Application-Driven Traffic Differentiation

Networked application typically open multiple connections to one or more servers over which information is transported to maintain the virtual objects that the application user interacts with. These connections share a common bottleneck that is restricted in capacity. One of the main problems is to share the bottleneck capacity via a congestion control algorithm that decides at which throughput can be supported over each connection. For the (many) applications that require low latency transport, the recently proposed ``Low Latency Low Loss Scalable'' (L4S) standard offers a solution. One undesired feature of L4S in this context is that it provides the same capacity share to each of the connections. In the present paper we extend L4S to enable the application to decide how the available bottleneck capacity should be shared between the connections that it maintains, still guaranteeing low latency.