Stochastic model-driven capacity planning framework for multi-access edge computing

Multi-access edge computing (MEC) offers cloud computing capabilities and IT services situated at the Radio Access Network (RAN) in the mobile users' proximity. Applications could offload their computation-intensive tasks to the MEC servers. Consequently, MEC significantly diminishes the mean response time and job rejection probability compared to conventional Mobile Cloud Computing (MCC). Cost-performance trade-off is one of the major concerns of the system designers. Low performance leads to the Service Level Agreement (SLA) violation and disappoints the service consumers. On the other hand, reaching high performance by augmenting the number of servers in the MEC and Cloud sides incur more infrastructure and other operational costs. In this paper, we formulate the mentioned cost-performance trade-off into an optimization problem. We demonstrate that the optimization problem is integer and non-linear. Moreover, we propose a capacity planning framework to determine the optimal number of servers in the MEC and Cloud sides, minimizing the Total Cost of Ownership (TCO) with SLA satisfaction. The proposed capacity planning framework gains from the simulated annealing algorithm to obtain a globally optimum solution. Furthermore, we deploy the stochastic performance model to measure mean response time and job rejection probability at each iteration. Numerical results reveal that the proposed framework determines the optimal solution within a reasonable time.