On Dynamic Pricing in Market-Based Resource Management for Clusters

Traditional cluster resource management systems (RMSs) focused on metrics such as maximizing throughput and minimizing mean waiting time, but failed to capture utility of users and revenue for providers. Consequently, researchers have been examining the appropriateness of market-based resource management techniques. This paper presents a market-based RMS for clusters. A dynamic pricing mechanism is proposed, with which the cluster owner can determine the level of sharing within a cluster. In addition, a scheduling algorithm that incorporates the pricing function to enforce market-based resource allocation is introduced. Finally, simulation is carried out to analyze the performance of the market-based RMS. Results show that the proposed resource allocation strategy improves the revenue of the cluster owner and user satisfaction.