Transfer Probability-Based Job Reallocation Method for Heterogeneous Edge Clouds

Abstract

In mobile edge computing (MEC), efficient job allocation is essential to optimize system performance and reduce reliance on cloud computing. Edge servers, deployed at base stations, must handle user-submitted jobs without overloading, which would otherwise lead to excessive job transfers to the cloud. Current k-means-based server-placement and job-allocation methods primarily minimize communication costs but fail to handle heterogeneous server performance. This oversight results in load imbalances where low-performance servers become overloaded, increasing unnecessary cloud transfers and network congestion. Such methods also do not address k-means’ sensitivity to initialization, which impacts job-distribution efficiency. To overcome these limitations, we propose a joint optimization method for integrating edge-server placement and job allocation with the objective of minimizing transfer probability in heterogeneous MEC environments. The method integrates a k-means++-based initial placement algorithm to reduce initialization sensitivity and dynamic job-reallocation algorithm that adjusts assignments on the basis of transfer probability. Extensive simulations demonstrate that our method reduces job overflow and cloud transfers compared with conventional methods. Real-world millimeter-wave communication experiments also confirm the effectiveness of the proposed method in practical MEC environments.