1. Katz, Stoica F04 EECS 122: Introduction to Computer Networks TCP Variations Computer Science Division Department of Electrical Engineering and Computer Sciences University of California, Berkeley Berkeley, CA 94720-1776

2. Katz, Stoica F04 2 Today’s Lecture: 11 Network (IP) Application Transport Link Physical 2 7, 8, 9 10, 11 17, 18, 19 14, 15, 16 21, 22, 23 25 6

3. Katz, Stoica F04 3 Outline  TCP congestion control -Quick Review -TCP flavors -Equation-based congestion control -Impact of losses -Cheating  Router-based support -RED -ECN -Fair Queueing -XCP

4. Katz, Stoica F04 4 Quick Review  Slow-Start: cwnd++ upon every new ACK  Congestion avoidance: AIMD if cwnd > ssthresh -ACK: cwnd = cwnd + 1/cwnd -Drop: ssthresh =cwnd/2 and cwnd=1  Fast Recovery: -duplicate ACKS: cwnd=cwnd/2 -Timeout: cwnd=1

5. Katz, Stoica F04 5 TCP Flavors  TCP-Tahoe -cwnd =1 whenever drop is detected  TCP-Reno -cwnd =1 on timeout -cwnd = cwnd/2 on dupack  TCP-newReno -TCP-Reno + improved fast recovery  TCP-Vegas, TCP-SACK

6. Katz, Stoica F04 6 TCP Vegas  Improved timeout mechanism  Decrease cwnd only for losses sent at current rate -avoids reducing rate twice  Congestion avoidance phase: -compare Actual rate (A) to Expected rate (E) -if E-A > , decrease cwnd linearly -if E-A < , increase cwnd linearly -rate measurements ~ delay measurements -see textbook for details!

7. Katz, Stoica F04 7 TCP-SACK  SACK = Selective Acknowledgements  ACK packets identify exactly which packets have arrived  Makes recovery from multiple losses much easier

8. Katz, Stoica F04 8 Standards?  How can all these algorithms coexist?  Don’t we need a single, uniform standard?  What happens if I’m using Reno and you are using Tahoe, and we try to communicate?

9. Katz, Stoica F04 9 Equation-Based CC  Simple scenario -assume a drop every k’th RTT (for some large k) -w, w+1, w+2,...w+k-1 DROP (w+k-1)/2, (w+k-1)/2+1,...  Observations: -In steady state: w= (w+k-1)/2 so w=k-1 -Average window: 1.5(k-1) -Total packets between drops: 1.5k(k-1) -Drop probability: p = 1/[1.5k(k-1)]  Throughput: T ~ (1/RTT)*sqrt(3/2p)

10. Katz, Stoica F04 10 Equation-Based CC  Idea: -Forget complicated increase/decrease algorithms -Use this equation T(p) directly!  Approach: -measure drop rate (don’t need ACKs for this) -send drop rate p to source -source sends at rate T(p)  Good for streaming audio/video that can’t tolerate the high variability of TCP’s sending rate

11. Katz, Stoica F04 11 Question!  Why use the TCP equation?  Why not use any equation for T(p)?

12. Katz, Stoica F04 12 Cheating  Three main ways to cheat: -increasing cwnd faster than 1 per RTT -using large initial cwnd -Opening many connections

13. Katz, Stoica F04 13 Increasing cwnd Faster AB x DE y Limit rates: x = 2y C x y x increases by 2 per RTT y increases by 1 per RTT

14. Katz, Stoica F04 14 Increasing cwnd Faster AB x DE y

15. Katz, Stoica F04 15 Larger Initial cwnd AB x DE y x starts SS with cwnd = 4 y starts SS with cwnd = 1

16. Katz, Stoica F04 16 Open Many Connections AB x DE y Assume A starts 10 connections to B D starts 1 connection to E Each connection gets about the same throughput Then A gets 10 times more throughput than D

17. Katz, Stoica F04 17 Cheating and Game Theory AB x DE y 22, 2210, 35 35, 1015, 15 (x, y) A Increases by 1 Increases by 5 D  Increases by 1 Increases by 5 Individual incentives: cheating pays Social incentives: better off without cheating Classic PD: resolution depends on accountability Too aggressive  Losses  Throughput falls

18. Katz, Stoica F04 18 Lossy Links  TCP assumes that all losses are due to congestion  What happens when the link is lossy?  Recall that Tput ~ 1/sqrt(p) where p is loss prob.  This applies even for non-congestion losses

19. Katz, Stoica F04 19 Example p = 0 p = 1% p = 10%

20. Katz, Stoica F04 What can routers do to help?

21. Katz, Stoica F04 21 Paradox  Routers are in middle of action  But traditional routers are very passive in terms of congestion control -FIFO -Drop-tail

22. Katz, Stoica F04 22 FIFO: First-In First-Out  Maintain a queue to store all packets  Send packet at the head of the queue Queued packets Arriving packet Next to transmit

23. Katz, Stoica F04 23 Tail-drop Buffer Management  Drop packets only when buffer is full  Drop arriving packet Arriving packet Next to transmit Drop

24. Katz, Stoica F04 24 Ways Routers Can Help  Packet scheduling: non-FIFO scheduling  Packet dropping: -not drop-tail -not only when buffer is full  Congestion signaling

25. Katz, Stoica F04 25 Question!  Why not use infinite buffers? -no packet drops!

26. Katz, Stoica F04 26 The Buffer Size Quandary  Small buffers: -often drop packets due to bursts -but have small delays  Large buffers: -reduce number of packet drops (due to bursts) -but increase delays  Can we have the best of both worlds?

27. Katz, Stoica F04 27 Random Early Detection (RED)  Basic premise: -router should signal congestion when the queue first starts building up (by dropping a packet) -but router should give flows time to reduce their sending rates before dropping more packets  Therefore, packet drops should be: -early: don’t wait for queue to overflow -random: don’t drop all packets in burst, but space drops out

28. Katz, Stoica F04 28 RED  FIFO scheduling  Buffer management: -Probabilistically discard packets -Probability is computed as a function of average queue length (why average?) Discard Probability Average Queue Length 0 1 min_thmax_th queue_len

29. Katz, Stoica F04 29 RED (cont’d)  min_th – minimum threshold  max_th – maximum threshold  avg_len – average queue length -avg_len = (1-w)*avg_len + w*sample_len Discard Probability Average Queue Length 0 1 min_thmax_th queue_len

30. Katz, Stoica F04 30 RED (cont’d)  If (avg_len < min_th)  enqueue packet  If (avg_len > max_th)  drop packet  If (avg_len >= min_th and avg_len < max_th)  enqueue packet with probability P Discard Probability (P) Average Queue Length 0 1 min_thmax_th queue_len

31. Katz, Stoica F04 31 RED (cont’d)  P = max_P*(avg_len – min_th)/(max_th – min_th)  Improvements to spread the drops (see textbook) Discard Probability Average Queue Length 0 1 min_thmax_th queue_len avg_len P max_P

32. Katz, Stoica F04 32 Average vs Instantaneous Queue

33. Katz, Stoica F04 33 RED Advantages  High network utilization with low delays  Average queue length small, but capable of absorbing large bursts  Many refinements to basic algorithm make it more adaptive (requires less tuning)

34. Katz, Stoica F04 34 Explicit Congestion Notification  Rather than drop packets to signal congestion, router can send an explicit signal  Explicit congestion notification (ECN): -instead of optionally dropping packet, router sets a bit in the packet header -If data packet has bit set, then ACK has ECN bit set  Backward compatibility: -bit in header indicates if host implements ECN -note that not all routers need to implement ECN

35. Katz, Stoica F04 35 Picture W W/2 AB

36. Katz, Stoica F04 36 ECN Advantages  No need for retransmitting optionally dropped packets  No confusion between congestion losses and corruption losses

37. Katz, Stoica F04 37 Remaining Problem  Internet vulnerable to CC cheaters!  Single CC standard can’t satisfy all applications -EBCC might answer this point  Goal: -make Internet invulnerable to cheaters -allow end users to use whatever congestion control they want  How?

38. Katz, Stoica F04 38 One Approach: Nagle (1987)  Round-robin among different flows -one queue per flow

39. Katz, Stoica F04 39 Round-Robin Discussion  Advantages: protection among flows -Misbehaving flows will not affect the performance of well- behaving flows Misbehaving flow – a flow that does not implement any congestion control -FIFO does not have such a property  Disadvantages: -More complex than FIFO: per flow queue/state -Biased toward large packets – a flow receives service proportional to the number of packets

40. Katz, Stoica F04 40 Solution?  Bit-by-bit round robin  Can you do this in practice?  No, packets cannot be preempted (why?)  …we can only approximate it

41. Katz, Stoica F04 41 Fair Queueing (FQ)  Define a fluid flow system: a system in which flows are served bit-by-bit  Then serve packets in the increasing order of their deadlines  Advantages -Each flow will receive exactly its fair rate  Note: -FQ achieves max-min fairness

42. Katz, Stoica F04 42 Max-Min Fairness  Denote -C – link capacity -N – number of flows -r i – arrival rate  Max-min fair rate computation: 1.compute C/N 2.if there are flows i such that r i <= C/N, update C and N 3.if no, f = C/N; terminate 4.go to 1  A flow can receive at most the fair rate, i.e., min(f, r i )

43. Katz, Stoica F04 43 Example  C = 10; r 1 = 8, r 2 = 6, r 3 = 2; N = 3  C/3 = 3.33  C = C – r3 = 8; N = 2  C/2 = 4; f = 4 8 6 2 4 4 2 f = 4 : min(8, 4) = 4 min(6, 4) = 4 min(2, 4) = 2 10

44. Katz, Stoica F04 44 Implementing Fair Queueing  Idea: serve packets in the order in which they would have finished transmission in the fluid flow system

45. Katz, Stoica F04 45 Example 12345 1234 12 3 12 4 34 5 56 12132344 56 556 Flow 1 (arrival traffic) Flow 2 (arrival traffic) Service in fluid flow system Packet system time

46. Katz, Stoica F04 46 System Virtual Time: V(t)  Measure service, instead of time  V(t) slope – rate at which every active flow receives service -C – link capacity -N(t) – number of active flows in fluid flow system at time t 12 3 12 4 34 5 56 Service in fluid flow system time V(t)

47. Katz, Stoica F04 47 Fair Queueing Implementation  Define - - finishing time of packet k of flow i (in system virtual time reference system) - - arrival time of packet k of flow i - - length of packet k of flow i  The finishing time of packet k+1 of flow i is

48. Katz, Stoica F04 48 FQ Advantages  FQ protect well-behaved flows from ill-behaved flows  Example: 1 UDP (10 Mbps) and 31 TCP’s sharing a 10 Mbps link

49. Katz, Stoica F04 49 Alternative Implementations of Max-Min  Deficit round-robin  Core-stateless fair queueing -label packets with rate -drop according to rates -check at ingress to make sure rates are truthful  Approximate fair dropping -keep small sample of previous packets -estimate rates based on these -apply dropping as above -wins because few large flows per-elephant state, not per-mouse state  RED-PD: not max-min, but punishes big cheaters

50. Katz, Stoica F04 50 Big Picture  FQ does not eliminate congestion  it just manages the congestion  You need both end-host congestion control and router support for congestion control -end-host congestion control to adapt -router congestion control to protect/isolate

51. Katz, Stoica F04 51 Explicit Rate Signaling (XCP)  Each packet contains: cwnd, RTT, feedback field  Routers indicate to flows whether to increase or decrease: -give explicit rates for increase/decrease amounts -feedback is carried back to source in ACK  Separation of concerns: -aggregate load -allocation among flows

52. Katz, Stoica F04 52 XCP (continued)  Aggregate: -measures spare capacity and avg queue size -computes desired aggregate change: D=aRS-bQ  Allocation: -uses AIMD -positive feedback is same for all flows -negative feedback is proportional to current rate -when D=0, reshuffle bandwidth -all changes normalized by RTT want equal rates, not equal windows

53. Katz, Stoica F04 53 XCP (continued)  Challenge: -how to give per-flow feedback without per-flow state? -do you keep track of which flows you’ve signaled and which you haven’t?  Solution: -figure out desired change -divide from expected number of packets from flow in time interval -give each packet share of rate adjustment -flow totals up all rate adjustment