Joint Optimization of Stateful VNF Placement and Routing Scheduling in Software-Defined Networks

Abstract

With the widely adoption of innovative Software-defined Networking (SDN) and Network Function Virtualiza-tion (NFV) technologies, wireless network management is much simplified and network resources are possibly allocated in an optimal manner. In software-defined NFV networks, when unicast session flows are required to get across multiple instances of a Virtual Network Function (VNF), the consistency of internal VNF-instance state must be guaranteed while scheduling the traffic flows. Thus, how to allocate VNF instances while satisfying the quality-of-service (QoS) demand of multiple unicast sessions, and how to address routing-planing problem between client flows and control flows become the significant challenges. In this paper, we are motivated to study the joint optimization towards the state synchronization among VNF instances and routing scheduling, with the objective to minimize the VNF deployment cost and rule occupation cost. This problem is first formulated as a Mixed Integer Linear Programming (MILP) that is NP-hard. We then devise a two-phase fast algorithm that can deploy VNF instances and select appropriate routing paths for the different flows. Finally, extensive evaluation results show that the proposed two-phase algorithm outperforms the benchmark algorithms significantly.