Delay Sensitive Virtual Network Function Placement and Routing

Network function virtualization (NFV) is a recent concept for sharing multiple resources from one or more physical network infrastructures in order to improve flexibility of design, deployment and management of networking services. Virtual Network functions (VNFs) are hosted in virtual machines for the purpose of serving a user and/or a tenant request where the order of the requested functions are as important as the service itself. Therefore, the placement, chaining and routing of VNFs have a major impact on the overall networking cost which can be translated into financial cost, latency, resources utilization or other suitable metrics. In this paper, we focus on the accumulated delay assuming multipath routing of flows and the assignment of service chains in virtual networks. We formulate the problem of placement and routing as a Mixed Integer Linear Programming (MILP) problem, considering not only the delay cost generated by serving user and/or tenant demand, but also the cost of routing using a bi-objective optimization problem. The proposed solution aims to minimize the overall delay and the simulation results show a considerable gain in performance and a better use of network resources compared to conventional shortest path routing based solutions.