology and Hodrick-Prescott technology.At the same time,the experimental results verify that HUS is more suitable as a memory leak performance indicators that cause software aging detection.