In today’s Internet, the predominant number of hosts have multiple network devices built-in. Typically, each of these devices provides at least one path to access the Internet. In Addition to that, content is ofen availiabe from diffent endpoints, e.g., different CDN sites. Finally, there are ofen multiple protocol alternatives or variants available, e.g., IPv4 and IPv6, TCP or QUIC over UDP.
Each path, endpoint, or element of a protocol stack available adds an option the endpoint can choose from. Thus, we call each of them a transport option. For each communication or transfer, an endpoint can choose a set of transport options which we then refer to as a transport con guration.
The goal of this project is to evaluate how transport diversity can be exploited to improve applications' network performance from within the client. It analyses the overall design space, but focuses the evaluation on automated, application-aware path selection.
In the past two years, ITEF QUIC has gained a lot of attention and has a good chance of replacing TCP as transport for many web services. With its multiple streams within a single connection de-coupling of retransmission and congestion control, it is an ideal platform for future transport evolution.
One of the most traffic-heavy kinds of web services is video streaming sites as Youtube. Despite video streaming, in principle, does not need reliable transport, these services usually operate over HTTP/TCP.
The goal of this project is to add unreliable and partial reliable transport to QUIC and evaluate its benefits on video streaming QoE.