VCONF: a reinforcement learning approach to virtual machines auto-configuration
ICAC '09 Proceedings of the 6th international conference on Autonomic computing
A Reinforcement Learning Approach to Online Web Systems Auto-configuration
ICDCS '09 Proceedings of the 2009 29th IEEE International Conference on Distributed Computing Systems
Hi-index | 0.00 |
In cloud computing, virtual machines (VMs) tend to be configured and reconfigured on the fly for the provisioning of an elastic computing environment. Their resident applications often have a number of performance-critical parameters that need to be configured in response to the change of workload and VM resources. The interplay of these two levels of configuration requires autoconfiguration of the cloud systems in a coordinated manner. In this paper, we propose a framework, namely CoTuner, for this purpose. At the heart of the framework is a self-learning and automated control approach. The approach is a hybrid of reinforcement learning and the Nelder-Mead optimization techniques. It is further enhanced by the use of systems knowledge to accelerate the learning process. Experimental results demonstrate that the approach gains more than 15% performance improvement over independent tuning of VMs and application parameters. It can drive the cloud system into an optimal or near-optimal configuration in less than 50 trial-and-error iterations.