MPTCP Meets Big Data: Customizing Transmission Strategy for Various Data Flows

摘要

In recent years, network transmission carries big data of various types and sizes, which often have different transmission requirements. Transmission of data brings challenges to traditional transport layer protocols; meanwhile, performance improvements at endpoints and networks provide opportunities for new transport protocol design. Among the new designed transport layer protocols, multipath TCP is a promising one that provides better performance with multiple paths transmitting data simultaneously. However, when dealing with different types and sizes of data, MPTCP does not always work well, as all data streams are processed based on the same transmission strategy, which is simple but not always optimal. To fill this gap, in this article, we first give optimal strategies for different data flows, and propose a cross-layer solution for MPTCP, named FSA-MPTCP, to automatically provide flexible and optimal transmission strategies for different data flows. Finally, we conduct some experiments, and the results show that FSAMPTCP adaptively provides better performance for different kinds of flows.