Netset: Automating Network Performance Evaluation

Performance measurement and comparison is integral to almost any kind of networking research. However, we currently lack a general framework under which network-based tests can be carried out. Thus researchers tend to design their tests based on their own ingenuity, making it difficult to repeat these tests and compare testing results from different sources. This has seriously stymied innovation in the networking research community. New ideas are difficult to be verified and then widely deployed since different tests produce different or even conflicting results. In this paper, we propose Netset, a software framework to automate network performance evaluation. It makes network-based tests portable, repeatable, convenient and gives realistic and comparable results. It leverages existent models and tools but is also extensible. We have implemented a proof-of-concept prototype of Netset and used it to carry out some TCP tests. Our initial experiences with Netset demonstrate its advantages. We believe that a standard benchmark tool like Netset will facilitate a more objective evaluation of new network protocols and benefit the whole networking research community.