Automated Multi-Tier System Design for Service Availability

Creating a cost-effective large-scale multi-tier Enterprise service requires judicious selection and configuration of infrastructure elements and mechanisms. The minimum cost solution that satisfies business requirements for service availability and performance should be identified. Emerging self-managed computing utility environments demand an automated solution to this problem. This solution must be integrated with the utility controller, which typically virtualizes resources for services, thus hiding from them information about the characteristics of the underlying physical infrastructure. In this paper, we present AVED, our initial version of an engine that automatically designs a cost-effective service infrastructure which will meet the service’s availability requirements. AVED explores a design space consisting of multiple combinations of hardware/software configurations presenting various tradeoffs among cost, availability, and performance. AVED can be used to generate a complete picture of the cost-availability tradeoff for the infrastructure design. We also describe how AVED can be integrated with utility computing environments to improve the automation of service lifecycles.