An automated parameter optimizer for data transfer performance testing

Abstract

This work presents an automated tool for optimizing control parameters in performance testing of big data transfer over long-fat network connections. Supporting both TCP and UDT protocols, the tool identifies the optimal configurations to enhance the efficiency of large-scale data transfers. A stochastic approximation algorithm is employed for parameter optimization, streamlining the protocol and parameter selection. The tool has been evaluated in various network scenarios, including long-haul connections in real-world high-performance networks. Its modular design also enables straightforward integration of additional data transfer protocols and alternative optimization methods.