Comparative Analysis of Various Game Theory Algorithms for CPU and GPU Resource Allocation

In this study, the challenge of dynamic resource allocation concerning CPU (Central Processing Unit) and GPU (Graphical Processing Unit) is addressed. Generally, the default configurations allocate a fixed amount of resources to processes, irrespective of their actual demand. This often results in significant resource wastage. To mitigate this issue and enhance both efficiency and fairness, game theory concepts can be incorporated into the resource allocation process, allowing for a more dynamic approach that reduces resource wastage. In this context, five different game theory models including Nash Bargaining, Shapley Value, Nash Equilibrium, Congestion Game, and VCG (Vickrey Clarke Groves) Auction are implemented across ten different processes with varying CPU and GPU demands. Among these models, the VCG Auction has proven to be the most effective choice when prioritizing efficiency over fairness. Conversely, when fairness is the primary focus, both the Nash Equilibrium and Shapley Value offer a better balance between equitable distribution and efficient resource utilization.