期刊
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
卷 71, 期 3, 页码 3144-3154出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TVT.2021.3139026
关键词
Task analysis; Quality of service; Bandwidth; Scheduling; 5G mobile communication; Resource management; Uplink; 5G networks; multi-access edge computing; network slicing; packet scheduling
5G networks aim to deliver low latency data and support diverse services, leading to the development of the MEC paradigm. In 5G MEC systems, communication and computing resources can be partitioned among network slices to support various applications. The use of the EWFQ/LC scheduler improves system fairness and meets latency constraints efficiently.
The fifth generation (5G) cellular networks aim to deliver data with low/ultra-low latency to users and support diverse services which require two main types of resources: communication and computing resources. These motivated the development of the Multi-access Edge Computing (MEC) paradigm because it can execute tasks' computation closer to users to reduce latency. To satisfy quality of services (QoS) requirements of different applications, both communication and computing resources in 5G MEC systems can be hard partitioned or soft partitioned among multiple network slices where each slice supports one or more services. A typical soft partition approach is using scheduling. Furthermore, packets in the multi-resource system can either be scheduled individually for each resource using a discrete resource scheduling approach or scheduled collectively only once by combined resource scheduling. In this paper, we propose a combined resource scheduler, called Extended Weighted Fair Queueing with Latency Constraint (EWFQ/LC), to schedule packets among network slices subject to system fairness and their latency constraints. By exploiting the virtual finish time feature inherent in weight fair queueing, EWFQ/LC schedules packets according to the fairness when their latency constraints can be met while according to these latency constraints when they are possibly violated. Based on simulations of realistic heterogeneous traffic scenarios, we show that EWFQ/LC reduces latency by as much as 73% compared to scheduling each resource discretely and maintains latency satisfaction ratio above 90%. More importantly, EWFQ/LC is able to accomplish these improvements with greater fairness in resource consumption especially under heavy traffic conditions, i.e., traffic with stricter latency constraints do not excessively over-consume resources of other traffic classes.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据