Adaptive Monitoring for Virtual Machine Based Reconfigurable Enterprise Systems

Abstract

Dynamic monitoring adaptation to system reconfigurations such as server scale-out is required especially for the virtual machine based flexible IT systems. For any states of the systems, the monitoring server must provide fresh information with a stabilized load. In this paper, we propose an adaptive monitoring method that generates the monitoring schedule for each state of the target systems. The schedule regulates the processes for updating information cache to keep the required freshness and to stabilize the monitoring load. Since the problem for schedule generation is classified in NP-hard, we propose an approximation algorithm. Results from experiments with system reconfigurations show that the adaptive monitoring method improves the variation coefficients of CPU usages and network traffics of monitoring server by at most 80%.