Fair multiresource allocation with access constraint in cloud–edge systems

Abstract

Fairness is an important quality-of-service criterion for shared computing systems and is a focus of research on multiresource allocation in cloud–edge collaborative computing systems. However, the access restrictions of edge servers and the compression of data transmission between cloud and edge servers greatly affect fairness among users. In this article, we propose a mechanism, task share fairness in cloud–edge collaborative computing systems (TSF-CE), to realize fair multiresource allocation of cloud–edge collaborative computing systems. TSF-CE comprehensively considers the particularities of data compression, bandwidth resources, and access constraints of edge servers. The task share is introduced to measure fairness among users. We prove that TSF-CE satisfies the desirable properties of Pareto efficiency, a weak sharing incentive, envy-freeness and strategy-proofness. We design an offline algorithm for TSF-CE and prove that the algorithm can obtain the fairest allocation. The results of simulations driven by Google and the Alibaba Cluster Trace show that TSF-CE can maximize the minimum task share, ensure the fairness of allocations, and improve the running efficiency and resource utilization. The simulation results also show that TSF-CE is Pareto efficient and has a weak sharing incentive.