1 / 21

TCP Performance Over Wireless Links

TCP Performance Over Wireless Links. Read: - Hari Balakrishnan et. al , “A Comparison of Mechanisms for Improving TCP Performance over Wireless Links …”, SIGCOMM 1996. Ibrahim Matta et. al, “Open Issues on TCP for Mobile Computing”, WCMB 2002.

bishop
Download Presentation

TCP Performance Over Wireless Links

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 Performance Over Wireless Links Read: - Hari Balakrishnan et. al , “A Comparison of Mechanisms for Improving TCP Performance over Wireless Links …”, SIGCOMM 1996. Ibrahim Matta et. al, “Open Issues on TCP for Mobile Computing”, WCMB 2002. - S. Biaz et al. “De-Randomizing” Congestion Losses to Improve TCP Performance over Wired-Wireless Networks”, to appear in June 2005 IEEE/ACM Transactions on Networking

  2. Impact of the Packet Error Rate on TCP • FromSally Floyd, HighSpeed TCP for Large Congestion Windows. RFC 3649 Packet Drop Rate P Congestion Window W RTTs Between Losses ------------------ ------------------- ------------------- 10^-2 12 8 10^-3 38 25 10^-4 120 80 10^-5 379 252 10^-6 1200 800 10^-7 3795 2530 10^-8 12000 8000 10^-9 37948 25298 10^-10 120000 80000 Table 2: TCP Response Function for Standard TCP. The average congestion window W in MSS-sized segments is given as a function of the packet drop rate P. TCP Wireless

  3. Impact of the Bit Error Rate on TCP • AfterSally Floyd, HighSpeed TCP for Large Congestion Windows. RFC 3649 Bit Error Rate BER Congestion Window W RTTs Between Losses ------------------ ------------------- ------------------- 10^-6 12 8 10^-7 38 25 10^-8 120 80 10^-9 379 252 10^-10 1200 800 10^-11 3795 2530 10^-12 12000 8000 10^-13 37948 25298 10^-14 120000 80000 Table 2: TCP Response Function for Standard TCP. The average congestion window W in MSS-sized segments is given as a function of the packet drop rate P. TCP Wireless

  4. Impact of the Bit Error Rate on TCP • Consider an MSS of 1 KB and an RTT of 100ms, what should be the BER to reach a throughput of 54 Mbps? TCP Wireless

  5. Impact of Wireless Losses on TCP • The congestion window is unnecessarily decreased • Retransmission timer is biased towards less responsiveness (TCP sender wrongly “senses” congestion.). TCP Wireless

  6. Strategies to Improve TCP Wireless • Do not use losses as congestion indication (delay/throughput based congestion control such as TCP Westwood) • Differentiate congestion losses from random wireless losses • End to end techniques • Network feedback TCP Wireless

  7. Metrics of Accuracy • Accuracy of diagnosing correctly: • Congestion losses (Ac) • Random wireless losses (Aw) • Ngc: Number of congestion losses correctly diagnosed as such • Nc: Total number of congestion losses TCP Wireless

  8. Metrics of Accuracy (2) • Example of a TCP connection with: • 60 congestion drops of which 40 were diagnosed correctly  Ac = 40/60 = 0.67 • 20 random wireless losses of which 15 were diagnosed correctly  Aw = 15/20 = 0.75 • Normal TCP • Ac = 1 • Aw = 0 TCP Wireless

  9. Expected Improvement TCP Wireless

  10. Expected Improvement (RTT) TCP Wireless

  11. “Middle man” Techniques

  12. Indirect TCP (I-TCP) (Bakre et. al) • TCP connection broken into two TCP connection • Each TCP connection adapts to specific environment: mss, congestion avoidance, timer.. Mobile Wired Network Sender Basestation TCP Wired TCP Wireless TCP Wireless

  13. Snoop (Balakrishnan et. al) • One TCP connection • A Snoop agent on basestation: • “Shields” sender from wireless losses (in fact, hides losses) • Retransmit on wireless link in case of random wireless drops. Mobile Wired Network Sender Basestation Snoop TCP Wireless

  14. Snoop (Balakrishnan et. al) (2) • Snoop buffers at basestation packets until acked by mobile. Snoop may retransmit packets. • Snoop perfectly diagnoses congestion losses and wireless losses. • When out-of-order packet reach Snoop, Snoop generates duplicate acks (congestion losses). • When dupacks reach Snoop (wireless losses), Snoop filters these dupacks and retransmits packets. Mobile Wired Network Sender Basestation Snoop TCP Wireless

  15. Explicit Bad State Notification (EBSN) (Bakshi et. al) • Basestation keeps track of channel state and feeds it back to sender • Sender diagnoses losses using ESBN and reacts appropriately Mobile Wired Network Sender Basestation EBSN TCP Wireless

  16. Explicit Transport Error Notification (ETEN) (Krishnan et. al) • Link layer (at any hop) would notify endpoint of losses other than congestion • Endpoint takes appropriate action Mobile Wired Network Sender Basestation TCP Wireless

  17. Explicit Congestion Notification (ECN) (Floyd and Jacobson) • Can ECN be used to infer/diagnose the cause of a loss? Random or congestion • Bad especially when TCP sender is ECN responsive. TCP Wireless

  18. “De-randomize” Congestion Losses (Biaz and Vaidya):Biased Queue Management • Packets of the same flow are marked with different discard priorities: 1 every k packets is marked “out”, all others marked “in” • When congestion occurs: drop FIRST packet marked “out”. • Good results: Ac close to 1 and Aw close to 0.75 TCP Wireless

  19. Congested router Biased Queue Management TCP Wireless

  20. At the Receiver:How to Interpret Losses? • Consider there were S packets in flight and r packets are lost. • Let X be a random variable = number of packets marked “out” within the r lost packets • What is the probability P(X=x) that x packet among the r packets are marked “out”? TCP Wireless

  21. At the Receiver:How to Interpret Losses? (2) • P(X=x) has a hypergeometric distribution • Examples: • Case 1: S = 24, k = 8, P (X = 3) = 1/2024 (Random ?) • Case 2: S = 24, k = 8, P (X = 0) = 0.65 (Random?) TCP Wireless

More Related