Demand Pricing & Resource Allocation in Market-Based Compute Grids: A Model and Initial Results

Market-based compute grids encompass service providers offering limited resources to potential users with varying quality of service demands and willingness to pay. Providers face problems of pricing and allocating resources to maximize revenue. Previous research proposed and analyzed a method for allocating resources based on joint optimization of access control and scheduling strategies. This paper proposes a tractable analytical model for joint optimization of job pricing and scheduling strategies with the objective of maximizing provider revenue. The paper provides initial results for the case of a single provider serving price-sensitive users whose utilities decay linearly with increasing service delay. The paper also shows that providers must combine both pricing and admission control to achieve maximum revenue.