Using UDP for Internet Transport Evolution

The increasing use of middleboxes (e.g., NATs, firewalls) in the Internet has made it harder and harder to deploy new transport or higher layer protocols, or even extensions to existing ones. Current work to address this Internet transport ossification has led to renewed interest in UDP as an encapsulation for making novel transport protocols deployable in the Internet. Examples include Google's QUIC and the WebRTC data channel. The common assumption made by these approaches is that encapsulation over UDP works in the present Internet. This paper presents a measurement study to examine this assumption, and provides guidance for protocol design based on our measurements.
The key question is "can we run new transport protocols for the Internet over UDP?" We find that the answer is largely "yes": UDP works on most networks, and impairments are generally confined to access networks. This allows relatively simple fallback strategies to work around it. Our answer is based on a twofold methodology. First, we use the RIPE Atlas platform to basically check UDP connectivity and first-packet latency. Second, we deploy copycat, a new tool for comparing TCP loss, latency, and throughput with UDP by generating TCP-shaped traffic with UDP headers.