An Evolutionary Hybrid Scheduling Algorithm for Computational Grids

Grids promote user collaboration through flexible, co-ordinated sharing of distributed resources to solve a single large problem. Grid scheduling, similar to resource discovery and monitoring, is inherently more complex in Grid environments. We propose two approaches for solving Grid scheduling problems with the simultaneous objectives of maximizing the number of workflow executions and minimizing the waiting time variance among tasks of each workflow. One is the multiple objective Niched Pareto Genetic Algorithm (NPGA) that involves evolution during a comprehensive search and work on multiple solutions. After the Genetic search, we strengthen the search using Simulated Annealing as a local search meta-heuristic. For comparison, we evaluate other scheduling, such as, Tabu Search (TS), Simulated annealing (SA), and Discrete Particle Swarm Optimization (Discrete PSO). Results show that our proposed evolutionary Hybrid scheduling involving NPGA with an SA search, works better than other scheduling in considering workflow execution time within a deadline and waiting time variance in tasks with minimal iterations.