UCAT: User Centric Adaptive Transmission for Meeting Diverse Network Demands

With the development of network, many emerging network applications such as telemedicine and Virtual Reality (VR) have put forward higher requirements on network bandwidth. However, the conventional coarse-grained service model of networks faces challenges in meeting the diverse and dynamic bandwidth requirements posed by these applications. For instance, while simple text browsing demands minimal bandwidth, bandwidth-intensive applications like VR require substantially higher resources. In this paper, we propose UCAT, an efficient multipath transmission architecture for user bandwidth demand adaptation. Specifically, UCAT leverages in-band network telemetry (INT) to collect network status information and generate multiple forwarding paths, which splits user traffic into flowlets and performs weighted scheduling. By customizing the packet transmission format using P4, UCAT achieves bandwidth demand awareness, enabling users with a single network interface card to achieve higher transmission bandwidth. Extensive experimental results show that UCAT can effectively meet user bandwidth requirements, reduce disordered packets, and improve network throughput.