Bandwidth Allocation for Layered Multicasted Video Hugh M.Smith, Matt W. Mutka and Eric Torng Dep. Of Computer Science and Engineering, Michigan State.

Bandwidth Allocation for Layered Multicasted Video

Hugh M.Smith, Matt W. Mutka and Eric TorngDep. Of Computer Science and Engineering, Michigan State University

IEEE International Conference on Multimedia Computing and Systems

Intro.

• Sender based rate-adaptation– The sender uses a single channel to transmit the video s

ignal to all receivers,and the sender adjusts the transmission rate of this single channel based upon network or receiver feedback.

• Receiver based rate control protocol– The sender stripes the video signal across multiple mult

icast channels, and each receiver selectively adds and drops channels to meet their individual need.

Intro(cont’d)

• Layered Multicast Control Protocol(LMCP)– New protocol which melds the strengths of the t

wo previous approaches.

• Present and analyze three different algorithms the sender might use for computing the transmission rates for each channel.

Multiple-Layered video encoding and transmission

• Striping is used to transmit a video signal across multiple channels.

• Splitting a single video signal into multiple segments and spreading the segments across multiple multicast channels.

• For this, the video decoder must be able to decode the stream without requiring all segments to be present(Hierarchical or embedded encoding)

Receiver Driven Layered Multicast(RLM)

• This Protocol focuses on how receivers add and drops network connections in order to meet their bandwidth restrictions.

• This approach assumes the network supports IP-Multicast with fast join-leave capabilities as developed in the IGMP v2.

• Add additional layers,RLM uses “join-experiments”

Layered Multicast Control Protocol(LMCP)

• LMCP combines the strength of both the sender based rate-adaptation protocol and the RLM protocol

• The receivers not only perform the basic RLM protocol, they also dynamically approximate their available bandwidth and provide this as feedback to the sender

LMCP sender transmission rate algorithm

• Given n receivers where receiver i can receive ri units of transmission,find a set of k cumulative transmission rates tj

• Minmizing where tj is the max cumulative transmission rate such that

• That is, finding an optimal mapping from

n

i

ji tr1

)(

ij rt

ji tTrRwhereTR set and

• Dynamic programming(optimal)– Runtime complexity O(kn3), space usage O(n2)

• Non-optimal divide and conquer– O( *kn),for reasonable k(k<6)and n>>k, this approa

che has a significant runtime performance advantage over dynamic programming

• Fixed Percentile– Tj=Roffset*(j) ,where offset=

k

k2

k

n

Statistical analysis

Simulations

Conclusions

• LMCP protocol incorporates both sender and receiver rate-adaptation approaches.

• The receivers continue to add and drop layers as their available resources dictate,and also send bottleneck rates to the sender as feedback

• the sender uses this feedback to adjust the transmission rates for each of the multicast channels

Conclusions(cont’d)

• These two algorithm were dynamic;One algorithm produces optimal transmission rates while a second algorithm produces near optimal transmission rates but is a more scalable approach.

• Analysis showed that both algorithms deliver a high percentage-used for all three distributions

• in simulations with 5 transmission channels,this protocol achieved an average percentage-used of 87%.

Future Work

• Looking at simulations the examine how the LMCP protocol shares the network with other applications.

• Examining the effects of different router and network characteristics on the algorithm

• Specially,looking at different queuing technologies such as RED(random Early Detection Gateways for congestion avoidance)