Multi-resource maximin share fair allocation in the cloud-edge collaborative computing system with bandwidth demand compression

The problem studied in this paper is the maximin share fair allocation of multiple resources in a cloud-edge collaborative computing system with bandwidth demand compression. For this problem, we propose a mechanism called maximin share fairness (MMS-BC) with bandwidth demand compression in the computing system. The MMS-BC mechanism provides two ways for each user to process tasks in the computing system: the user directly uploads the task to the cloud server for processing and storage, and the user first transfers the task to the edge server for processing and then uploads it to the cloud server for storage. When users handle the same task in two different ways, the bandwidth resources consumed by the two methods are definitely different. The bandwidth demand compression comes from this. The MMS-BC mechanism satisfies the following four commonly used properties: Pareto efficiency, envy-freeness, proportionality and strategy-proofness. To evaluate the performance of MMS-BC, we conduct a simulation experiment with a small-scale dataset and Alibaba cluster traces. The experimental results indicate that MMS-BC outperforms other similar mechanisms in terms of resource utilization, the number of tasks users can process, and runtime. Additionally, we conducted experiments on the impact of bandwidth demand compression on the mechanism. Experiments show that the mechanism with bandwidth demand compression greatly improves the number of tasks that users can process and resource utilization.