iPerfOPS: A Tool for Machine Learning-Based Optimization Through Protocol Selection

Previous work established the importance of selecting the right network protocol for new foreground traffic, based on the current background traffic. The interactions between protocols, such as TCP-CUBIC and TCP-BBR for congestion control, affect fairness and throughput on shared networks. Fortunately, machine-learned (ML) classifiers can be used to identify the current background protocols, then optimization through protocol selection (OPS) can be used to improve performance on shared wide-area networks (WAN). We describe the design, implementation, and evaluation of iPerfOPS, the first tool that uses OPS to perform bulk-data transfer. The new tool is a substantially modified version of the well-known iPerf tool, and is an end-to-end implementation that incorporates previous research results. iPerfOPS introduces (1) a reliable data-transfer capability to iPerf, and (2) an implementation of OPS. We describe some empirical evaluations of iPerfOPS and discuss some of the practical implementation details required to achieve high performance. iPerfOPS shows that it is possible, within one tool, to classify the background network protocols such that high throughput and fairness are achieved.