1. Internet-based Video Content Distribution
Anirban Mahanti
Principal Researcher
Networks Research Group, NICTA

2. Is video the next big thing on the Internet?
Flashback to the 1990s
Is video the next big thing on the Internet?
Yes, video-on-demand and live streaming applications will generate large amounts of traffic.
What architectural support do we need to support Internet video application?
Best effort, point-to-point, Internet not suitable for video
Need point-to-multipoint (multicast) with quality-of-service support in the network

3. Early Video-related Research …
Integrated & differentiated services
Control/signaling protocols
Native IP multicast
Application-layer multicast
Video scheduling protocols
Rate control protocols
Layered video coding
Peer-to-peer systems …

4. Early Video-on-Demand Trials
E.g., Cambridge trial, Time Warner Trial (1994)

5. Request Aggregation Protocols (1994-2002)
Serve multiple client requests together, either partially or entirely
Network transmissions may be multicast (save both server and network resources) or unicast (save only server resources such as disk bandwidth)
Consider on-demand streaming of a hot file. We want to minimize:
“server bandwidth” usage for the file
Maximum start-up delay for the file

6. A (Simple) Batching Example
Playback rate = 1 Mbps, duration = 90 minutes
Group requests in non-overlapping intervals of 30 minutes:
Max. startup delay = 30 minutes
Bandwidth required = 3 channels = 3 Mbps
Bandwidth increases linearly with decrease in startup delay 30 60 90
120
150
180
210
240
Time (minutes)
Channel 1
Channel 2
Channel 3

7. Periodic Broadcast Example
Partition the media file into 2 segments with relative sizes {1, 2}. For a 90 min. movie:
Segment 1 = 30 min, Segment 2 = 60 min
Max. startup delay = 30 min
Bandwidth required = 2 channels = 2 Mbps
Disadvantage: requires increased client capabilities
Time (minutes) 1 2 30 60 90
120
150
180
Channel 1
Channel 2

8. Lower Bound for Periodic Broadcast

9. Optimized Periodic Broadcast (2001)
r = segment transmission rate,
s = max. # streams client listens to concurrently
b = client data rate = s x r
Our performance is better than SB. Performance appears to improve with increasing “s”, ie, as the streaming rate on the channels decreases. Why? We investigate this next.
“Scalable On-demand Media Streaming with Packet Loss Recovery”, In Proc of ACM SIGCOMM 2001.

10. Video Landscape (1990s, Early 2000s)
Lack of compelling content
Lack of broadband penetration
Videos played using media players
Web browser, media player integration?

11. February 2005: Arrival of YouTube
Start of the Video era …
Ordinary Web users become content publishers (democratization of the Web)
65,000 new videos uploaded and 100 million videos viewed per day (in July 2006)
48 hours of new video uploaded each minute (May 2010)

12. 2005/06: Social Media Revolution
Content publishing democratized
Online social networks allow users to share content
Contexts: friends, family, work
Conversation is ubiquitous
Twitter, Facebook, Discussion forums, ‘Comments’ on traditional media articles, Fan fiction

13. Time Magazine Person of the Year (2006)

14. BBC iPlayer (2007): Internet TV Service
The original iPlayer release used peer-to-peer technology.

15. Today We Have …

16. What were the enabler? Technology Business models
Broadband penetration
Hardware, software
Media player & browser integration
Business models
Availability of content (UGC + others)
Advertisement
Premium content
Tapping new audience

17. Video Traffic Today
Source: Sandvine (2011)

18. The Future of Video Traffic
Video will account for an even larger share of Internet traffic
More video traffic on cellular networks
Integration of TV, social media, and smart devices

19. “Elephants” and “Mice” of the Internet
“YouTube Workload Characterization: A View from the Edge”, In Proc of ACM Internet Measurement Conference, 2007

20. Long-Tail and Content Distribution
“YouTube Workload Characterization: A View from the Edge”, In Proc of ACM Internet Measurement Conference, 2007

21. Churn in Video Popularity
“Characterizing Web-based Video Sharing Services”, ACM Trans on Web, 2010.

22. The Future of Video Traffic
Video will account for an even larger share of Internet traffic
More video traffic on cellular networks
Integration of TV, social media, and smart devices

23. Increase of Video in Mobile Networks

24. Consequences
wwwsmh.com.au

25. The Future of Video Traffic
Video will account for an even larger share of Internet traffic
More video traffic on cellular networks
Integration of TV, social media, and smart devices

26. The NICTA Social TV Project
Contact Sebastien Ardon nicta.com)

27. The Good Old Television …
“Lounge room Entertainment” has traditionally been a social/ family affair
People experience media together by sharing physical space

28. Social TV: Social Media meets TV
Blur between Broadcast, User-Generated, Online
In the Lounge and on Devices
Devices interactions
TV still king: the largest viewing surface in the house
Personalization is pervasive
Contextual-dependent: person, time, space, who’s there.
The ‘Conversation’ is restored to the lounge room
Real-time and non real-time interactions
Social aspect is restored, but not necessarily through sharing physical space, -> sharing virtual space.
Plethora of media offering means better personalization technology is needed (e.g. Recommendations, collaborative filtering)

29. Services Examples
Broadcast TV + Twitter in remote with automatic hash tag
TV auto-join Facebook fan page when starting to watch a show. Live conversation takes place on fan page.
Personalized group television: commonly liked shows, from common friends
Smartphone remote & User Input:
Check-in a TV show using your phone as remote

30. Social TV: What are the barriers?
On-Demand TV: cost of network bandwidth for Content Providers
From the broadcast to the on-demand model
Recommendation Systems
Live TV
Catch-up TV
Technical Hurdles
Multi-device interactions, lack of standards

31. Video workload differs significantly from traditional (Web) workloads
Concluding Remarks
Video and social media likely to be a significant driver for next-generation services and applications
Video workload differs significantly from traditional (Web) workloads
Scalability/cost, quality of experience guarantees
Support for multi-device interactions
Standards
New imaginative applications