With the development of new technologies such as Access Traffic Steering, Switching and Splitting (ATSSS) proposed by the 3GPP, there is a growing interest in getting multipath capabilities within the QUIC protocol. Today, there was a call on the QUIC WG IETF mailing list to work on the standardization of its multipath extensions to serve this use case.
This is a good occasion to remind us how our multipath extensions proposal evolved during these last two years. While the initial draft modified the cleartext header to include a “Path ID”, the current one makes the use of multiple paths undetectable from an external viewpoint. The core multipath mechanisms blend naturally into QUIC’s current design and open new possibilities to leverage asymmetrical networks such as satellites.
The use of multiple paths introduces its own set of challenges, such as how hosts should take advantage of several network paths. With our previous experiences with Multipath TCP and Multipath QUIC, we believe these algorithms must be adapted to their serving use cases. A few days ago, I publicly defended my Ph.D. thesis about “Flexible Multipath Transport Protocols”. In my work, I explored various solutions to adapt the path management and the packet scheduling to both the application and the network conditions. I also describe the evolution of the design of the Multipath extensions for the QUIC protocol based on numerous experiments based on Experimental Design and the in-the-wild runs using the MultipathTester app.