Research On Service Placement And Service Migration Strategies In Mobile Edge Computing

Driven by the vigorous development of wireless communication technology and mobile Internet,the total number of Internet users and their network devices has increased dramatically,computing-intensive and latency-critical applications have emerged one after another as well.To support such applications on mobile devices with limited resources,the computing paradigm has evolved from centralized mobile cloud computing to distributed mobile edge computing.In mobile edge computing networks,edge devices with certain storage space and computing capability pre-place application services and process user requests corresponding to these services,to reduce service acquisition latency and relieve core network traffic.However,the storage space of an individual edge cloud is insufficient to store all services and its computing capability is also not enough to satisfy all requests.Secondly,the resource capacity of edge devices is heterogeneous,and their computing capability and storage capacity may not match,which may cause network congestion.On the other hand,due to the limited coverage of edge devices and end-user mobility,the user's Qo E may degrade and even interrupt ongoing services.Service migration is an effective approach to ensure Qo E,but it generates overhead in terms of signaling messages and data transfer between edge devices.Thus,this thesis mainly focuses on service placement and service migration in edge computing networks.The research content and results of this thesis are as follows:1.Research on joint service placement and request routing strategies.First,the cloud radio access network architecture is integrated into edge computing networks to establish a collaboration mechanism,and a two-timescale framework is embedded to achieve the trade-off between operating expenses and response timeliness,thereby the problem of joint service placement and request routing in edge computing networks is formulated as a pure integer linear programming problem,which maximizes system utility under storage and computing resource constraints of edge devices.Then,we not only equivalent the service placement subproblem to the submodular function maximization problem and design a greedy-based algorithm with approximate guarantee,but also transform the request routing subproblem into a multi-knapsack problem with assignment restrictions and apply a linear programming relaxation-based heuristics to solve it.Finally,the numerical results show that the proposed solution is always better than the benchmark algorithm,reaching90% of the optimal performance in services homogeneous case and 76% in services heterogeneous case.2.Research on mobility-aware service migration strategy.This thesis proposes a service migration prediction framework that determines when and where to perform service migration according to user mobility patterns.The framework consists of three schemes.One is a handoff time estimation scheme,which aims to estimate the transit time of each data center along the user's destination path.The second is a transfer throughput estimation scheme,which aims to estimate the data transfer throughput of each data center along the path.Based on the handoff time estimation scheme and the transfer throughput estimation scheme,the best data center with the highest average transfer throughput at a certain moment or during the time window can be determined.The third is a service migration management scheme,which divides user-requested service into several portions and migrates to these best data centers.Finally,the simulation results show that the proposed solution achieves the lowest data transmission latency compared with the benchmark algorithm and effectively reduces migration overhead.