基于负载再分配的边缘计算任务均衡调度策略

Balanced Scheduling Strategy for Edge Computing Tasks Based on Load Redistribution

摘要:

目前5G异构网络的计算卸载机制在边缘节点资源方面和电力业务调度算法时延与负载均衡方面存在问题。综合考虑边缘服务器负载均衡和业务计算时延，针对min min算法与max min算法的不足，文中提出了基于电力业务优先级二次再分配的均衡调度算法。该算法根据业务的最大容忍时延划分业务的优先级，并且根据最小完成时间标准差确定长短业务比例，实现业务的预分配。基于各个边缘服务器的负载情况，该算法再将高负载边缘服务器上的业务分配给负载低的服务器，实现二次再分配。仿真结果表明，与典型的算法相比较，所提算法缓解了现有技术中存在的以上问题，降低了计算时延，实现负载均衡，提高了系统资源利用率。

Abstract:

The current computational offloading mechanism of 5G heterogeneous network has problems in terms of edge node resources and latency and load balancing of power service scheduling algorithm. The balanced scheduling algorithm based on secondary redistribution of power service priorities is proposed for the shortcomings of min min algorithm and max min algorithm, comprehensively considering the load balancing and service computation delay of edge server. The algorithm divides the priority of services according to the maximum tolerated delay of services, and determines the ratio of long and short services according to the standard deviation of minimum completion time to realize the pre allocation of services. Based on the load of each edge server, the algorithm then allocates the services on the high load edge servers to the low load servers to achieve secondary redistribution. Simulation results show that, compared with typical algorithms, the proposed algorithm alleviates the above problems in the existing technologies, reduces computational latency, achieves load balancing, and improves system resource utilization.