Let's Share the Resource When We're Co-Located: Colocation Edge Computing

• Published in: IEEE transactions on vehicular technology 2020-05, Vol.69 (5), p.5618-5633
• Main Authors: Nguyen, Minh N. H., Zaw, Chit Wutyee, Kim, Kitae, Tran, Nguyen H., Hong, Choong Seon
• Format: Article
• Language: English
• Subjects: Algorithms; Bandwidth; bandwidth allocation; Bandwidths; Channel allocation; Colocation edge computing; Computational modeling; Computer simulation; Convergence; Data centers; Delays; Edge computing; Energy consumption; Greedy algorithms; Mobile communication systems; Mobile computing; Network latency; Operating costs; Operators; Radio equipment; server scaling management; Servers; user association; Wireless networks
• ISSN: 0018-9545; 1939-9359
• DOI: 10.1109/TVT.2020.2982679

Multi-access Edge Computing (MEC) is recently acknowledged as one of the key pillars for the next revolution of mobile communications area, where the convergence of IT and telecommunications network provides the low latency and computation capability for cellular base stations (BSs). As a result, this is a great opportunity for mobile network operators by deploying new services and applications at BSs. Nevertheless, huge capital and operational cost can challenge mobile operators for the deployment of new BSs and MEC micro-datacenters. Colocation Edge Computing (ColoMEC) is a new concept where multiple operators share not only the same BS tower but also their radio and computation resources colocated at the edge sites. In order to reduce the operational cost of a ColoMEC system, the limited bandwidth at over-utilized colocation BSs can be extended by sharing the bandwidth among BSs, while shared MEC micro-datacenters can be scaled based on the arrival traffic loads. Thus, sharing the BS infrastructure, bandwidth, and MEC micro-datacenters among the co-located mobile operators can be an economical solution to provide high-performance services with low expenses by exploiting the temporal and spatial difference in traffic loads. Turning this vision into reality, we study a joint bandwidth allocation sharing and MEC micro-datacenter scaling in ColoMEC management problem (\mathbf {ColoMEC-MP}). To solve \mathbf {ColoMEC-MP} problem, we propose an algorithm based on proximal block coordinate descent technique by iteratively solving the decoupled convex subproblems (i.e., user association, bandwidth allocation, and MEC micro-datacenter scaling) with additional proximal terms. To improve the convergence of the proposed algorithm, we propose a greedy initialization for the user association which is based on the link capacity at each user. Our simulation demonstrates the superiority of the algorithm in terms of the operational cost compared with fixed service rate of shared MEC micro-datacenters strategies.