A Novel QoS-Based Co-Allocation Model in Computational Grid

Abstract

In grid systems, co-allocation is a key infrastructure to schedule heterogeneous and distributed resources for high-level applications. Although it has been widely studied, QoS-based resource co-allocation still remains an unsolved issue. In this paper, we propose a novel QoS-based co-allocation model for the grid applications with constraints to budget and deadline. A new concept, called virtual resource agent, is introduced into the co- allocation policy. Theoretical analysis indicates that virtual resource agent can provide quantitative QoS guarantee for applications in terms of budget and deadline. Experimental results show that the proposed model can significantly reduce deadline violation rate and increase the benefits of resource providers compared with other three co-allocation policies.