Migration of VNF instances for service continuity

Abstract

Fog computing computes and stores data locally based upon distributed architectures that extend cloud to the edge of the network. This enables fog paradigms to be suitable for delay-sensitive applications that demand various network functions (NFs). These NFs here are provided using network function virtualization (NFV) to allow elastic NF deployment on the fog nodes, that is, virtual network functions (VNFs). Now a major challenge for NFV-based fog architectures is the limited resources, which vulnerate them to premature saturation and load imbalance. Hence resource management schemes are necessary for fog networks, in particular load migration and redistribution. Along this, an efficient migration scheme is proposed here using the coordinated generalized pattern search (CGPS) heuristic method. The scheme migrates excess load from heavily-loaded (HL) nodes to lightly-loaded (LL) nodes in proximity. The scheme adopts two criteria, termed as exclusive and common use of VNF instances. Results show that the proposed scheme achieves high migration success at reduced delays, as well as alleviating resources with low migration frequency.