Secure and task abortion aware GA-based hybrid metaheuristics for grid scheduling

Abstract

In traditional distributed computing the users and owners of the computational resources usually belong to the same administrative domain. Therefore security and reliability of the resources are not concerned in such a setting. These issues need to be addressed in scheduling in the Computational Grid systems, where the users and distributed resource clusters work in different autonomous domains. In this paper we present a non-cooperative symmetric game to address the requirements for the security and reliability. The game model takes into account the realistic feature that Grid users usually act independently. The users' cost of playing the game is interpreted as a total cost of the secure job execution, which can be aborted due the machines unreliability and Grid dynamics. The Grid users game is transformed into a bi-level optimization problem, which is solved by four hybrid genetic-based heuristics. We have experimentally evaluated the approach using a Grid simulator under the heterogeneity, the large-scale and dynamics conditions. The relative performance of four hybrid schedulers is measured through the makespan and flowtime metrics. The obtained results suggest that it is worth for the Grid users to pay some additional cost of the verification of the security conditions and possible task abortion in order to achieve an efficient allocation of tasks to the trustful and reliable resources. © 2010 Springer-Verlag.