Combinatorial auction based resource co-allocation model for grids

Abstract

Resource co-allocation problem is one of the challenging problems in grids. In order to model this problem, a new combinatorial auction based resource co-allocation (CABRC) approach is proposed. This economy based model provides efficient allocation of resources in a grid environment by allowing bidders to submit bids on the combinations of different resource types. In order to solve the model, CABRC problem is defined and formulated using integer programming. It is proved that CABRC problem is NP-hard and since optimum solutions may take tremendous amount of time to be found, two new greedy heuristics based on price per unit criteria are proposed. A software package that consists of an artificial test case generator, an optimum solver, an upper bound estimator and three greedy heuristic solvers for CABRC problem is coded. Since there is no real world data for testing the solvers, performance of algorithms are compared using a comprehensive test suite which is produced by the test case generator. Proposed two polynomial time heuristic solvers produce promising results of 97.3 per cent and 99.2 per cent average performance relative to the optimum solution respectively.