Orchestrating network function virtualization platform: Migration or re-instantiation?

Abstract

Network function virtualization (NFV) provokes the evolution of network functions to overcome various challenges facing the network service providers (NSPs). To exploit the advantages of the virtualization technology, NFV platforms should use the cloud environment to provide their services. Typically, an NFV service is represented by a service function chain (SFC) that consists of multiple virtualized network functions (VNFs). Hosting and orchestrating these VNFs in a cloud environment are challenging tasks. In this paper, we discuss the VNF orchestration problem from the perspective of VNF's migration and re-instantiation mechanism to achieve carrier grade-aware NFV services in a cloud-based platform. This paper also provides detailed insights on the NFV system modeling, building blocks, and various challenges hindering its cloud adoption. Also, a novel mixed integer linear programming (MILP) optimization model is proposed as a solution to facilitate the NFV platform orchestration in a cloud environment. The model decides between triggering either VNF's migration or re instantiation while achieving minimal downtime of the VNF, satisfying carrier grade requirements, and finding an optimal placement for the migrated or re-instantiated VNF that minimizes the SFC delays. The proposed model is compared to two availability-agnostic greedy algorithms. The simulation results show that finding an optimized decision whether to migrate or re-instantiate a VNF while associating it with an optimal placement can achieve a minimal VNF's downtime and SFCs delays and can enhance the quality of service accordingly.