1 / 15

TCP/Vegas

TCP/Vegas. Yang Richard Yang 10/1/2001. Review AIMD Congestion Control. What Are the Desired Properties of a Congestion Avoidance Scheme?. Efficiency (fully utilization) Fairness (resource sharing) Distributedness (no central knowledge for scalability)

thomascruz
Download Presentation

TCP/Vegas

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. TCP/Vegas Yang Richard Yang 10/1/2001

  2. Review AIMD Congestion Control

  3. What Are the Desired Properties of a Congestion Avoidance Scheme? • Efficiency (fully utilization) • Fairness (resource sharing) • Distributedness (no central knowledge for scalability) • Convergence (fast convergence after disturbance, low oscillation) • responsiveness (speed to new state, lower or higher) • smoothness

  4. Linear control • The simplest control strategy

  5. overload x0 x0 convergence: efficiency convergence: efficiency x0 x0 convergence: fairness convergence

  6. underload x0 x0 convergence: efficiency convergence: efficiency x0 x0 convergence: fairness convergence

  7. AIMD State Space x2 Fairness Line x0 Efficiency Line x1

  8. Other Special Cases

  9. TCP/Vegas

  10. Algorithms • Retransmit • first DUPACK & Timeout (finer grained timeout) • multiple losses: detect early, and no multiple reductions • Congestion avoidance • Slow start • increase every other RTT (congestion detection in the middle)

  11. Issues: Rerouting • If the propagation delay increases due to a new route, the flow should increase its window size to fill the pipe; however, the sender can interpret the increase as a signal of congestion and decrease its window size, which is undesired

  12. Issues: Persistent Congestion • Remember the objective of Vegas is to keep w = x d + , where x is rate, d is delay, and  . Then if a new flow arrives and many flows are already sharing the link, the flow will observe a large d. This large d is interpreted as propagation delay instead of queuing delay. Later, when some flows leave, the new flow will interpret the reduction of congestion as not enough window size and increase window size to maintain a persistent congestion

  13. Issues: Fairness • TCP/Vegas grabs less bandwidth than TCP/Reno. The best way to think the problem is to think about a FIFO queue. The larger share of the queue occupied by a flow, the more bandwidth achieved by a flow. SO how to delay TCP/Vegas in a TCP/Reno world? • TCP/Vegas is largely insensitive to different round-trip time, and therefore has a different fairness semantics

More Related