MOClusVNF
Jing Chen, Jia Chen, Renkun Hu · 2019
Network functions virtualization (NFV) is becoming a prevailing design for future Internet by decoupling software implementations of network functions from their hosts or hardware. One crucial issue in NFV is to improve resource allocation problem that saves cost and to improve the flexible management of telecommunications service providers (TSPs). Orchestrating VNFs includes VNFs placement (network-level) and routing of flows (flow-level) through ordered VNFs. The challenge of developing NFV orchestration (NFVO) scheme is to improve utilization of the available network resources while meeting the performance requirements of flows. In this paper, multi-objective clustering based VNF (MOClusVNF) is proposed which can balance the tradeoff between multi-objective including maximizing the number of flows admitted to the network, minimizing the path stretch, and balancing the load among VNF instances (VNFIs). MOClusVNF is proposed to cluster VNFs based on their correlation, and determine whether to launch new VNFI or reuse established VNFI to improve the utilization of VNFIs. Moreover, VNFIs in the same cluster are deployed on the same server or its neighbor servers with one and two hop distance to reduce occupied link bandwidth. Simulation results show that the proposed MOClusVNF can balance the load among NFVIs while maximizing the number of flows admitted to the network. Moreover, the simulation results show the scheme can reduce the occupied link bandwidth utilization by 7.5% comparing with greedy method without violating the service level agreement (SLA) for flows.