Abstract

Nowadays, network function virtualization (NFV) receives widespread concerns from the whole society for its promising application. The main challenge for the deployment of VNF that comes along is the resource allocation of demanded network services in NFV-based network infrastructures. Current offline resource mapping and scheduling algorithms are impractical for continuous VNF requests in data centers, because of the high time consumption. In this paper, we firstly present a resource sharing schema for VNF, including the automatic monitoring and fast switching mechanisms, which support multiple VNFs effectively share resource in one node. This VNF resource sharing schema improves the acceptance ratio of VNF requests as well as the system utilization. And then, we propose an automatic NFV resource allocation mechanism AutoVNF to optimize the VNF deployment. AutoVNF includes mapping and scheduling algorithms, which automatically allocates available nodes to VNF requests and schedules the execution of VNFs in one VNF queue. Simulations show that AutoVNF has a good performance on acceptance ratio, average flow time, and total cost compared with well-known scheduling algorithms.