QoS-Aware Joint Power Allocation and Task Offloading in a MEC/NFV-enabled C-RAN Network

Research paper by Mohsen Tajallifar, Sina Ebrahimi, Mohammad Reza Javan, Nader Mokari, Luca Chiaraviglio

Indexed on: 18 Aug '21Published on: 30 Nov '19Published in: arXiv - Computer Science - Networking and Internet Architecture


In this paper, we propose a novel resource management scheme that jointly allocates the transmit power and network function virtualization (NFV) resources in a centralized radio access network (C-RAN) architecture consisting of several RRHs and a BBU pool. The BBU pool is connected to a set of NFV-enabled nodes (including the multi-access\mobile edge computing (MEC)-enabled node) to which the requested tasks of user equipments (UEs) are offloaded. We formulate an optimization problem taking into account the transmission, execution, and propagation delays of each task with the aim to allocate the transmission power in radio and computational resources in nodes such that the user's maximum tolerable latency is satisfied. Since the proposed optimization problem is highly non-convex, we adopt the alternate search method (ASM) to achieve a sub-optimal solution. Moreover, a novel heuristic algorithm is proposed to jointly manage the allocated computational resources and placement of the tasks derived by ASM. We also propose a new admission control mechanism for finding the set of tasks that can be served by the available resources. Furthermore, a disjoint method that separately allocates the transmission power and the computational resources are proposed as the baseline of comparison. The simulation results show that the joint method outperforms the disjoint task offloading and power allocation. Specifically, for a specific setup, the joint method can provide 50 percent of acceptance ratio while the acceptance ratio of the disjoint method is 25 percent when the maximum tolerable latency of tasks is equally divided between power allocation and task offloading steps. Moreover, with the proposed heuristic algorithm, the value of acceptance ratio is equal to 1 for maximum tolerable latencies larger than 50 ms, while without the heuristic algorithm, the acceptance ratio decreases to 65 percent.