The design, usage, and performance of GridUFO: A Grid based Unified Framework for Optimization

We present GridUFO (Grid based Unified Framework for Optimization), a Service Oriented Architecture (SOA) compliant Problem Solving Environment (PSE) that allows the user to implement/share metaheuristics based optimization algorithms over a Grid. GridUFO eradicates the shortcomings of earlier projects and provides a unified approach for using algorithm/problem pair over a Grid in the easiest possible “plug & play” manner. This framework allows the users to concentrate on the actual application development, by hiding all the complexities involved in a Grid, without compromising on the functionality and flexibility promised by a Grid. This paper provides a detailed overview of the GridUFO infrastructure, specifically the way it deals with optimization algoirthms and objective functions, handles Service Level Agreements (SLAs), and follows SOA. We also present various results achieved, that demonstrate both the utility and performance of GridUFO under various application workloads and scenarios.