Self-management for neural dynamics in brain-like information processing
ICAC '09 Proceedings of the 6th international conference on Autonomic computing
IBM zEnterprise unified resource manager platform performance management
IBM Journal of Research and Development
Hi-index | 0.00 |
Autonomic control managers can remove the need for manual system configuration in order to achieve good performance and efficient resource utilization. However, simple controllers based on reconfiguration actions tied to thresholds, or 'if-then' rules, themselves need to be configured and tuned in order to adapt the controller behavior to the expected workload characteristic. In this paper we present an experimental study of zero-configuration policies that can be automatically tuned based on analytical models of the system under control. In particular, we have designed and implemented a threshold-free self-configuration policy for a distributed workflow execution engine and compared it with a standard PID controller. The experimental results included in the paper show that using such apolicy the controller can tune itself in addition to reconfiguring the distributed engine and the proposed policy out-performs simpler policies that require manual and error-prone tuning of their parameters.