BOF: Kernel networking and TCP's (ir)relevance in the latest brave new world
Session information has not yet been published for this event.
One Line Summary
Let's talk about the latest trend in transport protocol development which is to expressly avoid relying on the kernel
Defining transport protocols running on userspace over UDP is an emerging trend. QUIC is starting to be deployed, SPUD is being discussed in IETF, there are efforts to implement TCP and SCTP over UDP in userspace. A key motivation for this trend is to avoid the kernel like the plague. As stated in the QUIC wiki: “As a userspace transport atop UDP, QUIC allows innovations which have proven difficult to deploy with existing protocols as they are hampered by legacy clients and middleboxes, or by prolonged Operating System development and deployment cycles.”
In this BOF, we will raise at least two questions
1) Is there anything we can do from the kernel perspective to address the “prolonged development and development cycles” issue?
2) Other than continuing to provide functional UDP sockets, is there anything we should do in the kernel to support transport protocols in userspace model?
TCP, networking, QUIC, SPUD, UDP
Software engineer at Facebook and Linux netdev contributor.