Home // International Journal On Advances in Intelligent Systems, volume 1, number 1, 2008 // View article

Polling Schedule Optimization for Adaptive Monitoring to Scalable Enterprise Systems

Authors:
Fumio Machida
Masahiro Kawato
Yoshiharu Maeno

Keywords: Adaptive monitoring, Polling schedule, Virtualization, System reconfigurations, Information freshness

Abstract:
Adaptive monitoring is a promising technique to automate configurations of a monitoring server in enterprise systems according to the dynamic system reconfigurations such as server scale-out and virtual machine migration. Even after the system reconfiguration, the monitoring server need to be configured properly for providing the fresh information to clients with stabilized server load. In this paper, we propose an adaptive monitoring system that automatically changes the monitoring schedule to satisfy the required freshness under the limited server load after system reconfigurations. The adaptive monitoring system consists of a polling-based monitoring architecture and an algorithm for polling schedule generation. Since the problem for polling schedule generation is classified in NP-hard, we propose an approximation algorithm. According to the results from the experiments with real system reconfiguration scenarios, the adaptive monitoring system improves the variation coefficients of changes of CPU usages and network traffics in the monitoring server by at most 80%. We extend the proposed adaptive monitoring system to be scalable by introducing a hierarchical architecture.

Pages: 11 to 22

Copyright: Copyright (c) to authors, 2008. Used with permission.

Publication date: February 24, 2009

Published in: journal

ISSN: 1942-2679