Time-dependent networking (TDN)
The design of the Internet has many integral parts that has a negative effect on the perceived latency for the end user. In our work on time-dependent networking, we seek to reduce the latencies for time-dependent applications without redesigning the existing Internet infrastructure.
Many interactive applications produce signalling traffic upon events generated by the users or other communicating systems. This makes the traffic patterns likely to be different from what we see in bulk traffic generated by moving large quantities of data. The time-dependent applications often generate packets with high intertransmission time and small packet sizes, what we call “thin streams”. One of our main fields of focus is to study how such streams fare in the internet, how competing traffic adversely affect the latency for such streams and how we can reduce the experienced latency for such traffic.
For reliable protocols like TCP, the way retransmissions are handled affects the recovery latency for thin streams. We are working on reducing the experienced latency from recovery by making TCP mechanisms that are more attuned to thin-stream traffic. We also work on applying redundancy as a way of reducing latency.
Our ultimate goal is that to make incremental changes to the current Internet architecture that will reduce the experienced latency, aiming for near-immediate response for time-dependent applications. Currently, we are pursuing this topic through the RITE EU project (www.riteproject.eu) and the TimeIn project.