A First Step Towards Production-Ready Network Function Storage: Benchmarking with NFSB

Abstract

Recently, multiple research teams aimed to address the scalability and fault-tolerance issues of Network Functions (NFs) by implementing a storage system (or distributed shared object) to share state across NF instances. Each system demonstrated promising performance under specific scenarios. However, there are many more technical challenges to be addressed to build a production-ready storage system for NFs.In this paper, we enumerate these challenges, explain how some of them were addressed by previous systems, and discuss alternative promising solutions to tackle them. Most challenges remain as open problems, while others deserve more thorough study. The first goal of the paper is to provide a broad research agenda around this topic, by discussing more problems than solutions. Moreover, we advocate that the first step towards a production-ready system is to build a realistic yet extensible benchmarking tool to deploy, test, and analyze "network function storages" comprehensively. Toward this end, we present our tool – Network Function Storage Benchmarking (NFSB). Finally, we present some preliminary results using NFSB.