面向多用户资源均衡分配的MEC计算卸载策略

Computation offloading strategy for balanced-resource allocation in the multi-user mobile edge computing environment

移动边缘计算在资源受限地区的物联网部署上面临能源、算力、容量等资源限制的挑战,为保障网络QoS,提出一种任务分组和资源分配联合优化的两阶段计算卸载策略(TSCOS)。以卸载决策和资源分配为约束,设计联合优化卸载算法来降低时延、提高资源利用率。第一阶段基于任务偏好的待卸载计算任务分组模型,与候选服务器进行预匹配,提高计算卸载的精准度以降低匹配时间开销;第二阶段运用Gale Shapley算法快速计算多对多博弈最优匹配解集,实现边缘网络负载均衡。与距离分组模型相比,任务偏好的分组模型可以提高卸载精准度以及计算资源的利用率。将TSCOS策略与随机游走策略(RS)、贪心策略(GE)以及动态资源调度策略(FFS+IPFS)对比,实验结果表明,TSCOS在服务器负载均衡方面相比,方差缩小达到2~4倍,任务成功率提高5%~15%;在相同的能耗下,能效比提高5%~10%。任务平均接收率和卸载任务处理成功率达92%和96%。

Due to edge server equipped with the limited resource,such as energy,computing power,capacity and so on,the mobile edge computing is difficult to be deployed in resource-constrained areas.To ensure the quality of service,a Two-Stage Computing Offloading Strategy(TSCOS)combining with task clustering and resource allocation was proposed.Joint optimization offloading algorithm was designed by taking offloading decision and resource allocation as constraints to reduce delay and improve resource utilization rate.Through pre-matching the offloaded tasks with the candidate servers,a grouping model based on task preference was designed to improve the accuracy and matching time during offloading phase.Then,an improved Gale Shapley method was used to realize the load balance of edge network by quickly calculating the optimal solution set of many-to-many game matches.Compared with distance grouping model,the task preference grouping model could improve the utilization of computing resources.To evaluate the performance of the proposed mechanism,the TSCOS was compared with random walk strategy,greedy strategy and dynamic resource scheduling strategy(FFS+IPFS).The result showed that the edge server load balancing variance reduced by 2~4 times,and the task completion success rate was improved 5%~15%.Under the same energy consumption,the energy-efficient was improved 5%~10%.Moreover,the average acceptance rate of offloading tasks and the average success rate were 92%and 96%respectively.