Design And Implementation Of Network Function Virtualization Performance Interference Measurement System

Traditionally,network operators use dedicated hardware to provide network services.To save expensive infrastructure investment and operating expense(OPEX),as well as improve flexibility and efficiency of service deployment,network function virtualization(NFV)decouples network functions from the dedicated hardware and enables them running as applications on commodity servers.Due to VM's ease of duplication and migration,network operators tend to deploy virtual network functions(VNFs)in virtual machines(VMs).Efforts are paid to provide efficient VNF placement approaches,aiming at minimizing the resource cost of VNF deployment and reducing the latency of service chain.However,existing placement approaches may result in hardware resource competition and performance interference among co-located VNFs,leading to performance degradation.In this paper,we measure the performance interference among co-located VNFs to identify the influenced factors.From the aspect of packet processing,we classify VNFs into different types,and analyze how VNFs' combinations and the characteristics of flow affect the performance interference.To find the root causes of performance interference,we collect and analyze resource utilization to identify the competitive resources and quantify the effects of hardware factors.Through measurements,we find that the performance interference among co-located VNFs is ubiquitous,which causes performance degradation,in terms of VNFs' throughput,ranging from 12.36% to 50.3%.And we disclose that the contentions of network I/O bandwidth and inter-domain interconnect play key roles in performance interference.Based on our measurement results,we make quantitative and qualitative analysis,and give advice on how to mitigate performance interference among co-located VNFs and design more efficient VNF placement approaches.