VNF Placement for Service Function Chains with Strong Low-Delay Restrictions in Edge Computing Networks

Pilar Manzanares-Lopez,Juan Pedro Muñoz-Gea,Josemaria Malgosa-Sanahuja

doi:10.3390/app10186573

Pilar Manzanares-Lopez, Juan Pedro Muñoz-Gea + Show 1 more

Open Access

https://doi.org/10.3390/app10186573

Copy DOI

Journal: Applied Sciences	Publication Date: Sep 20, 2020
Citations: 4	License type: CC BY 4.0

Affiliation: Universidad Politécnica de Cartagena

Abstract

The edge computing paradigm, allowing the location of network services close to end users, defines new network scenarios. One of them considers the existence of micro data centers, with reduced resources but located closer to service requesters, to complement remote cloud data centers. This hierarchical and geo-distributed architecture allows the definition of different time constraints that can be taken into account when mapping services into data centers. This feature is especially useful in the Virtual Network Function (VNF) placement problem, where the network functions composing a Service Function Chain (SFC) may require more or less strong delay restrictions. We propose the ModPG (Modified Priority-based Greedy) heuristic, a VNF placement solution that weighs the latency, bandwidth, and resource restrictions, but also the instantiation cost of VNFs. ModPG is an improved solution of a previous proposal (called PG). Although both heuristics share the same optimization target, that is the reduction of the total substrate resource cost, the ModPG heuristic identifies and solves a limitation of the PG solution: the mapping of sets of SFCs that include a significant proportion of SFC requests with strong low-delay restrictions. Unlike PG heuristic performance evaluation, where the amount of SFC requests with strong low-delay restrictions is not considered as a factor to be analyzed, in this work, both solutions are compared considering the presence of 1%, 15%, and 25% of this type of SFC request. Results show that the ModPG heuristic optimizes the target cost similarly to the original proposal, and at the same time, it offers a better performance when a significant number of low-delay demanding SFC requests are present.

Highlights

Edge Computing (EC) [1,2,3] is considered as a key supporting technology for the emergingInternet of Things (IoT) and 5G networks
The results show that the Modified Priority-based Greedy (ModPG) heuristic optimizes the target cost to the original proposal, and at the same time, it reduces the amount of non-allocated Service Function Chain (SFC) requests
We propose an alternative to the Priority-based Greedy (PG) heuristic that improves the Virtual Network Function (VNF) placement results, reducing the number of non-allocated SFC request (SFCr) and maintaining similar total cost

Summary

Introduction

Edge Computing (EC) [1,2,3] is considered as a key supporting technology for the emerging. Internet of Things (IoT) and 5G networks. Computing services are shifted to the edge of the Internet ideally within one hop from mobile devices and other smart devices [4]. The traffic demands of existing services and, most importantly, new services such as Virtual. Reality (VR), Augmented Reality (AR), public security, smart cities, or connected cars pose challenges to remote resource-rich computing centers or clouds. The cloud is often remotely located and far from the users, and the data transfer delays between users and the cloud can be long and unpredictable. Bringing services closer to the edge network reduces the backhaul costs and solves the low latency requirements of the services.

Objectives

Results

Conclusion