Game Theoretical Model Applied to Scheduling in Grid Computing

We consider a grid computational model which consist of a number of computation nodes and a number of users. Each user generates a computation load (jobs) requesting computational and communication resources. A deadline for each job is also defined. We propose a scheduling algorithm which is based on Iterated Prisoner's Dilemma (IPD) under the Random Pairing game, where nodes (players) of the grid system decide about their behavior: cooperate or defect. In this game players play a game with randomly chosen players and receive payoffs. Each player has strategies which define its decision. Genetic algorithm (GA) is used to evolve strategies to optimize a criterion related to scheduling problem. In this paper we show that GA is able to discover a strategy in the IPD model providing a cooperation between node-players, which permits to solve scheduling problem in grid.