Optimizing Operational Costs of NFV Service Chaining

Network Function Virtualization (NFV) is a novel paradigm that enables flexible and scalable implementation of network services on cloud infrastructure. One of the key factors in the success of NFV is the ability to dynamically allocate physical resources according to the actual demand. Yet, orchestrating service chains that require high traffic throughput is a very complex task and most existing solutions either concentrate on hand crafted tuning of the servers' configuration to get the desire performance (using hardware accelerators such as DPDK or SRIOV) or on theoretical placement functions that assumes that the operational cost is part of the input. In this work we try to bridge this gap by studying near optimal deployment strategies for such service chains that provide the needed performance yet minimizing the operation resource overhead. Since many network functions deals with data plane, the cost associated with virtual switching is an important factor of the overall operational cost. The amount of resources required to support virtual switching depends on the way service chains are deployed (implying intra- or inter-servers) and the amount of network traffic that goes through the service chains. In this paper, we present an algorithm to deploy service chains in NFV-based infrastructure that both guarantees the needed performance and minimizes the operational cost of the environment (and thus achieves a better utilization of the available resources). The algorithm is based on extensive measurements of actual cost and performance of service chains in a realistic NFV environment. Our performance evaluation indicates that the new algorithm significantly reduces virtual switching resource utilization when compared to the de-facto standard placement used in OpenStack/Nova. Moreover, we show that these resources can be used to support more services, thus allowing a much better acceptance ratio of network services.