Reasonably Migrating Virtual Machine in NFV-Featured Networks

Abstract

Combining with software-defined networking and virtualization technology, Network Function virtualization (NFV) has been proposed as an important technology for constructing scalable network. VNF (Virtual Network Function) redeployment is a critical step for dealing with network evolve. Due to the inner state consistency constraints, VNF migration to a new location is major challenge for redeployment. Previous works about VNF migration primarily focused on VNF internal state transfer, which required reprograming VNF software and designed new transfer protocols. Extra work and more bugs could be introduced due to software develop workload. In this paper, we propose to migrate the entire VM (Virtual Machine) rather than VNF to simplify the development workload. Combining with Consolidated Middlebox, it is reasonable to migrate the VM which contains corresponding VNF to adapt to network evolve. However, transfer overhead introduced by VM migration should be minimized. Therefore, we tend to fulfill good VNF migration and minimize the time cost for the migration, meet the network bandwidth requirement as well. As this optimization problem is NP-hard, two heuristic algorithms are proposed aimed at various network scenarios. The paper is concluded with the simulation results and future direction.