DistNSE: Distributed network service embedding across multiple providers

Abstract

Network function virtualization (NFV) brings significant benefits by outsourcing network functions (NFs) to virtualized datacenters and enabling new business models, e.g. NF as a service (NFaaS). In this respect, network service embedding (i.e., the mapping of service chains) raises significant challenges due to the need to preserve correctness and satisfy NF location dependencies. The latter creates the need for service mapping across multiple NF Providers (NFPs), given the limited footprint of each NFP. This further exacerbates the network service embedding (NSE) problem, due to the NFPs' local policies and the confidentiality of network topology and resource availability. To address these issues, we present DistNSE, a distributed NSE framework that enables the collaboration among NFPs for NSE, while maintaining the privacy and autonomy of each NFP. DistNSE ensures competitive pricing by letting different NFPs compete for different NFs of a service chain. To this end, DistNSE decomposes NSE in two steps: (i) service chain partitioning using an algorithm that fosters competition across NFPs, and (ii) the assignment of chain segments to datacenter networks, while ensuring compliance with the NFP's policy. We further couple the proposed embedding methods with a communication protocol for interoperability and collaboration among the participating NFPs. We use simulations to assess the efficiency of DistNSE and identify significant gains over an existing distributed embedding framework (i.e., Polyvine).