1 / 22

Medium Start in TCP-Friendly Rate Control Protocol

This article provides an overview of TFRC, a congestion control scheme for transport layer protocols. It discusses the benefits and drawbacks of TFRC and introduces the Medium Start algorithm, which aims to reach the fair sending rate faster while still performing conservative probing. The article also examines the implementation and evaluation of Medium Start in ns-2 simulation.

nchappelle
Download Presentation

Medium Start in TCP-Friendly Rate Control Protocol

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. Medium Start in TCP-Friendly Rate Control Protocol CS 217 Class Project Spring 04 Peter Leong & Michael Welch

  2. TFRC Overview • Congestion control scheme (CCID3) for transport layer protocols • Mostly used for unreliable data flows • Competes fairly with TCP for bandwidth • Lower throughput variation over time than TCP • Streaming multimedia • Telephony data • RFC 3448

  3. TFRC Overview (cont) • Streaming data are sent using UDP without congestion control • Results an unfair sharing of the network bandwidth • Transports without congestion control mechanisms run a higher risk of being selectively dropped by a router • Provides feedback to sender about network condition

  4. TFRC Algorithm • The receiver measures the loss event rate and feeds this information back to the sender. • The sender also uses these feedback messages to measure the round-trip time (RTT). • The loss event rate and RTT are then fed into TFRC's throughput equation, giving the acceptable transmit rate. • The sender then adjusts its transmit rate to match the calculated rate.

  5. TFRC Algorithm -- Handling of idle periods • Sender has a nofeedback timer • Receiving a feedback packet resets the nofeedback timer to expire after 2*RTT • Nofeedback timer expiration triggers halving of the send rate and timer reset • e.g. An idle period of 10*RTT = halving send rate 5 times • Exponential decrease of window size

  6. TFRC Algorithm -- Handling of idle periods (cont) • If the sender subsequently starts to send again, a normal slowstart phase will occur until the transmit rate reaches calculated fair rate. • An idle period of several RTTs is not uncommon for voice-over-IP data transmission • This means we often find ourselves sending data in slowstart (below our last measured fair rate) • In general, we slowstart because we do not know current network conditions

  7. Medium Start • Assuming that fair rate is a good indicator of the actual available bandwidth within a reasonable time period, can we do better? • With TFRC, we have a recent sample of the network through our fair rate calculation • We can use this information to slowstart faster  Medium Start

  8. Medium Start Algorithm • Set T_idle := now – T_last_send • F := (30 min - min(max(T_idle, 10 min), 30 min)) / 20 min • Set ssthresh := F*rcvrate • Set rate := max(4*size/RTT, rate) • Set ssmult = 4 • Perform traditional slowstart algorithm, including backoff mechanisms

  9. Medium Start Evaluation • We implemented Medium Start in ns-2 • Compared achieved sending rate of TFRC with Medium Start vs. Stock TFRC • Varied the loss rate, competing traffic type and density, and the source sending model. • Results:

  10. Evaluation Topology • Node 0 : TFRC Sender • Node 1 : Competing traffic source • Node 2 : Router • Node 3 : TFRC Receiver • Links are 10Mb • Losses occur on the 2-3 link only • Router is DropTail • Packetsize = 500 bytes

  11. Evaluation 1 • CBR sender • No competing traffic • 0.1 Mbps sending rate • Link loss 15% • X axis is time • Y axis is sending rate in bytes

  12. Evaluation 1 - Stock TFRC

  13. Evaluation 1 – Medium Start

  14. Evaluation 2 • Exponentially distributed sender • Mean burst time = 10s • Mean idle time = 10s • No competing traffic • 0.1 Mbps burst sending rate • Link loss 15% • X axis is time • Y axis is sending rate in bytes

  15. Evaluation 2 – Stock TFRC

  16. Evaluation 2 – Medium Start

  17. Evaluation 3 • Exponentially distributed sender • Mean burst time = 10s • Mean idle time = 10s • 1 Mbps burst sending rate • Exponentially distributed competing traffic • Mean burst time = 5s • Mean idle time = 5s • 5 Mbps burst sending rate • Link loss 10% • X axis is time • Y axis is sending rate in bytes

  18. Evaluation 3 – Stock TFRC

  19. Evaluation 3 – Medium Start

  20. Benefits of Medium Start • Utilizes recent measurements to reach our fair sending rate faster • While still performing some conservative probing • Allows higher fidelity for streaming media applications • Allows higher utilization of the channel

  21. Drawbacks of Medium Start • Most test results show an increase in loss rate compared to Stock TFRC • Relies on stable network dynamics • May cause competing flows to experience additional loss  less TCP Friendly?

  22. Problems and Open Issues • ns-2 implementation of TFRC seems to deviate from the RFC • Simulation does not allow us to perceive the improvement Medium Start may have on streaming multimedia and internet telephony • An actual implementation would allow us to tweak the algorithm’s parameters through experimentation

More Related