Transcript A Primer on Multimedia Network & YouTube & Netflix Case Studies

A Quick Primer on
Multimedia Networking
• Multimedia vs. (conventional) Data Applications
– analog “continuous” media: encoding, decoding & playback
– service requirements
• Classifying multimedia applications
– Streaming (stored) multimedia
– Live multimedia broadcasting
– Interactive multimedia applications
• Making the best of best effort service: streaming
Stored Multimedia over “Best-Effort “Internet
– client buffering, rate adaption, etc.
• Large-scale video delivery over the Internet: YouTube
and Netflix case studies (optional)
Required Readings:
csci4221 Textbook, sections 7.1-7.2
CSci4211:
Multimedia Networking
1
Multimedia and Quality of Service
Multimedia applications:
network audio and video
(“continuous media”)
QoS
network provides
application with level of
performance needed for
application to function.
CSci4211:
Multimedia Networking
2
Digital Audio
• Sampling the analog signal
– Sample at some fixed rate
– Each sample is an arbitrary real number
• Quantizing each sample
– Round each sample to one of a finite number of values
– Represent each sample in a fixed number of bits
4 bit representation
(values 0-15)
CSci4211:
Multimedia Networking
3
Audio Examples
• Speech
– Sampling rate: 8000 samples/second
– Sample size: 8 bits per sample
– Rate: 64 kbps
• Compact Disc (CD)
– Sampling rate: 44,100 samples/second
– Sample size: 16 bits per sample
– Rate: 705.6 kbps for mono,
1.411 Mbps for stereo
CSci4211:
Multimedia Networking
4
Why Audio Compression
• Audio data requires too much bandwidth
– Speech: 64 kbps is too high for a dial-up modem user
– Stereo music: 1.411 Mbps exceeds most access rates
• Compression to reduce the size
– Remove redundancy
– Remove details that human tend not to perceive
• Example audio formats
– Speech: GSM (13 kbps), G.729 (8 kbps), and G.723.3 (6.4
and 5.3 kbps)
– Stereo music: MPEG 1 layer 3 (MP3) at 96 kbps, 128
kbps, and 160 kbps
CSci4211:
Multimedia Networking
5
A few words about audio compression
• Analog signal sampled at
constant rate
– telephone: 8,000
samples/sec
– CD music: 44,100
samples/sec
• Each sample quantized, i.e.,
rounded
– e.g., 28=256 possible
quantized values
• Each quantized value
represented by bits
– 8 bits for 256 values
• Example: 8,000
samples/sec, 256 quantized
values --> 64,000 bps
• Receiver converts it back
to analog signal:
– some quality reduction
Example rates
• CD: 1.411 Mbps
• MP3: 96, 128, 160 kbps
• Internet telephony: 5.3 13 kbps
6
Digital Video
• Sampling the analog signal
– Sample at some fixed rate (e.g., 24 or 30 times per sec)
– Each sample is an image
• Quantizing each sample
– Representing an image as an array of picture elements
– Each pixel is a mixture of colors (red, green, and blue)
– E.g., 24 bits, with 8 bits per color
7
The
2272 x 1704
hand
CSci5221:
The
320 x 240
hand
Multimedia
8
A few words about video compression
• Video is sequence of
images displayed at
constant rate
Examples:
• MPEG 1 (CD-ROM) 1.5
Mbps
– e.g. 24 images/sec
• MPEG2 (DVD) 3-6 Mbps
• Digital image is array of
• MPEG4 (often used in
pixels
Internet, < 1 Mbps)
• Each pixel represented
Research:
by bits
• Layered (scalable) video
• Redundancy
– adapt layers to available
bandwidth
– spatial
– temporal
CSci4211:
Multimedia Networking
9
Video Compression: Within an Image
• Image compression
– Exploit spatial redundancy (e.g., regions of same color)
– Exploit aspects humans tend not to notice
• Common image compression formats
– Joint Pictures Expert Group (JPEG)
– Graphical Interchange Format (GIF)
Uncompressed: 167 KB
Good quality: 46 KB
CSci4211:
Multimedia Networking
Poor quality: 9 KB
10
Video Compression: Across Images
• Compression across images
– Exploit temporal redundancy across images
• Common video compression formats
– MPEG 1: CD-ROM quality video (1.5 Mbps)
– MPEG 2: high-quality DVD video (3-6 Mbps)
– Proprietary protocols like QuickTime and RealNetworks
CSci4211:
Multimedia Networking
11
MM Networking Applications
Classes of MM applications:
1) Streaming stored audio
and video
2) Streaming live audio and
video
3) Real-time interactive
audio and video
Jitter is the variability
of packet delays within
the same packet stream
CSci4211:
Fundamental
characteristics:
• Typically delay sensitive
– end-to-end delay
– delay jitter
• But loss tolerant:
infrequent losses cause
minor glitches
• Antithesis of data,
which are loss intolerant
but delay tolerant.
Multimedia Networking
12
Application Classes
• Streaming
– Clients request audio/video files from servers
and pipeline reception over the network and
display
– Interactive: user can control operation (similar
to VCR: pause, resume, fast forward, rewind,
etc.)
– Delay: from client request until display start
can be 1 to 10 seconds
CSci4211:
Multimedia Networking
13
Application Classes (more)
• Unidirectional Real-Time:
– E.g., real-time video broadcasting of a sport event
– similar to existing TV and radio stations, but delivery on
the network
– Non-interactive, just listen/view
• Interactive Real-Time :
– Phone conversation or video conference
– E.g., skype, Google handout, VoIP & SIP, …
– More stringent delay requirement than Streaming and
Unidirectional because of real-time nature
– Video: < 150 msec acceptable
– Audio: < 150 msec good, <400 msec acceptable
CSci4211:
Multimedia Networking
14
Multimedia Over Today’s Internet
TCP/UDP/IP: “best-effort service”
• no guarantees on delay, loss
?
?
?
?
?
?
But you said multimedia apps requires ?
QoS and level of performance to be
?
? effective!
?
?
Today’s Internet multimedia applications
use application-level techniques to mitigate
(as best possible) effects of delay, loss
CSci4211:
Multimedia Networking
15
Challenges
• TCP/UDP/IP suite provides best-effort, no
guarantees on expectation or variance of packet
delay
• Streaming applications delay of 5 to 10 seconds is
typical and has been acceptable, but performance
deteriorate if links are congested (transoceanic)
• Real-Time Interactive requirements on delay and
its jitter have been satisfied by over-provisioning
(providing plenty of bandwidth), what will happen
when the load increases?...
CSci4211:
Multimedia Networking
16
Internet (Stored) Multimedia:
Simplest Approach
• audio or video stored in file
• files transferred as HTTP object
– received in entirety at client
– then passed to player
audio, video not streamed:
• no, “pipelining,” long delays until playout!
CSci4211:
Multimedia Networking
17
Streaming Stored Multimedia
Streaming:
• media stored at source
• transmitted to client
• streaming: client playout begins
before all data has arrived
• timing constraint for still-to-be
transmitted data: in time for playout
CSci4211:
Multimedia Networking
18
Streaming Stored Multimedia:
What is it?
1. Video
pre-recorded
2. video
sent
network
delay
3. video received,
played out at client
time
streaming: at this time, client
playing out early part of video,
while server still sending later
part of video
CSci4211:
Multimedia Networking
19
Internet multimedia: Streaming Approach
• browser GETs metafile
• browser launches player, passing metafile
• player contacts server
• server streams audio/video to player
CSci4211:
Multimedia Networking
20
Streaming from a streaming server
• This architecture allows for non-HTTP protocol
between server and media player
• Can also use UDP instead of TCP.
CSci4211:
Multimedia Networking
21
Streaming Stored Multimedia
Application-level streaming
techniques for making the
best out of best effort
service:
– client side buffering
– use of UDP versus TCP
– multiple encodings of
multimedia
Media Player
•
•
•
•
jitter removal
decompression
error concealment
graphical user interface
w/ controls for
interactivity
22
Streaming Multimedia: UDP or TCP?
UDP
• server sends at rate appropriate for client (oblivious to
network congestion !)
– often send rate = encoding rate = constant rate
– then, fill rate = constant rate - packet loss
• short playout delay (2-5 seconds) to compensate for network
delay jitter
• error recover: time permitting
TCP
•
•
•
•
send at maximum possible rate under TCP
fill rate fluctuates due to TCP congestion control
larger playout delay: smooth TCP delivery rate
HTTP/TCP passes more easily through firewalls
23
Streaming Multimedia: Client Buffering
video sent at
Certain bit rates
variable
network
delay
constant bit
rate video
playout at client
buffered
video
client video
reception
client playout
delay
time
• Client-side buffering, playout delay compensate
for network-added delay, delay jitter
24
Streaming Multimedia: Client Buffering
“constant
”
drain
rate,
d
variable
fill
rate, x(t)
buffered
video
• Client-side buffering, playout delay compensate
for network-added delay, delay jitter
CSci4211:
Multimedia Networking
25
Streaming Stored Multimedia:
Interactivity
• VCR-like functionality: client can
pause, rewind, FF, push slider bar
– 10 sec initial delay OK
– 1-2 sec until command effect OK
• timing constraint for still-to-be transmitted data:
in time for playout
CSci4211:
Multimedia Networking
26
Streaming Live Multimedia
Examples:
• Internet radio talk show
• Live sporting event
Streaming
• playback buffer
• playback can lag tens of seconds after
transmission
• still have timing constraint
Interactivity
• fast forward impossible
• rewind, pause possible!
CSci4211:
Multimedia Networking
27
Interactive, Real-Time Multimedia
• applications: IP telephony,
video conference, distributed
interactive worlds
• end-end delay requirements:
– audio: < 150 msec good, < 400 msec OK
• includes application-level (packetization) and network
delays
• higher delays noticeable, impair interactivity
• session initialization
– how does callee advertise its IP address, port number,
encoding algorithms?
CSci4211:
Multimedia Networking
28
Large-scale Internet Video Delivery:
YouTube & Netflix Case Studies
Based on two active measurement studies we have
conducted
•Reverse-engineering YouTube Delivery Cloud
– Google’s New YouTube Architectural Design
•Unreeling Netflix Video Streaming Service
– Cloud-sourcing: Amazon Cloud Services & CDNs
CSci4211:
Multimedia Networking
29
YouTube Video Delivery Basics
Front end
web-servers
2. HTTP reply
containing html to
construct the web page 4. HTTP reply
and a link to stream the FLV stream
FLV file
Video-servers
(front end)
Internet
1. HTTP GET request
for video URL
3. HTTP GET
request
for FLV stream
User
CSci4211:
Multimedia Networking
30
www.youtube.com
CSci4211:
Multimedia Networking
31
Embedded Flash Video
CSci4211:
Multimedia Networking
32
Google’s New YouTube
Video Delivery Architecture
Three components
• Videos and video id space
• Physical cache hierarchy
•
 three tiers: primary,
secondary, & tertiary
 primary caches:
“Google locations” vs.
“ISP locations”
Implications:
Layered organization of
a) YouTube videos are not replicated at all locations!
namespaces
b) only replicated at (5) tertiary cache locations
 representing “logical”
c) Google
likely utilizes some form of location-aware
video servers
load-balancing
(among primary cache locations)
 five
“anycast” namespaces
 two “unicast” namespaces
CSci4211:
Multimedia Networking
33
YouTube Video Id Space
• Each YouTube video is assigned a unique id
e.g., http://www.youtube.com/watch?v=tObjCw_WgKs
• Each video id is 11 char string
• first 10 chars can be any alpha-numeric values [0-9, a-z,
A-Z] plus “-” and “_”
• last char can be one of the 16 chars {0, 4, 8, ..., A, E, ...}


Video id space size: 6411
Video id’s are randomly
distributed in the id
space
CSci4211:
Multimedia Networking
34
Physical Cache Hierarchy & Locations
~ 50 cache locations
• ~40 primary locations
• including ~10 non-
Google ISP locations
• 8 secondary locations
• 5 tertiary locations
Geo-locations using
• city codes in unicast hostnames,
e.g., r1.sjc01g01.c.youtube.com
• low latency from PLnodes (< 3ms)
• clustering of IP addresses using
latency matrix
P: primary
S: secondary
T: Tertiary
CSci4211:
Multimedia Networking
35
Layered Namespace Organization
Two types of namespaces
– Five “anycast” namespaces
• lscache: “visible” primary ns
• each ns representing fixed #
of “logical” servers
• logical servers mapped to
physical servers via DNS
– 2 “unicast” namespaces
• rhost: google locations
• rhostisp: ISP locations
• mapped to a single server
CSci4211:
Multimedia Networking
36
YouTube Video Delivery Dynamics:
Summary
• Locality-aware DNS resolution
• Handling load balancing & hotspots
– DNS change
– Dynamic HTTP redirection
– local vs. higher cache tier
• Handling cache misses
– Background fetch
– Dynamic HTTP redirection
CSci4211:
Multimedia Networking
37
What Makes Netflix Interesting?
• Commercial, feature-length movies and TV shows
 and not free; subscription-based
• Nonetheless, Netflix is huge!


25 million subscribers and ~20,000 titles (and growing)
consumes 30% of peak-time downstream bandwidth in North
America
• A prime example of cloud-sourced architecture
 Maintains only a small “in-house” facility for key functions




e.g., subscriber management (account creation, payment, …)

user authentication, video search, video storage, …

Akamai, Level 3 and Limelight
Majority of functions are sourced to Amazon cloud
(EC2/S3)
DNS service is sourced to UltraDNS
Leverage multiple CDNs (content-distribution networks) for
video delivery
CSci4211:
Multimedia Networking
Netflix Architecture
• Netflix has its own “data center” for certain crucial operations
(e.g., user registration, billing, …)
• Most web-based user-video interaction, computation/storage
operations are cloud-sourced to Amazon AWS
• Video delivery is out/cloud-sourced to 3 CDNs
• Users need to use MS Silverlight player for video streaming
CSci4211:
Multimedia Networking
Netflix Videos and Video Chunks
• Netflix uses a numeric ID to identify each movie
– IDs are variable length (6-8 digits): 213530, 1001192,
70221086
– video IDs do not seem to be evenly distributed in the
ID space
– these video IDs are not used in playback operations
• Each movie is encoded in multiple quality levels, each is
identified by a numeric ID (9 digits)
– various numeric IDs associated with the same movie appear to
have no obvious relations
CSci4211:
Multimedia Networking
40
Netflix Videos and Video Chunks
• Videos are divided in “chunks” (of roughly 4 secs),
specified using (byte) “range/xxx-xxx?” in the URL path:
Limelight:
http://netflix-094.vo.llnwd.net/s/stor3/384/534975384.ismv/range/057689?p=58&e=1311456547&h=2caca6fb4cc2c522e657006cf69d4ace
Akamai:
http://netflix094.as.nflximg.com.edgesuite.net/sa53/384/534975384.ismv/range/0
-57689?token=1311456547_411862e41a33dc93ee71e2e3b3fd8534
Level3:
http://nflx.i.ad483241.x.lcdn.nflximg.com/384/534975384.ismv/range/057689?etime=20110723212907&movieHash=094&encoded=06847414df0656e6
97cbd
• Netflix uses a version of (MPEG-)DASH for video
streaming
CSci4211:
Multimedia Networking
41
DASH: dynamic adaptive streaming over HTTP
• Not really a protocol; it provides formats to enable efficient
and high-quality delivery of streaming services over the
Internet
– Enable HTTP-CDNs; reuse of existing technology (codec, DRM,…)
– Move “intelligence” to client: device capability, bandwidth
adaptation, …
• In particular, it specifies Media Presentation Description (MPD)
42
CSci4211:
Multimedia Networking
Ack & ©: Thomas Stockhammer
DASH Data Model and Manifest Files
• DASH MPD:
Segment Info
Initialization Segment
http://www.e.com/ahs-5.3gp
Media Presentation
Period, start=0s
Media Segment 1
Period,
•start=100
•baseURL=http://www.e.com/
…
…
Period, start=100s
Representation 1
Representation 2
Period, start=295s
…
100kbit/s
•bandwidth=500kbit/s
•width 640, height 480
…
Segment Info
500kbit/s
…
Representation 1
…
duration=10s
Template:
./ahs-5-$Index$.3gs
start=0s
http://www.e.com/ahs-5-1.3gs
Media Segment 2
start=10s
http://www.e.com/ahs-5-2.3gs
Media Segment 3
start=20s
http://www.e.com/ahs-5-3.3gh
…
• Segment Indexing: MPD only; MPD+segment; segment only
Media Segment 20
start=190s
Segment Index in MPD only
seg1.mp4
<MPD>
seg2.mp4
...
<URL
...
sourceURL="seg1.mp4"/>
<URL
sourceURL="seg2.mp4"/>
<MPD>
</MPD>
seg.mp4
...
<URL sourceURL="seg.mp4" range="0-499"/>
<URL sourceURL="seg.mp4" range="500CSci4211:
Multimedia Networking
999"/>
Ack
</MPD>
http://www.e.com/ahs-5-20.3gs
43
& ©: Thomas Stockhammer
Netflix Manifest Files
• A manifest file contains metadata
• Netflix manifest files contain a lot of information
o
o
o
Available bitrates for audio, video and trickplay
MPD and URLs pointing to CDNs
CDNs and their "rankings"
<nccp:cdn>
<nccp:name>level3</nccp:name>
<nccp:cdnid>6</nccp:cdnid>
<nccp:rank>1</nccp:rank>
<nccp:weight>140</nccp:weight>
</nccp:cdn>
<nccp:cdn>
<nccp:name>limelight</nccp:name>
<nccp:cdnid>4</nccp:cdnid>
<nccp:rank>2</nccp:rank>
<nccp:weight>120</nccp:weight>
</nccp:cdn>
<nccp:cdn>
<nccp:name>akamai</nccp:name>
<nccp:cdnid>9</nccp:cdnid>
<nccp:rank>3</nccp:rank>
<nccp:weight>100</nccp:weight>
</nccp:cdn>
CSci4211:
Multimedia Networking
44
Netflix Manifest Files …
A section of the manifest containing the base URLs, pointing to
CDNs
<nccp:downloadurls>
<nccp:downloadurl>
<nccp:expiration>1311456547</nccp:expiration>
<nccp:cdnid>9</nccp:cdnid>
<nccp:url>http://netflix094.as.nflximg.com.edgesuite.net/sa73/531/943233531.ismv?token=131145
6547_e329d4271a7ff72019a550dec8ce3840</nccp:url>
</nccp:downloadurl>
<nccp:downloadurl>
<nccp:expiration>1311456547</nccp:expiration>
<nccp:cdnid>4</nccp:cdnid>
<nccp:url>http://netflix094.vo.llnwd.net/s/stor3/531/943233531.ismv?p=58&amp;e=1311456547&amp;h=8adaa52cd06db9
219790bbdb323fc6b8</nccp:url>
</nccp:downloadurl>
<nccp:downloadurl>
<nccp:expiration>1311456547</nccp:expiration>
<nccp:cdnid>6</nccp:cdnid>
<nccp:url>http://nflx.i.ad483241.x.lcdn.nflximg.com/531/943233531.ismv?etime=20110723212907
&amp;movieHash=094&amp;encoded=0473c433ff6dc2f7f2f4a</nccp:url>
CSci4211:
Multimedia Networking
45
</nccp:downloadurl>
</nccp:downloadurls>
Netflix: Adapting to Bandwidth Changes
• Two possible approaches


Increase/decrease quality level using DASH
Switch CDNs
• Experiments


Play a movie and systematically throttle available
bandwidth
Observe server addresses and video quality
• Bandwidth throttling using the “dummynet” tool


Throttling done on the client side by limiting how fast it
can download from any given CDN server
First throttle the most preferred CDN server, keep
throttling other servers as they get selected
CSci4211:
Multimedia Networking
46
Adapting to Bandwidth Changes
• Lower quality levels in response to lower bandwidth
• Switch CDN only when minimum quality level cannot be
supported
• Netflix seems to use multiple CDNs only for failover
purposes!
CSci4211:
Multimedia Networking
47
CDN Bandwidth Measurement
• Use both local residential hosts and PlanetLab nodes
 13 residential hosts and 100s PlanetLab nodes are used
 Each host downloads small chunks of Netflix videos from all
three CDN servers by replaying URLs from the manifest files
• Experiments are done for several hours every day for about
3 weeks
 total experiment duration was divided into 16 second intervals
 the clients downloaded chunks from CDN 1, 2 and 3 at the
beginning of seconds 0, 4 and 8.
 at the beginning of the 12th second, the clients tried to
download the chunks from all three CDNs simultaneously
• Measure bandwidth to 3 CDNs separately as well as the
combined bandwidth
• Perform analysis at three time-scales



average over the entire period
daily averages
instantaneous bandwidth
CSci4211:
Multimedia Networking
48
There is no Single Best CDN
CSci4211:
Multimedia Networking
49