Competitive Processors Allocation in 2D Mesh Connected Multicomputer Networks: A Dynamic Game Approach

“Partitionable parallel machine” is an efficient computing system for executing parallel applications. In this paper, the problem of processors sharing among multiple users is addressed where users' requests are executed in a 2D mesh-connected multicomputer simultaneously. The problem is formulated as an oligopoly market where users compete to rent free processors. The system consists of self-interested users with different needs and utility functions. Each user engages in an auction with others to achieve their needs of processors. The most significant challenge for the users is how to select the optimal number of requested processors for winning the auction. A noncooperative game is used to obtain the optimal processors allocation for users. Game theory is employed to address this challenge for competing users. Nash equilibrium is considered as the solution of this game. The numerical results reveal the dynamics of distributed dynamic adaptation of processors sharing strategies.