Availability-Aware VNF Placement and Request Routing in MEC-Enabled 5G Networks

TL;DR

This paper addresses VNF placement and request routing in MEC-enabled 5G networks for uRLLC applications, proposing approximation algorithms that balance availability, latency, and resource efficiency, with validated simulation and prototype results.

Contribution

It introduces a novel randomized rounding and greedy heuristic approach for VNF placement that considers availability constraints in 5G MEC networks.

Findings

Proposed algorithms achieve within 5-10% of optimal total reward.

Existing schemes perform poorly without availability considerations.

Redundant VNF placement improves packet delivery and latency in prototypes.

Abstract

In this paper, we study the virtual network function (VNF) placement problem in mobile edge computing (MEC)-enabled 5G networks to meet the stringent reliability and latency requirements of uRLLC applications. We pose it as a constrained optimization problem, which is NP-hard, to maximize the total reward obtained by a network service provider by serving uRLLC service requests. We propose an approximated randomized rounding approach to solve the NP-hard optimization problem in polynomial time. We prove that the proposed randomized approach achieves performance guarantees while violating the resource constraints boundedly. Furthermore, we present a greedy-heuristic approach to tackle the violations of resource constraints. Simulation results show that the proposed randomized rounding and greedy approaches achieve a total reward which is within 5% and 10% of the optimal solution, respectively. Furthermore, we compare the proposed greedy approach with the existing schemes that do not consider the availability requirements. We observe that the existing schemes perform poorly in terms of total reward, as negligence to the availability requirements negatively impacts the number of successfully served requests. These findings highlight the trade-off between availability and resource efficiency in latency-sensitive uRLLC applications. We also implement a software prototype of a 5G network using open-source software platforms with redundant placement of VNFs. The results on packet delivery ratio and latency obtained from the prototype implementation are also improved in the redundant VNFs with different failure probabilities.