An Alternative Mechanism for Multiresource Fair Allocation in Heterogeneous Cloud Computing Systems

ABSTRACT Finding a fair allocation is an important issue in many application areas. In a heterogeneous cloud computing system, users may have different requirements, and servers may also have different configurations. The first proposed fair allocation mechanism for heterogeneous cloud computing systems, called DRFH, is based on dominant resource fairness. However, the DRFH mechanism does not satisfy the properties of strong shared incentives and independence of dummy servers. In this article, we propose a simple mechanism, called the maximin share‐based mechanism in a heterogeneous cloud computing system (MMSH), which maximizes the minimum ratio of the user's utility to the maximin share. Because the MMSH mechanism can be formulated as a linear program, a MMSH allocation can be found in polynomial time. Moreover, we prove that MMSH satisfies all the desirable properties including Pareto efficiency, strong sharing incentives, envy‐freeness, group strategy‐proofness, and independence of dummy servers. Using the Alibaba trace to conduct data simulations, the experimental results indicate that in most cases, the allocation generated by the MMSH mechanism has a higher resource utilization rate.