review of "Congestion Avoidance and Control"

From: Jenny Liu (jen@cs.washington.edu)
Date: Sun Oct 17 2004 - 12:38:26 PDT

  • Next message: Shobhit Raj Mathur: "review of paper 9"

    "Congestion Avoidance and Control" introduces five algorithms for incorporation into TCP based on the idea that it is desirable to maintain conservation of packets along a network. The paper finds exactly 3 ways in which conservation of packets can fail, and addresses each of them in turn. The observation that the conservation of packets property implies robustness in the face of congestion is a good one. Having the end points respond to congestion issues minimizes complexity in the system and eliminates the need to keep state about the rate of congestion within the network. However, there are some drawbacks to having congestion avoidance happen at the endpoints as well. Because the host uses dropped packets to figure out that the network is congested, there may be some delay between when congestion occurs and when the host actually figures out that congestion has occured. Also, putting congestion avoidance and control at the endpoints leaves the network vulnerable to malicious hosts who don't play by the rules and who might try to flood the network on purpose. Putting more of the responsibility for congestion avoidance and control inside the network rather than at the endpoints might help to strike a balance betweeen the trade-offs of having all the responsiblity at either the ends or the middle.


  • Next message: Shobhit Raj Mathur: "review of paper 9"

    This archive was generated by hypermail 2.1.6 : Sun Oct 17 2004 - 12:38:23 PDT