Identifying performance bottlenecks in software data planes for cloud-based NFV services

Abstract

Network Function Visualization (NFV) is transforming the market for computer networks. Most proposed NFV solutions have been implemented and tested in cloud computing environments. In this context, both hardware and software-based features have been used to improve the performance of Virtual Network Functions (VNFs) by speeding up packet processing. However, there are still essential research challenges that need to be tackled to provide better performance experiences for NFV Services, such as detecting and diagnosing performance bottlenecks. However, due to the characteristics inherited from both Cloud and NFV environments, the detection and diagnose of performance problems is a complex task. In this work, we proposed PerfChecker, a monitoring tool that aims at detecting and diagnosing performance bottlenecks in Cloud-based NFV environments. We implemented a PerfChecker prototype for OpenStack and performed some experiments demonstrating that it can assist the cloud infrastructure operator to improve the performance of NFV services.