Packet-Scale Congestion Control

The Packet-Scale Congestion Control is a new paradigm that allows end-to-end transport protocols to scale to ultra-high speed networks.  While most existing protocols can attain a single-stream throughput of no more than a few Gigabits-per-second, this paradigm promises to scale to up to Terabit-and-higher speeds.

It does so by shedding the legacy framework of RTT-scale operation that is adopted by most existing congestion-control protocols — this allows protocols based on the packet-scale paradigm to operate at much finer timescales than are currently possible. The paradigm itself relies on two main ideas: (i) fine-scale probing that creates finely-controlled inter-packet spacing at the sender and observes changes in these at the receiver to estimate the current available bandwidth in the network, and (ii) probing-without-overloading, which exploits the fine-scale of probing to probe for a wide-range of rates within an RTT, without causing persistent queuing at bottleneck links.  The paradigm also helps truly achieve RTT fairness and friendliness to conventional TCP traffic — two goals that have so far remained allusive to high-speed transport protocols.

The paradigm, however, brings up new challenges that are not dominant ones for RTT-scale protocols. These include:

• The sensitivity of the paradigm to “noise” in the end-to-end delays experienced by packets.
• The implementation of fine-scaled inter-packet spacing in current end-systems.
• The stability, sensitivity, and fairness of the paradigm under stressful traffic conditions.

These are the subject of our current research in this project.