Survey Adaptive Video Multicast over the Internet

1
SurveyAdaptive Video Multicast over the Internet

• Jiangchuan Liu ? Bo Li and Ya-Qin Zhang
• IEEE Multimedia, January-March 2003

2
Outline

• Introduction
• Adaptation approaches for video multicast
• Support for video coding techniques
• Evaluation procedures and metrics
• Conclusions

3
Introduction

• Because network conditions are dynamic and
  theres no quality of service (QOS) guarantee in
  the current best-effort Internet, bandwidth
  adaptability becomes an essential requirement.
• Two challenges for adaptive video multicast over
  the internet
• The size of a multicast session
• Internet is a heterogeneous network

4
Introduction (cont.)

• An adaptation algorithm should thus maximize the
  total system throughput and fairly distribute
  video data to each receiver commensurate with its
  individual demand.

5
Adaptation Approaches for Video Multicast

• Single-rate adaptation (single-rate ,end-to-end)
• Simulcast (multirate, end-to-end)
• Layered adaptation (multirate, end-to-end)
• Agent-base adaptation (multirate, active)

6
Single-Rate Adaptation

• The sender maintain one video stream and adjusts
  the rate based on the receivers states.
• To reflect the dynamic network condition ,the
  state information should be sent back to the
  sender on time.

7
Single-Rate Adaptation (cont.)

• To solve feedback implosion problem
• Sender sends out a randomly generated 16-bit key
  and a number indicating how many digits of the
  key are significant.
• Each receiver also generates a key and responds
  to the solicitation only if the keys and the
  states match.
• The receivers in a session arent fairly treated.

8
Simulcast
9
Simulcast (cont.)

• A representative simulcast protocol is the
  destination set grouping (DSG) protocol.
• In DSG, the source maintains three streams
  carrying low-, medium- ,and high-quality versions
  of the original video.
• The receiver can change to another stream when
  the current one cant satisfy its requirement.

10
Layered Adaptation

• To infer the receivers capacities together with
  their locations though an end-to-end probing
  algorithm.
• A commonly used decomposition scheme is
  cumulative laying
• Base layer
• Enhancement layer

11
Layered Adaptation (cont.)

• Network-Driven
• Base layer is highest priority
• Enhancement layer is lower priority
• When there isnt enough bandwidth, routers
  discard the lowest-priority packets, thereby
  preventing loss of the base layer or
  high-priority enhancement layers.

12
Layered Adaptation (cont.)

• Receiver-Driven
• A receiver periodically joins a higher layers to
  explore the available bandwidth.
• If packet loss exceeds some threshold after the
  join experiment, when congestion occurs-the
  receiver should leave the group. Otherwise it
  will stay at the new subscription level.

13
Layered Adaptation (cont.)

• Receiver-Driven Layered Congestion Control (RLC)
• The sender temporarily increases the sending rate
  on a layer, and a receiver will join a higher
  layer only if theres no packet loss during this
  experiment.

14
Agent-Base Adaptation

• We can best support or enhance many applications
  using information or intelligent services only
  available inside a network.
• Two service has embodied in the design of RTP
• RTP-level mixer
• Translator

15
Agent-Base Adaptation (cont.)
16
Support for Video Coding Techniques
17
Support for Video Coding Techniques (cont.)
18
Support for Video Coding Techniques (cont.)

• Transcoding
• A straightforward approach for transcoding is to
  decompress the video stream, process it , and
  then recompress it.

19
Support for Video Coding Techniques (cont.)

• Scalable Coding

20
Support for Video Coding Techniques(cont.)

• Object Scalability

21
Evaluation Procedures and Metrics

• There are two fundamental issues we must address
  in simulation
• The systems offer load
• Its performance characteristics
• Source traffic model
• Use two parameters A and P
• A is the average number of packets generated per
  interval.
• P is characterizes the packet distribution in the
  interval.

22
Evaluation Procedures and Metrics (cont.)

• Local metrics
• Bandwidth
• Loss rate
• Delay
• Delay jitter
• PSNR
• Global metrics
• Convergence time
• Fairness index

23
Conclusion
24
Conclusion (cont.)

• Agent deployment is mainly subject to the
  management policies or market strategies of
  network operators and service providers.