1. Lecture 8 Congestion Control EECS 122 University of California Berkeley

2. EECS 122Walrand2 TOCTOC: Congestion Control The Problem Questions Approaches TCP: Algorithm TCP Refinements Summary

3. EECS 122Walrand3 Congestion ControlCongestion Control: The Problem Flows share links: How to share the links bandwidth?

4. EECS 122Walrand4 Congestion ControlCongestion Control: Questions What should be the ideal sharing? Does it matter? Does it matter Discovering available bandwidth What is fair? What is fair

5. EECS 122Walrand5 QuestionsQuestions: Does it matter? Congestion occurs Access link  Slow link (56k, DSL, T1, wireless, …) Access network  E.g., behind the DSLAM Can improve treatment of flows E.g., one flow should not get a much smaller fraction of bandwidth Some flows might need some guaranteed bandwidth

6. EECS 122Walrand6 QuestionsQuestions: Available bandwidth? Example: = router = host 3 = link with bandwidth of 3Mbps (same for 6 and 10) x, y, z = throughput of flows

7. EECS 122Walrand7 QuestionsQuestions: Available bandwidth? Example: Assume C  D with rate y and E  F with rate z How does A “discover” the available bandwidth to B? Some approaches: 1.ReservationReservation 2.Adapt to congestionAdapt 3.Test for sufficient bandwidthTest 4.Pricing congestionPricing

8. EECS 122Walrand8 QuestionsQuestions: Available bandwidth? Example: Assume C  D with rate y and E  F with rate z How does A “discover” the available bandwidth to B? Some approaches: 1.ReservationReservation 2.Adapt to congestionAdapt 3.Test for sufficient bandwidthTest 4.Pricing congestionPricing

9. EECS 122Walrand9 Available bandwidthAvailable bandwidth: Reservation 1.Routers (or manager) keep track of reserved rates 2.A requests a rate R to B from the network 3.The network figures out if R is available 4.If R is available, routers (or manager) update reservations and confirm to A 5.Note: Complex, Slow, Requires enforcement, Renegotiations, Pricing

10. EECS 122Walrand10 Available bandwidthAvailable bandwidth: Adapt 1.Transmit and slow down if congestion occur 2.Example: Initially: x= 0, y = 3, z = 3 Then A increases its rate; C and E notice congestion and slow down Later, C stops: A and E increase rates 3.Notes: No guarantees: throughput may drop Key question: how to adapt rates

11. EECS 122Walrand11 Available bandwidthAvailable bandwidth: Test 1.Assume flows require at most 1Mbps (e.g., video) 2.Routers monitor their rates to see if they have at least 1 Mbps of available bandwidth; they mark packets otherwise 3.If A wants a new flow to B, it sends test packets to B 4.If routers do not mark test packets, then A can start its new flow; otherwise, A does not start it 5.Advantages: 1.relatively simple 2.guarantee

12. EECS 122Walrand12 Available bandwidthAvailable bandwidth: Pricing Example: When they get saturated, routers mark packets If a flow with rate R uses saturated links, it gets marks with rate R Each mark costs one unit Source slows down if price becomes excessive x= 1+, y = 2+, z = 2+  pA = 1 + 1; pC = pE = 2 x = 2+, y = 1+, z = 1+  pA = 2 + 2; pC = pE = 1

13. EECS 122Walrand13 QuestionsQuestions: What is Fair? Example: x = y = z = 1.5: fair in max-min sense x = 0, y = z = 3: maximizes x + y + z 5x = 4y = 4z: equalizes resources flows use with x = 1.33, y = z = 1.67 What if A  B needs 2Mbps? (and is willing to pay for it)

14. EECS 122Walrand14 Congestion ControlCongestion Control: Approaches Telephone Network: Reservation Transmission Control Protocol (TCP) Adapt rate to congestion Algorithm for adaptation attempts to be fair … User Datagram Protocol (UDP) Transmit and hope for the best Various proposals for Internet: Reservation Pricing Test Note: Either by hosts or between domains

15. EECS 122Walrand15 Congestion ControlCongestion Control: TCP Algorithm Principles Example Multiple Sources A Bad AlgorithmA Bad Algorithm: AIAD AIMD: Additive Increase – Multiplicative Decrease Why AIAD Fails

16. EECS 122Walrand16 TCP AlgorithmTCP Algorithm: Principles We focus on the “standard” TCP (reno) Idea: Not congested => increase rate Congested => slow down Questions: How to detect congestion?  Missing ACKs How to increase/slow down?  AIMD

17. EECS 122Walrand17 TCP AlgorithmTCP Algorithm: Example No congestion  x increases by one packet/RTT every RTT Congestion  decrease x by factor 2 AB x C = 50 pkts/RTT

18. EECS 122Walrand18 TCP AlgorithmTCP Algorithm: Multiple Sources AB x C = 50 pkts/RTT DE No congestion  rate increases by one packet/RTT every RTT Congestion  decrease rate by factor 2 Rates equalize  fair share y

19. EECS 122Walrand19 TCP AlgorithmTCP Algorithm: Bad Algorithm AB x C = 50 pkts/RTT DE No congestion  x increases by one packet/RTT every RTT Congestion  decrease x by 1 y

20. EECS 122Walrand20 TCP AlgorithmTCP Algorithm: AIMD AB x C DE y Limit rates: x = y

21. EECS 122Walrand21 TCP AlgorithmTCP Algorithm: Why AIAD Fails AB x C DE y Limit rates: x and y depend on initial values

22. EECS 122Walrand22 Congestion ControlCongestion Control: TCP Refinements Fast Retransmit Fast RecoveryFast Recovery: 1 st Look Fast RecoveryFast Recovery: 2 nd Look Slow Start Window Updates Flow Control Summary

23. EECS 122Walrand23 RefinementsRefinements: Fast Retransmit n n+1 n+2 Cumulative ACKs: ACK # = next expected # n+1 n+3 n+1 3 rd duplicated ACK:  likely packet loss  retransmit n+1 timeout

24. EECS 122Walrand24 Refinements Refinements : Fast Recovery (1) Timeout  Reset Window = 1 unit (MSS) 3 rd Dup ACK  Window/2 Window Slope = 1 MSS/RTT 3 rd Dup ACK n n/2 Timeout 1 Moderate congestion (subsequent pkts arrived) Severe congestion

25. EECS 122Walrand25 Refinements Refinements : Fast Recovery (2) Window adjustment is tricky: Want W  W/2 1 n ssthresh = W/2 W = ssthresh + 3 W = W + 1 at each Dup Ack n - 3 W = n + W/2 W/2 outstanding packets: n+1, …, n+W/2 W/2 outstanding packets: n+1, …, n+W/2 W W = ssthresh

26. EECS 122Walrand26 Refinements Refinements : Slow Start Objective: Discover available bdw fast Solution: Exponential increase of window W Timeout 1 n n/2 Threshold 65KB exp Additive Slope = 1/RTT

27. EECS 122Walrand27 Refinements Refinements : Window Updates Exponential: W = W + 1 at each ACK: W = 1W = 2W = 4W = 8 Additive: W = W + 1/W at each ACK: W = 8 + 1/8 W = 8.125 + 1/8.125  8 + 2/8 W  8 + 8/8 = 9 W  9 + 9/9 = 10

28. EECS 122Walrand28 Refinements Refinements : Flow Control Objective: Avoid saturating destination Algorithm: Receiver avertizes window RAW RAW window = min{RAW – OUT, W} where OUT = Oustanding = Last sent – last ACKed W = Cong. Window from AIMD + refinements [ACK | RAW | …]

29. EECS 122Walrand29 Refinements Refinements : Summary W 1 65KB Actual window = min{RAW - OUT, W} X 0.5 TO 3DA X 0.5 3DA TO X 0.5 SSCA SS CA 3 3

30. EECS 122Walrand30 Congestion ControlCongestion Control: Summary Slow Start: Discover available bandwidth Congestion Avoidance: AIMD  Tries to be fair Refinements: Fast Retransmit: 3DA Fast Recovery: Reset W to W/2 (instead of W = 1) [More precisely: ssthresh = W/2, W = W + 1 per DA, W = ssthresh when get new ACK.] TO: set ssthresh = W/2, W = 1, SS until W = ssthresh, then CA Timers: Timeout = Average + 4 Deviations If time out  Timeout x 2 Reset after new packet or new ACK Flow Control: Window = min{RAW – OUT, W}