A Pragmatic and Realistic Approach for Providing Adapted Transport Services for All Applications

Abstract

Transmission Control Protocol (TCP) services are not suitable for all Internet applications. Also, the performance provided by TCP is suboptimal for many networks. Given this situation, several alternative protocols and new transport layer architectures have been proposed to provide efficient end-to-end communication services for QoS-constrained applications. Yet, these solutions are not deployed or adopted because they only target new applications or force the rewriting of classic socket-based (aka “legacy”) applications. At the same time, new communication paradigms (SDN/NFV, etc.) and technologies (eBPF, DPDK, RDMA, etc.) have recently emerged. These evolutions can break the ossification of the Transport layer. However, they are unequally distributed on the Internet, posing a heterogeneity problem. This article proposes a software framework (named VTL for Virtual Transport Layer) that aims to take advantage of these new technologies according to their availability, with the goal of offering the most appropriate Transport services according to the application and network context. Relying on existing protocols or Transport layer frameworks, the VTL makes it possible to deploy the best protocol solution on the end hosts or the data paths in response to the QoS needs of the applications, whether or not they are aware of the existence of the VTL.