Presentation is loading. Please wait.

Presentation is loading. Please wait.

A Comparison of Layering and Stream Replication Video Multicast Schemes Taehyun Kim and Mostafa H. Ammar Networking and Telecommunications Group Georgia.

Similar presentations


Presentation on theme: "A Comparison of Layering and Stream Replication Video Multicast Schemes Taehyun Kim and Mostafa H. Ammar Networking and Telecommunications Group Georgia."— Presentation transcript:

1 A Comparison of Layering and Stream Replication Video Multicast Schemes Taehyun Kim and Mostafa H. Ammar Networking and Telecommunications Group Georgia Institute of Technology Atlanta, Georgia

2 Research Goal  A systematic comparison of video multicasting schemes designed to deal with heterogeneous receivers  Replicated streams  Cumulative layering  Non-cumulative layering

3 Stream Replication  Multiple video streams  Same content with different data rates  Receiver subscribes to only one stream  Example  DSG (Cheung, Ammar, and Li, 1996)  SureStream of RealNetworks   Intelligent streaming of Microsoft 

4 Replicated Stream Multicast

5 Cumulative Layering  1 base layer + enhancement layers  Base layer  Independently decoded  Enhancement layer  Decoded with lower layers  Improve the video quality  Example  RLM (McCanne, Jacobson, Vetterli, 1996)  LVMR (Li, Paul, and Ammar, 1998)  MPEG-2/4, H.263 scalability modes

6 Layered Video Multicast

7 Layering or Replication?  Common wisdom states: “Layering is better than replication”  But it depends on  Layering bandwidth penalty  Specifics of encoding  Protocol complexity  Topological placement of receivers

8 Bandwidth Penalty  Information theoretic results  R(P,  2 )  R(P,  1,  2 )  Packetization overhead  Syntactically independent layering  Picture header  GOP information  Macroblock information

9 Experimental Comparison

10 Comparison by DP J. Kimura, F. A. Tobagi, J. M. Pulido, P. J. Emstad, "Perceived quality and bandwidth characterization of layered MPEG-2 video encoding", Proc. of the SPIE, Boston, MA, Sept. 1999"

11 Providing a Fair Comparison  Need to insure that each scheme is optimized  Two dimensions  Selection of stream/layer rates  Assignments of streams/layers to receivers

12 Rate allocation  Cumulative layering  Optimal receiver partitioning algorithm (Yang, Kim, and Lam)  Stream replication  Cumulative rate allocation

13 Stream assignment  Cumulative layering  Assign as many layers as possible  Stream replication  Greedy algorithm

14 Comparison Methodology  Model of network  Topology  Available bandwidth  Placement of source and receivers  Determine optimal stream rates and allocation  Evaluate performance

15 Performance Metrics  Average reception rate  Total bandwidth usage  Average effective reception rate  Efficiency 

16 Network Topology  GT-ITM  Number of server = 1  Number of receivers = 1,640  Number of transit domains = 10  Number of layers = 8  Amount of penalty = 25%

17 Data reception rate

18 Bandwidth usage

19 Effective reception rate

20 Efficiency

21 Effect of overhead

22 Effect of the number of layers

23 Clustered Distribution  Topology consideration  Layering favors clustered receivers  Stream replication favors randomly distributed receivers  Simulate when receivers are clustered within one transit domain

24 Effective reception rate

25 Protocol Complexity  Layered video multicasting  Multiple join for a receiver  Large multicast group size  Replicated stream video multicasting  One group for a receiver  Small multicast group size

26 Average group size

27 Conclusion  Identified the factors affecting relative merits of layering versus replication  Layering penalty  Specifics of the encoding  Topological placement  Protocol complexity  Developed stream assignment and rate allocation algorithm  Investigated the conditions under which each scheme is superior

28 Optimal Quality Adaptation for MPEG-4 Fine-Grained Scalable Video Taehyun Kim and Mostafa H. Ammar Networking and Telecommunications Group Georgia Institute of Technology Atlanta, Georgia

29 Related Work (1/2)  S. Nelakuditi, et al, “Providing smoother quality layered video stream,” NOSSDAV 2000  Goals  Achieving smoother quality for layered CBR video using receiver buffer  Minimizing quality variation (maximizing runs of continuous frames)

30 Algorithm  Forward scan  Switching between select and discard phase  Entering select phase if buffer is full  Entering discard phase if buffer is empty  Backward scan  Exploiting the residual buffer  Extending each run

31 Bandwidth Model

32 Experimental Result

33

34 Related Work (2/2)  D. Saparilla, et al, “Optimal streaming of layered video,” INFOCOM 2000  Goal  Investigating the bandwidth allocation problem to minimize loss probability  Modeling the source video and the available bandwidth by stochastic process

35 Main Result  Static policy  Allocating bandwidth in proportion to long run average data rate  Optimal for infinite length, independent layering  Threshold-based policy  If the base layer buffer is below a threshold, allocate bandwidth to the base layer

36 Research Goal of MPEG4 FGS Quality Adaptation  Maximization of the perceptual video quality by minimizing quality variation  Accommodation of the mismatch between  Rate variability of VBR video  Available bandwidth variability

37 MPEG4 FGS Hybrid Scalability  Base layer  Enhancement layer  FGS layer: improving video quality  FGST layer: improving temporal resolution

38 Rate Variability

39 Quality Adaptation Framework C[k]: transmission resource constraint X[k]: cumulative data size S[k]: cumulative selected data size d: threshold

40 Optimal Quality Adaptation  Threshold should be equal to the receiver buffer size to achieve  Minimum quality variability  Necessary condition of maximum bandwidth utilization

41 Online Adaptation  Estimating the threshold point without assuming the available bandwidth information in advance  The available bandwidth is estimated by an MA style linear estimator

42 Experiment Model

43 Bandwidth Variability  TCP  TFRC

44 Performance over TFRC  Threshold-based streaming (Infocom’00)  Online adaptation

45 Performance over TCP  Threshold-based streaming  Online adaptation

46 Conclusion  Accommodated the mismatch between the rate variability and the bandwidth variability  Developed an optimal quality adaptation scheme for MPEG4 FGS video to reduce quality variation  Investigated the perceptual quality of different algorithms and options


Download ppt "A Comparison of Layering and Stream Replication Video Multicast Schemes Taehyun Kim and Mostafa H. Ammar Networking and Telecommunications Group Georgia."

Similar presentations


Ads by Google