An on-line, business-oriented optimization of performance and availability for utility computing

Utility computing provides a pay-as-you-go approach to information systems in which application providers (e.g., web sites) can better manage their costs by adding capacity in response to increased demands and shedding capacity when it is no longer needed. This paper addresses application providers who use clusters of servers. Our work develops a framework to determine the number of servers that minimizes the sum of quality-of-service (QoS) costs resulting from service level penalties and server holding costs for the server cluster. The server characteristics considered are service rate, failure rates, repair rates, and costs. The contributions of this paper are: 1) a model for the performance and availability of an e-Commerce system that is consistent with data from a multisystem testbed with an e-Commerce workload; 2) a business-oriented cost model for resource allocation for application providers; 3) a closed form approximation for the optimal allocation of servers for an application provider based on the performance model in 1) and the cost model in 2); and 4) a simple criteria for utility owners and server manufacturers to make tradeoffs between server characteristics.