Modeling and analysis of different reconfiguration strategies for virtual network function placement and chaining with service classes identification.

Abstract

The Virtual Network Function Placement and Chaining problem (VNF-PC) is an important part of the Network

Functions Virtualization (NFV) based-technologies implementation. VNF-PC problem focuses on the allocation of customer

demands on the Substrate Network. Aiming to evaluate the impact of diverse modeling, various reconfiguration strategies based

on implicit steps in solving the VNF-PC are proposed: resizing virtual network instances, re-routing chaining, and repositioning

Network Functions (NF) instances on different servers. In addition, this work analyzes, compares, and discusses the advantages

and disadvantages of each proposed reconfiguration strategy in an online scenario. Traditionally, VNF-PC solutions from literature process requests generated at random and do not take into account real-world demands. Complementing the analyses of reconfiguration strategies, works from the literature are surveyed to identify commonly used Network Services (NS). Following that, these NS are classified into service classes, and used to generate realistic requests to be mapped in the experimental stage of the reconfiguration approaches. The experiments are conducted using realistic requests and a real-world network topology. An Integer Linear Programming model is used to process the requests. Simulations show that repositioning NF instances can generate up to ≈ 25% more profit than reconfiguring only the VNF instances, but the processing time increases by up to 99.99%.

On the other hand, resizing virtual network instances and rerouting the chaining had no significant impact on runtime.