Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Modelling IPTV Services

789 views

Published on

  • Be the first to comment

  • Be the first to like this

Modelling IPTV Services

  1. 1. Modelling IPTV Services Fernando Ramos* fernando.ramos@cl.cam.ac.uk *Together with Jon Crowcroft, Ian White, Richard Gibbens, Fei Song, P. Rodriguez
  2. 2. Outline   Introduction: what IPTV is not, and why do we care   Motivation to model IPTV services   The IPTV traffic model, in some detail (W.I.P.)   Conclusions Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  3. 3. What IPTV is not X p2p TV X web TV cable-like TV service X on-demand video service live TV IPTV is a cable-like TV service offered on top of an IP network Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  4. 4. Why do we care with IPTV?   One of the fastest growing television services in the world [1]   2005: 2 million users   2007: 14 million users   ...and growing   High bandwidth and strict QoS requirements   Big impact in the IP network [1]
Parks
Associates.
Tv
services
in
Europe:
Update
and
Outlook,
2008 Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  5. 5. Overview of an IPTV network Internet IP Network TV Head End Customer Premises TV . . STB . DSLAM PC Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  6. 6. Motivation – Why do we need a realistic IPTV Traffic Model?   Brand new service on top of an IP network   User behaviour very different from other IP-based applications Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  7. 7. Motivation – Why do we need a realistic IPTV Traffic Model?   To evaluate different delivery systems for IPTV   To evaluate different network architectures for IPTV Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  8. 8. The dataset   We have analysed real IPTV data from one of the largest IPTV service providers   ~ 6 months worth of data   ~ 250,000 customers   ~ 620 DSLAMs   ~ 150 TV channels   NB:We consider a user is zapping if he switches between 2 TV channels in less than 1 minute. Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  9. 9. IPTV Traffic Model   Workload characteristics   Zapping blocks containing a random number of switching events (zapping period)   Separated by watching/away periods of random length Zap blocks 30 Switching Event Number 25 20 15 10 5 Inter-zap block periods 0 06:00 07:12 08:24 09:36 10:48 12:00 13:12 14:24 15:36 Time of the day Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  10. 10. IPTV Traffic Model WATCHING MODE ZAPPING MODE Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  11. 11. IPTV Traffic Model - Detailed 1 1 10 100 1000 Proportion (CDF) Empirical data Gamma CDF Exponential CDF Gamma 2 CDF 0.1 Inter zap block interval (minutes) [100 DSLAMs, 40k users] Findings: Empirical data fits with 2 gamma and 1 exponential (consistent across regions) To do: Check consistency for different channels Check consistency for period of the day Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  12. 12. IPTV Traffic Model - Detailed 1 0.9 0.8 Proportion (CDF) 0.7 0.6 Empirical data 0.5 Gamma CDF 0.4 0.3 Geommetric CDF 0.2 Poisson CDF 0.1 0 1 10 Zap block size (log) [100 DSLAMs, 40k users] Findings: Empirical data fits with gamma distribution (consistent across regions) To do: Check consistency for period of the day Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  13. 13. IPTV Traffic Model - Detailed 1.E+00 Normalised channel access (log) 1 10 100 1.E-01 1.E-02 1.E-03 Normalised watching counter 1.E-04 Zipf-like distribution 1.E-05 1.E-06 Channel index sorted by popularity (log) [49 DSLAMs, 16k users] Findings: Popularity is a) Zipf-like for top channels, b) decays abruptly for non-popular ones. To do: Add dependency of previous channel. Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  14. 14. Conclusions   Preliminary results of an IPTV Workload model were presented   Some of the main findings:   Workload characteristics: Burst (zapping) periods separated by watch/way periods   Popularity: a) Zipf-like for top channels, b) decays fast for non- popular ones   Watching period empirical data fits with 2 gamma and 1 exponential distributions   Number of channels in a zap period fits with gamma distribution   See you at the SIGCOMM Poster Session!  Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009
  15. 15. THANK YOU! fernando.ramos@cl.cam.ac.uk Fernando Ramos, Cosener's Multi-Service Networks 9 July 2009

×