Assignment #3
Download
Report
Transcript Assignment #3
Networked Life: 20 Questions and Answers
(M. Chiang, Princeton University)
Q17: IPTV and Netflix: How can the
Internet Support Video?
Prof. Hongseok Kim
© 2014 Networking for Information Communications and Energy Lab.
Video watching is changing
Content type
»
User generated as well as licensed
When
»
DVR on IPTV, HBO Go …
Where
»
Anywhere with Internet connectivity
How
»
Almost any device
2
How much
Hulu : 31 million unique viewers in Feb. 2012
Comcast : 39 million, watching 205 million videos
US : 100 million IPTV users
YouTube and Netflix : half of Internet usage
3
classification1
Precoded
»
Vast majority is precoded
»
The content is already encoded and stored somewhere
Real time
»
Sports, news, weather
Two way interactive
−
Online gaming
−
Video conferencing
4
classification2
Download
Streaming
Netflix and YouTube
»
Device does not keep a local copy
5
classification3
Channelized
»
Follow the schedule of each channel accordingly
−
TV, DVR
On demand
»
VoD(Video on Demand)
−
YouTube, Netflix, some premium TV
NVoD
»
Near Video on Demand
»
Which staggers the same channel every few minutes, so that within a
latency tolerance of that few minutes, you get the experience of VoD
6
classification4
Multicast
Extreme form : broadcast
»
Everyone is in the multicast group
Unicast
7
IPTV
Over private and managed network, often with a set-top box
on consumer premise
IP convergence
Cost
Flexibility
compression
8
IPTV
9
VoI
Over public networks
Client-server without fee
»
YouTube, ABC and BBC
Client-server with fee
»
Netflix, Amazon Prime, Hulu Plus, HBO Go
P2P
10
VoI
11
Quality
Bit rate / distortion
»
Motion
»
Screen resolution
»
Viewing distance/screen size
»
Efficiency of compression
»
SD, HD, UltraHD
Delay
Jitter
12
Questions
How can the pipe take on so many bits per second?
How to keep track of video?
How to support quality of service over best effort network?
13
Layers
14
Video
Sequence of frames moving at a particular speed
Each frame is a picture consists of pixels
Each pixel is {colors, luminance} encoded in bits
Bit rate = bits per frame * number of frames per second
15
Compression
Remove redundancies in signals
Lossless compression
Lempel-ziv
Lossy compression
Tradeoff between compression ratio and resulting fidelity
16
Rate distortion curve
17
Which bitrate?
Distortion tolerable
Channel condition
Usage quota
18
What to compress
Redundancies
Frame-to-frame similarities
Human visual limitations
Transform coding
Statistical structures
Huffman coding
19
Standards
MPEG1: 1992
»
VCD, 1Mbps
MPEG2(H.262): 1996
»
DVD, 10Mbps
MP3
»
Standard for the online music industry
»
12:1 compression ratio
MPEG4: 2000
MPEG4 Part 10(H.264): 2004
»
HDTV, 15-20Mbps, Blu-ray, 40Mbps
H.261, QuickTime, Windows, Media Player, Flash, Real
Media…
20
Inter-frame prediction
21
Metrics
Bitrate efficiency
»
If GoP longer, bit rate becomes lower
Error resilience
»
If an I frame lost
Instant channel change
»
For channelized video content, the ability to change channels fast is
important
22
Example
23
I frame dropped
24
P frame dropped
25
B frame dropped
26
Application layer
IGMP
Membership-query
Membership-report
Leave-group(optional)
Unicast help
SIP
27
RTSP
28
Transport layer
UDP
Connectionless
Differences compared to TCP
No congestion control
No retransmission
Latency vs. reliability(tradeoff)
RTP
29
Another use of UDP
Network management protocols
SNMP
RIP
Number of states and number of parallel sessions
Handshake and tear down overhead
Header overhead
»
8bytes(UDP)<20bytes(TCP)
30
Latency-jitter tradeoff
31
Optimization
»
𝑷𝒊 : the time at which packet i is displayed to the user
»
𝑽𝒊 : the time at which packet i is transmitted(unit step)
»
𝑨𝒊 : the time at which packet i is received
32
Solution
33
34
35
Summary
IP has become the basis of video content
Content being decoupled from delivery channels and
devices
Quality of service provided through different mechanisms
Layering in action
36
Thank you!
Networking
Next
Information
Innovative
Communications
Creative
Energy
Envisioning