移动边缘计算网络中基于资源联合配置的计算任务卸载策略

发布时间：2018-04-10 01:01

本文选题：计算卸载　切入点：移动边缘计算网络　出处：《北京邮电大学》2017年硕士论文

【摘要】：随着移动互联网的发展,智能移动终端已经成为人们生活必不可少的工具,移动应用的功能越来越强大,可以满足人们生活、出行、购物、娱乐等方方面面的需求。但是移动终端的计算资源、存储资源和电池容量有限,并不能满足各种应用的需求。如何解决移动终端资源受限与应用需要消耗大量资源的矛盾成为移动通信网络所面临的重要挑战。目前,计算卸载技术可有效解决上述问题。该技术可将部分计算任务卸载到附近的云服务器上来运行,具有计算任务需求的终端只需发送计算任务并接收计算结果,无需占用本地计算及存储资源,因此可以有效解决移动终端资源受限问题。此外,计算卸载技术可以增强移动终端的续航能力。移动边缘计算网络由于其能够提供较强的边缘计算及存储服务,具有服务时延低等优势,因此在其上运行计算卸载技术,能够为用户提供高效的服务。然而,在该网络架构中,如何联合考虑无线传输资源和云端计算资源状态,为用户选择合适边缘计算节点来服务,值得研究。针对上述问题,本文主要工作内容如下:第一,提出了基于传输和计算资源联合决策的计算任务卸载方案。本文首先建立了移动边缘计算(Mobile Edge Computing, MEC)服务节点选择模型,该模型以最小化用户接入MEC服务节点的传输代价为目标,联合考虑了传输与计算资源的分配来进行计算任务卸载。其次,提出了基于传输和计算资源联合决策的计算任务卸载算法,该算法考虑了用户间公平性。最后,通过仿真证明了该算法的有效性,该算法能实现较低的传输代价,提高用户的公平性。第二,在考虑MEC服务节点资源受限条件下,提出了多用户、多MEC服务节点匹配方案。本文首先描述了多用户、多MEC服务节点的网络场景。其次,建立了基于资源受限的MEC服务节点匹配模型,该模型以拍卖理论为基础,目标是优化MEC服务节点收益,同时充分利用MEC服务节点资源。再次,提出了基于二维资源拍卖的用户-服务节点匹配算法,以为用户寻找到合适的计算任务卸载节点。最后通过仿真证明了该算法的有效性,该算法能使MEC服务节点在充分利用资源的同时获得较高的收益。
[Abstract]:With the development of mobile Internet, intelligent mobile terminal has become an indispensable tool in people's life. The function of mobile application is becoming more and more powerful, which can meet the needs of people's life, travel, shopping, entertainment and other aspects.However, the computing resources, storage resources and battery capacity of mobile terminals are limited and can not meet the needs of various applications.How to solve the contradiction between the limitation of mobile terminal resources and the need to consume a large amount of resources in applications has become an important challenge for mobile communication networks.At present, the calculation and unloading technology can effectively solve the above problems.This technique can unload some computing tasks to the nearby cloud server to run. The terminal with computing task only needs to send the calculation task and receive the calculation result, so it does not need to occupy the local computing and storage resources.Therefore, it can effectively solve the problem of resource limitation of mobile terminal.In addition, the computing unload technology can enhance the mobile terminal's ability to live.Mobile edge computing network can provide strong edge computing and storage services, and has the advantages of low service delay, so running computing uninstall technology on the mobile edge computing network can provide efficient services for users.However, in the network architecture, it is worth studying how to consider the state of wireless transmission resources and cloud computing resources, and select the appropriate edge computing nodes for users.The main work of this paper is as follows: firstly, a scheme of computing task unload based on the joint decision of transmission and computing resources is proposed.In this paper, a mobile Edge computing node selection model for mobile edge computing is established. The model aims at minimizing the transmission cost of users accessing MEC service nodes, and considers the allocation of transport and computing resources to unload computing tasks.Secondly, a computing task unload algorithm based on the joint decision of transmission and computing resources is proposed, which takes into account the fairness between users.Finally, the effectiveness of the algorithm is proved by simulation. The algorithm can achieve lower transmission cost and improve the fairness of users.Secondly, considering the resource constraints of MEC service nodes, a multi-user and multi-#en1# service node matching scheme is proposed.This paper first describes the network scenarios of multi-user and multi-MEC service nodes.Secondly, a resource-constrained MEC service node matching model is established. The model is based on auction theory and aims to optimize the MEC service node revenue and make full use of MEC service node resources.Thirdly, a user-service node matching algorithm based on two-dimensional resource auction is proposed to find the appropriate computing task to uninstall the node for the user.Finally, the effectiveness of the algorithm is proved by simulation, and the algorithm can make the MEC service node make full use of the resources and obtain a higher profit at the same time.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN929.5;TP393.0

【相似文献】