A routing and placement scheme for network function virtualization

Network Function Virtualization (NFV) and Software-Defined Networking (SDN) are two paradigms that have attracted much attention in the networking field. The first permits the implementation of Network Functions (NFs) on commodity servers located in datacenters. The second facilitates the management and routing of network flows by controllers. While recent work has mostly explored the use of NFV and SDN with the goal of minimizing the resources to satisfy a set of requested NFs, the application of these two paradigms has not been studied in scenarios where resources are limited. Those scenarios are typical when parts of the network or datacenters fail. Hence, this paper presents an optimization scheme based on integer linear programming (ILP) for the joint routing and placement of virtual NFs problem. Given a set of requests, each consisting of NFs and end points of the traffic flow, the objective of the scheme is the maximization of the number of NFs satisfied. At the same, the scheme minimizes both the routing and infrastructure costs to satisfy the requests. While numerical results demonstrate that the proposed ILP can be used in small/medium-sized networks, the paper also presents a low-complexity greedy heuristic approach for large networks.