Collaborative Service Placement for Mobile Edge Computing Applications

Mobile edge computing (MEC) can improve the quality of services and save the bandwidth of backhual networks, by placing application services in the base stations (BSS), which are endowed with computing resources and are in close proximity to user equipments (UEs). Since the capacity of an individual BS is limited, only a small number of service instances can be allowed for each BS at the same time. Meanwhile, in a densely deployed network, the coverage areas of adjacent BSS are overlapped. Therefore, these capacity-limited BSS can collaboratively optimize their service placements to improve the performance of MEC. In this paper, we investigate the collaborative service placement (CSP) problem in MEC, which aims to minimize the traffic load caused by service request forwarding. The CSP problem involves several difficult issues, including correlations of adjacent BSS' service placement decisions, joint service placement and UE association, and joint allocation of computing and radio resources. This makes the CSP problem be a complex combinatorial optimization problem. To solve the CSP problem, we propose an efficient decentralized algorithm based on the Matching Theory. It can optimize the decisions of service placement and BS-UE association for BSS, according to local interactions between BSS and UEs. Our proposed algorithm is practical for large-size networks, and its effectiveness is demonstrated by the simulation results.