ScaRL: Service Function Chain Allocation Based on Reinforcement Learning in Mobile Edge Computing

Network Function Virtualization (NFV) and Mobile Edge Computing (MEC) are the two core technologies in 5G, therefore the service function chain allocation (SFC-A) in MEC plays an important role in enhancing the quality of experience (QoE) of the end users. Different from the SFC-A in cloud computing, SFC-A in MEC is more challenging due to the limited computing resources at edge servers. In this paper, we investigated the SFC-A problem in MEC with limited resources, and formulate the problem as an integer linear program problem with the objective of minimizing both the transmission latency and processing latency. Furthermore, to overcome the inherent shortcomings of heuristic solutions, we propose an algorithm, ScaRL, which leverages reinforcement learning to make the optimal decision by taking advantage of its trial-and-error mechanism, reward mechanism and exploration-exploitation ability. A series of simulations verified that ScaRL is effective in reducing the average latency of SFC requests, as well as reducing the blocking rate to provide QoE guarantee.