Latency-aware and Survivable Mapping of VNFs in 5G Network Edge Cloud

TL;DR

This paper addresses the challenge of optimally placing Virtual Network Functions in 5G edge clouds by considering latency, survivability, and cost, proposing a heuristic solution for practical deployment.

Contribution

It introduces a novel ILP formulation for latency-aware and survivable VNF mapping in MEC, along with a simulated annealing heuristic to solve the NP-hard problem efficiently.

Findings

The heuristic achieves near-optimal solutions in polynomial time.

The approach effectively reduces latency and improves survivability in simulated 5G edge scenarios.

Results demonstrate cost-effective VNF placement close to optimal solutions.

Abstract

Network Functions Virtualization (NFV) and Multi-access Edge Computing (MEC) play crucial roles in 5G networks for dynamically provisioning diverse communication services with heterogeneous service requirements. In particular, while NFV improves flexibility and scalability by softwarizing physical network functions as Virtual Network Functions (VNFs), MEC enables to provide delay-sensitive/time-critical services by moving computing facilities to the network edge. However, these new paradigms introduce challenges in terms of latency, availability, and resource allocation. In this paper, we first explore MEC cloud facility location selection and then latency-aware placement of VNFs in different selected locations of NFV enabled MEC cloud facilities in order to meet the ultra-low latency requirements of different applications (e.g., Tactile Internet, virtual reality, and mission-critical applications). Furthermore, we also aim to guarantee the survivability of VNFs and an edge server against failures in resource limited MEC cloud facility due to software bugs, configuration faults, etc. To this end, we formulate the problem of latency-aware and survivable mapping of VNFs in different MEC cloud facilities as an Integer Linear Programming (ILP) to minimize the overall service provisioning cost, and show that the problem is NP-hard. Owing to the high computational complexity of solving the ILP, we propose a simulated annealing based heuristic algorithm to obtain near-optimal solution in polynomial time. With extensive simulations, we show the effectiveness of our proposed solution in a real-world network topology, which performs close to the optimal solution.