Data mining: practical machine learning tools and techniques with Java implementations
Data mining: practical machine learning tools and techniques with Java implementations
Preserving QoS of e-commerce sites through self-tuning: a performance model approach
Proceedings of the 3rd ACM conference on Electronic Commerce
An Architectural Evaluation of Java TPC-W
HPCA '01 Proceedings of the 7th International Symposium on High-Performance Computer Architecture
OnCall: Defeating Spikes with a Free-Market Application Cluster
ICAC '04 Proceedings of the First International Conference on Autonomic Computing
Dynamic Provisioning of Multi-tier Internet Applications
ICAC '05 Proceedings of the Second International Conference on Automatic Computing
Towards Self-Configuring Hardware for Distributed Computer Systems
ICAC '05 Proceedings of the Second International Conference on Automatic Computing
Advanced virtualization capabilities of POWER5 systems
IBM Journal of Research and Development - POWER5 and packaging
A Hybrid Reinforcement Learning Approach to Autonomic Resource Allocation
ICAC '06 Proceedings of the 2006 IEEE International Conference on Autonomic Computing
ICAC '06 Proceedings of the 2006 IEEE International Conference on Autonomic Computing
Autonomic Provisioning of Backend Databases in Dynamic Content Web Servers
ICAC '06 Proceedings of the 2006 IEEE International Conference on Autonomic Computing
Machine learning for on-line hardware reconfiguration
IJCAI'07 Proceedings of the 20th international joint conference on Artifical intelligence
Hi-index | 0.00 |
As the use of virtualisation and partitioning grows, it becomes possible to deploy a multi-tier web-based application with a variable amount of computing power. This introduces the possibility of provisioning only for a minimum workload, with the intention of renting more resources as necessary, but it also creates the problem of quickly and accurately identifying when more resources are needed or unneeded resources are being paid for. This paper presents a machine learning based approach to handling this problem. An autonomous adaptive agent learns to predict the gain (or loss) that would result from more (or less) resources; this agent uses only low-level system statistics, rather than relying on custom instrumentation of the operating system or middleware. Our agent is fully implemented and evaluated on a publicly available multi-machine, multi-process distributed system (the online transaction processing benchmark TPC-W). We show that our adaptive agent is competitive with any static choice of computing resources over a variety of test workloads. We also show that the agent outperforms each static choice in at least one case, implying that it is well suited for a situation where the workload is unknown.