1. Multimedia & Comm. Lab Rate control for ABR service in ATM Networks 98/9/30 Multimedia & Comm. Lab 정승훈

2. Multimedia & Comm. Lab 2 Contents  Introduction  ABR Service  Congestion Control Mechanisms  Source-level Rate Adaptation  Examples  Open Issues

3. Multimedia & Comm. Lab 3 Introduction  Classes of Service  Why need Congestion Control  Traffic Management Functions  What is Expected from Congestion Control

4. Multimedia & Comm. Lab 4 Classes of Service  ABR (Available bit rate) Follows feedback instructions. Network gives max throughput with minimum loss.  UBR (Unspecified bit rate) User sends whenever it wants. No feedback mechanism, No guarantee. Cells may be dropped during congestion.  CBR (Constant bit rate) Throughput, delay, and Jitter guaranteed.  VBR (Variable bit rate)

5. Multimedia & Comm. Lab 5 Why need Congestion Control  Will the congestion problem be solved when: Memory becomes cheap (infinite memory)? Links become cheap (very high speed links)? Processors become cheap?  Congestion is a dynamic problem Static solutions are not sufficient  Bandwidth explosion More unbalanced networks.  Buffer shortage is a symptom, not the cause.

6. Multimedia & Comm. Lab 6 Traffic Management Functions  Connection Admission Control (CAC) Verify that the requested bandwidth and QoS can be supported  Traffic Shaping Limit burst length, Space-out cells  Usage Parameter Control (UPC) Monitor and control traffic at the network entrance  Network Resource Management Scheduling, Queueing, Virtual path resource reservation

7. Multimedia & Comm. Lab 7 Traffic Management Functions  Priority Control Cell Loss Priority (CLP) = 1 cells may be dropped  Selective Cell Discarding Frame discard  Feedback Controls Network tell the source to increase or decrease its load Explicit forward congestion indication (EFCI) Explicit rate (ER) Backward explicit congestion notification (BECN)

8. Multimedia & Comm. Lab 8 What is expected ?  Objectives Support a set of QoS parameters and classes for all ATM services Minimize network and end-system complexity while maximizing network utilization  Selection Criteria Scalability Fairness Robustness Implementability

9. Multimedia & Comm. Lab 9 ABR Service  The Nature of the ABR Service  Some Early Debates  The Role of the Network  The Role of the End Systems

10. Multimedia & Comm. Lab 10 The Nature of the ABR Service  ABR Service ABR connections will share the available bandwidth The share of available bandwidth for each ABR connection is dynamic and may diminish down to a specified minimum cell rate (MCR) The dynamic nature of the ABR service can be seen from the feedback model The ABR service is appropriate only for applications which can adapt their rates to the time-varying available bandwidth and tolerate unpredictable cell delays a low or zero cell loss rate is guaranteed to users who adapts their rates properly

11. Multimedia & Comm. Lab 11 Some Early Debates  Open-Loop vs. Close-Loop  Credit-based vs. Rate-based  Binary Feedback vs. Explicit Feedback

12. Multimedia & Comm. Lab 12 Open-Loop vs. Close-Loop  Open-Loop not need end-to-end feedback prior-reservation and hop-to-hop flow control  Close-Loop the source adjust its cell rate in responding to the feedback information from the network. Too slow in high-speed networks But, ABR service is designed to use any bandwidth ATM Forum specified that feedback is necessary for ABR flow control

13. Multimedia & Comm. Lab 13 Credit-based vs. Rate-based  Credit-based hop-by-hop per-VC window Static : Full round-trip worth of credit per VC Adaptive : Credits depend upon activity  Rate-based End-to-end rate control Binary : Feedback via congestion bit in cells Explicit : Feedback via resource management (RM) cells

14. Multimedia & Comm. Lab 14 Credit vs. Rate Debate : Issues  Per-VC queueing Switch complexity, Nonscalable  Switch vs. end-system complexity  Zero cell loss  Isolation and misbehaving users  Buffer requirements Full round-trip per VC

15. Multimedia & Comm. Lab 15 Binary vs. Explicit Rate  Binary Feedback One-bit Feedback Explicit forward congestion indicator (EFCI) set to 0 at source Congested switch set EFCI to 1 Every nth cell, destination sends a RM cell to the source indicating increase amount or decrease factor

16. Multimedia & Comm. Lab 16 Binary vs. Explicit Rate  Explicit Rate Feedback Every N rm cells, the sources send a control cell The switches measure load over a period The destination returns the cell to the source The switches specify explicit rate in cell The source adjusts the transmission rate

17. Multimedia & Comm. Lab 17 Binary vs. Explicit Rate  ER feedback schemes have several advantages The switches know more information along the flow path Faster to get the source to the optimal operating point Policing is straight forward  Two ways for ER feedback Forward feedback Backward feedback

18. Multimedia & Comm. Lab 18 The Role of The Network  The network might provide information directly to the users No information A binary congestion indication  EFCI Detailed congestion indication  RM cells with queue levels and severity level Explicit bandwidth (or rate) information  RM cell with the current available bandwidth that can be adjusted by nodes along the connection in the forward direction  The destination returns the RM cell to the source with either and absolute rate or a relative rate adjustment

19. Multimedia & Comm. Lab 19 The role of the End systems  How the source and destination end systems work with Feedback information to adapt the source rate Negative Feedback  source increments its rate by default Positive Feedback  source decrements its rate by default Explicit Feedback  source maintains its rate by default

20. Multimedia & Comm. Lab 20 Congestion Control Mechanisms  Fairness  Binary Feedback EFCI PRCA  Explicit Rate Feedback EPRCA ERICA

21. Multimedia & Comm. Lab 21 Fairness  Max-Min available bandwidth = C / N  MCR plus equal share available bandwidth = MCR + (C - MCR) / N  Maximum of MCR or Max-Min share available bandwidth = max{MCR, Max-Min share}  Allocation proportional to MCR The bandwidth allocation for a connection is weighted proportional to its MCR  Weighted allocation

22. Multimedia & Comm. Lab 22 EFCI  Mechanism The network uses EFCI to convey congestion information (in the forward direction) Feedback is returned via RM cells from the destination end system to source. The sources adjust their rates by additive increase and multiplicative decrease (at periodic update intervals). The feedback is negative, the source increase their rates by default and decrease only if an RM cell is received

23. Multimedia & Comm. Lab 23 PRCA  Proportional rate control algorithm positive feedback RM cells are generated at a rate proportional to the source rate End system requires a means to discover when to generate an RM cell.  Every Nrm cells, only one cell with EFCI=0 Source Dest ATM node RM cell If EFCI=0 cell received EFCI=1EFCI=0 N rm

24. Multimedia & Comm. Lab 24 EPRCA  Enhanced proportional rate control algorithm Source behavior  The source sends data cells with EFCI set to 0 and sends RM cells every n data cells.  The RM cells contain desired explicit rate(ER), current cell rate (CCR) and congestion indication(CI).  The source initializes CCR to the allowed cell rate(ACR) and CI to 0. Source Dest ATM node RM cell CI=0 if no congestion CI=1 otherwise User cell EFCI=0 RM cell CI=1 N rm

25. Multimedia & Comm. Lab 25 EPRCA (cont’d) Switch behavior  computes a mean allowed cell rate(MACR) for all VCs using: MACR = (1 -  ) * MACR +  *CCR  and the fair share as a fraction of this average  The ER field in the returning RM cells are reduced to fair share.  May set the CI bit in the cells passing. Destination behavior  monitors the EFCI bits and mark the CI bit in the RM cell if the last seen data cell had EFCI bit set. Problems  congestion detection is based on the queue length.  This method is shown to result in unfairness.  Sources that start up late may get lower throughput than those start early

26. Multimedia & Comm. Lab 26 ERICA  Explicit Rate Indication for Congestion Avoidance Switch behavior  Set target rate at 95% of link bandwidth  Monitor input rate and number of active VCs k  Overload = Input rate / Target rate  VC’s share = VC’s current cell rate / Overload  Fairshare = Target rate / k  ER = Max(Fairshare, This VC’s share)  ER in Cell = Min(ER in Cell, ER) Features  Measured overload/load at switch.  Small queue lengths during steady state.  Fast response.

27. Multimedia & Comm. Lab 27 Source level rate adaptation  Architecture  Encoder-level rate shaping  Rate shaping for precoded video

28. Multimedia & Comm. Lab 28 Rate control architecture Quantizer Rate shaper Rate control ATM Networks Output buffer Quantization level index MPEG Codec Rate Shaper Feedback

29. Multimedia & Comm. Lab 29 Rate shaping for precoded video  Frame discarding  Selective Block dropping  Eliminate some DCT coefficients  Block dropping with Error concealment  Feature-based block dropping

30. Multimedia & Comm. Lab 30 Examples of Rate control  Explicit Backward Congestion Notification  Composite Rate Control Scheme  Weighted Max-Min Fairness  Shaped VBR

31. Multimedia & Comm. Lab 31 EBCN Using queue occupancy of the VP buffer of an ATM switch RM cells : increase / decrease quantizer step size q new = max[q old + q diff, q max ] TcTc TnTn Video Sources Server Buffer occupancy max 0 Explicit Backward Congestion Notification M. Ghanbari - Essex Univ. GLOBECOM ‘96

32. Multimedia & Comm. Lab 32 CRCS  Composite Rate Control Scheme From  S. Karademir - Garleton Univ in Canada  GLOBECOM ‘96 Congestion Notification  Explicit Feedback mechanism  Feedback cell Traffic prediction  Prediction parameter  feedback info.  Average transmission rate during the last two cycles.  Prediction Model  TES : nonlinear auto-regressive model

33. Multimedia & Comm. Lab 33 WMMF  Weighted Max-Min Fairness from  T.V. Lakshman - Bell Labs.  INFOCOM ‘97 Design Goals  simple admission control  high statistical multiplexing gain  frequent bandwidth negotiation  adaptation of source rates to match available bandwidth  maintain low end-to-end delays Key Idea  RCBR  associate a weight with each flow

34. Multimedia & Comm. Lab 34 WMMF (cont’d)  RCBR Renegotiated CBR  hybrid of the CBR and VBR  the simplicity of admission control for CBR  the greater statistical multiplexing gains of VBR. Key Points  short-term fluctuations are absorbed in local source buffers  long-term changes make the source renegotiate the bandwidth  Weighted fair share Key  Using difference of flow activity

35. Multimedia & Comm. Lab 35 WMMF (cont’d)  Source Adaptation Mechanism Demand Prediction  Discrete Auto-regressive model  X n+k =  +  k (X n -  )  correlation efficient  : mean number of cels per frame  Gamma-Beta Auto-regressive Model  Heymann ‘96 Encoder Rate Adaptation  rate adaptation function  avg = Tr avg - [  * (B p - SETPOINT) / T horizon ]  Tr avg : Transmission Rate  B p : Predicted Buffer

36. Multimedia & Comm. Lab 36 SVBR  Shaped VBR From  M. Hamdi - ENST  IEEE JSAC Aug. ‘97 Key Idea  CBR 의 장점과 VBR 의 장점을 혼합한 형태  비디오 전체에 대해 VBR 로 인코딩  CBR 의 단점인 buffer delay 를 제거  Bursty 한 부분의 영역에 대해서는 CBR 을 적용  VBR 의 burstiness 감소  Shaped Variable Bit Rate

37. Multimedia & Comm. Lab 37 SVBR (cont’d)  Rate Shaping Principle  하나의 GOP 에 할당되는 비트수의 최대값을 설정 (leak rate)  GoP 단위로 비트수를 계산하여 leak rate 를 넘으면 Quantization parameter Q 를 증가 GoP scale rate prediction  GoP 단위로 rate prediction 을 적용  다음 GoP 의 크기를 예측한 후, 해당 GoP 를 위한 Q 를 재조정  Q k+1 = Q k R k /R k+1  Q k : k 번째 GOP 의 quantization parameter  R k : k 번째 GOP 의 bit 수

38. Multimedia & Comm. Lab 38 Open Issues  Policing dynamic UPC control time lag estimation  Point-to-Multipoint Connections Branchpoint behavior  Priority service for RM cells  Virtual Source / destination

39. Multimedia & Comm. Lab 39 Conclusion  Rate control for ABR Services  Congestion Control algorithm  Source-level rate adaptation