Matryoshka: Joint Resource Scheduling for Cost-Efficient MEC in NGFI-Based C-RAN

In this paper, we consider MEC in NGFI-based C-RAN, a novel and practical MEC framework to facilitate the emerging mobile applications such as AR/VR and video surveillance. However, it is challenging to implement it in a cost-efficient manner (i.e., optimized operational expenditures and service performance), due to the coupled resource provision, service deployment and workload distribution. To solve this joint resource scheduling problem, we resort to rounding and decomposition to devise Matryoshka, a novel approximation algorithm with polynomial running time. Extensive data-driven simulations corroborate that Matryoshka achieves superior performance (e.g., 43% and 52% performance gain compared with two state-of-the-art works, CSPP and Octopus) and scales well to support a variety of system settings.