A Software Aging Study of the Xen Hypervisor Under I/O-Intensive Workload

Le résumé fourni par la source

Software aging is a phenomenon affecting long-running systems, characterized by the progressive degradation of performance and reliability, often caused by memory leaks, fragmentation, and resource exhaustion. This work presents an experimental analysis of aging symptoms in the Xen hypervisor, using a cyclic stress load based on virtual disk mount and unmount operations. To identify patterns of continuous degradation, we apply the nonparametric Mann-Kendall statistical test to key time series data of system resources and the virtualized environment. The experiment was conducted over 13 days, divided into sequential cycles of stress and recovery phases. Each stress phase involved continuous disk operations, generating pressure on the I/O subsystem and memory allocation routines. The results reveal statistically significant trends of increasing system resource consumption, indicating classic symptoms of software aging, such as memory leaks, CPU overload, and persistent fragmentation in critical processes. Furthermore, this study provides empirical evidence on the impacts of aging in cloud computing infrastructures, with particular focus on the bare-metal Xen hypervisor, a widely used virtualization solution.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.