Reducing Monitoring Overhead in Virtualized Environments Through Feature Selection

Cloud computing has emerged as a cost-effective paradigm for hosting and delivering services. Cloud providers adopt server consolidation strategies to achieve efficient management of resources. A drawback is that applications running on the same host compete for physical resources. Such interference can affect the performance of applications. Performance monitors are useful tools to detect or even predict performance degradation. However, the monitoring itself can be a source of contention. In this paper, we analyze the influence of performance monitoring overhead in virtualized environments. Furthermore, as a mean to reduce contention for shared resources, we propose an approach to reduce the dimensionality of the performance feature space.