Download presentation

Presentation is loading. Please wait.

Published byOlivia Moore Modified over 4 years ago

1
Clustering of Source/Channel Rate Allocations for Receiver-driven Multicast under a Limited Number of Streams Philip A. Chou, Microsoft Research Kannan Ramchandran, UC Berkeley

2
Multicast Sender sends one stream (of packets) into the network; the stream is replicated as necessary in the network to reach interested receivers. Scales to an unlimited number of receivers. Potentially good for Internet radio, TV, etc. RRRR S RR …

3
Problem: Heterogeneity Multicast over a wide area proceeds over a collection of packet erasure channels whose bandwidths and erasure processes are –Essentially unknown [except in broad statistical sense] –Variable from receiver to receiver –Time-varying RRRR S RR … Collection of packet erasure channels

4
Receiver-driven Multicast Sender sends multiple streams into the network, each tailored to a different channel characteristic (in terms of bandwidth or reliability) Each receiver subscribes to the stream that best match its channel characteristic (can switch over time) RRRR S RR …

5
Existing Work Deering (1988): IP Multicast McCanne (1996): Receiver-driven Layered Multicast –Sender sends out each layer in a separate stream –Each receiver subscribes to as many streams as will fit –Addresses bandwidth heterogeneity Tan & Zakhor (1999, 2000): RLM w/FEC Chou et al. (1999, 2000): RLM w/FEC+ARQ –Generalizes RLM to include layered channel codes –Receiver subscribes to optimal collection of source and parity layers (to minimize distortion for available rate) –Addresses both bandwidth & packet loss heterogeneity –Problem: dozens (or even hundreds) of streams

6
Clustering Receivers Questions to be answered: –How large should M be to serve most receivers well? –How can we design the collection of M streams? –How can a receiver decide which of the M streams to use? We will assume streams are all at the same bitrate. Redundancy is provided by FEC. Space of associated channels Distribution over channels 1 source parity 2…M

7
Existing FEC Systems Commercial systems (e.g., Windows Media) use systematic Reed-Solomon code to produce N-K parity packets for every K source packets The parameters (N,K) are chosen to match the packet loss characteristics for the channel source parity 100Kbps K N for more reliable channelsfor less reliable channels

8
Existing FEC Systems Priority Encoding Transmission (PET, Albanese et al., 1996) is similar, but it allows K to change across source layers with different importance. 100Kbps K2K2 K3K3 K1K1 parity source N

9
PET packetization Property: recover layer i iff receive K i packets (out of N) Albanese et al. (1996) use 3 layers (I,P,B), dont optimize Davis & Danskin (1996) optimize K i s for any number of layers for minimum distortion Mohr, Riskin, & Ladner (1999) assume fine grain scalability (e.g., SPIHT) and adjust breakpoints using greedy search Puri & Ramchandran (1999) optimize breakpoints using O(N) algorithm

10
Optimal Stream for a Known Channel Wolog assume N layers, layer i 1,…,N coded with K i =i. Let R=(R 0,R 1,…,R N ) be breakpoint vector, where R 0 0 and R 1,…,R N index the last byte in layers 1,…,N respectively. Let D(R 0 ), D(R 1 ), …, D(R N ) be the corresponding vector of distortions if R 0, R 1,…, R N source bytes are recovered. Let q=(q 0,q 1,…,q N ) be probability mass vector, where q k =Pr{1 st k of N layers recovered}=Pr{k of N packets received}. The effect of any stationary packet erasure channel on the receivers expected distortion is through q=(q 0,q 1,…,q N ). R0R0 R1R1 RNRN R D(R0)D(R0) D(R1)D(R1) D(RN)D(RN) Operational D(R) function

11
Expected Distortion and Rate Expected Distortion is Transmission rate (bytes per GOF) is where k =N/(k(k+1)) for k=1,…,N-1 and N =1. Finding R=(R 0,R 1,…,R N ) that minimizes D(R) s.t. R(R) R* can be found by minimizing D(R)+ R(R) for some using the O(N) algorithm of Puri & Ramchandran.

12
Optimal Stream for a Collection of Channels Let {q } be a collection of channels indexed by Є over which there is a distribution Expected distortion of PET packetization R for channel q is Overall expected distortion (w.r.t. ) is Hence to min D(R) s.t. R(R) R*, find q=q and use P&R.

13
Multiple Optimal Streams for a Collection of Channels Start with M streams with PET packetizations R 1,…,R M. Let m( ) be stream number to which receiver with channel q should subscribe. Optimal m( ) (minimizing overall expected distortion) is m( ) = argmin m D (R m ) = argmin m q,k D(R m,k ), which induces partition cells m ={ :m( )=m}. Optimal PET packetization R m for cell m is which can be solved by the Puri-Ramchandran algorithm. Repeat. 1 2 … M

14
Simulation Setup We simulate collection of iid packet erasure channels with N=40, ~ Beta(1,b), mean =1/(1+b)=.10,.15,.20. We assume D(R)= 2 -2cR, R = #bytes per GOF, R*=7/c. Find clusters with M=1,2,4,8,16,32. Beta(1,1/ -1) distribution

15
Simulation Results

16
Conclusion We have presented a clustering algorithm that finds the set of M streams having source/channel rate allocations (PET packetizations) that optimally covers the space of packet erasure channels under an arbitrary distribution –nearest neighbor performed by N-dim dot product –centroid is performed by O(N) algorithm For typical (?) distribution of channels, 4 streams can gain 4 out of a possible 5 dB (i.e., loses only 1 dB compared to an infinite number of streams).

Similar presentations

OK

Receiver-driven Layered Multicast S. McCanne, V. Jacobsen and M. Vetterli University of Calif, Berkeley and Lawrence Berkeley National Laboratory SIGCOMM.

Receiver-driven Layered Multicast S. McCanne, V. Jacobsen and M. Vetterli University of Calif, Berkeley and Lawrence Berkeley National Laboratory SIGCOMM.

© 2018 SlidePlayer.com Inc.

All rights reserved.

To make this website work, we log user data and share it with processors. To use this website, you must agree to our Privacy Policy, including cookie policy.

Ads by Google

Ppt on ready mix concrete plant Ppt on solar system download Ppt on mind reading phones Ppt on electricity in india Ppt on introduction to object-oriented programming encapsulation Slide backgrounds for ppt on social media Ppt on electricity from waste Ppt on greenhouse effect and global warming Ppt on organic farming in india Ppt on south indian cuisine