Transport of RealTime Traffic over the Internet

1
Transport of Real-Time Trafficover the Internet

• Bernd Girod
• Information Systems Laboratory
• Stanford University

2
THE MEANING OF FREE SPEECH The acquisition by
eBay of Skype is a helpful reminder to the
world's trillion-dollar telecoms industry that
all phone calls will eventually be free . . . .
. . Ultimatelyperhaps by 2010voice may become a
free internet application, with operators making
money from related internet applications like
IPTV . . .
Economist, September 2005
3
IPTV Rollout
Verizon 10M householdsby 2009
IPTV SBC18M households by 2007
IEEE Spectrum, Jan. 2005
4
Why Is Real-Time Transport Hard?
Internet is a best-effort network . . .
Congestion Insufficient rate to
communicate Packet loss Impairs perceptual
quality Delay Impairs interactivity of
services Telephony one way delay lt 150 ms
ITU-T Rec. G.114 Delay
jitter Obstructs continuous media playout
5
Outline of the Talk

• QoS vs. best effort
• Resource allocation for IPTV
• Rate-distortion optimized streaming
• Multi-path routing
• P2P multicasting of live video streams

6
How 1B Users Share the Internet
Rate r
TCP Throughput
maximum transfer unit
Growing congestion
data rate
packet loss rate
p
round trip time
0.001
0.0001
0.1
0.01

• Mahdavi, Floyd, 1997
• Floyd, Handley, Padhye, Widmer, 2000

7
QoS vs. Best Effort

• Reservation-ism
• Voice and video need guaranteed QoS (bandwidth,
  loss, delay)
• Implement admission control Busy tone when
  network is full
• Best effort is fine for data applications

• Best Effort-ism
• Best Effort good enough for all applications
• Real-time applications can be made adaptive to
  cope with any level of service
• Overprovisioning always solves the problem, and
  its cheaper than QoS guarantees

8
Simple Model of A Shared Link

• Link of capacity C is shared among k flows
• Fair sharing each flow uses data rate C/k
• Homogeneous flows with same utility function u(.)
• Total utility

C
Breslau, Shenker, 1998
9
Rigid Applications
u

• Utility u0 below of minimum bit-rate B
• Maximum total utility Uk is achieved by
  admitting at most k flows

1
C/k
B
Breslau, Shenker, 1998
10
Rigid Applications (cont.)

• Expected loss in total utility w/o admission
  control
• Gap DU is substantial when number of admissable
  flows k is small
• Gap DU usually disappears with growing capacity
  C? Overprovisioning can solve the problem!

Breslau, Shenker, 1998
11
Elastic Applications

• Elastic applications utility function u(k), such
  that total utility U(k)ku(C/k) increases with k
• Example u(C/k)1-aC/k
• All flows should be admitted best effort!

u
C/k
12
Video Compression

• H.264 video coding for 2 different testsequences

• Video is elastic application
• Rate must be adapted to network throughput
• How to achieve rate control for stored content or
  multicasting?
• Utility function depends on content should use
  unequal rate allocation

Good picture quality
Foreman Mobile
Bad picture quality
13
Different Utility Functions

• Example uk(rk)1-akrk
• With rkgt0 ? Karush-Kuhn-Tucker conditions
  (reverse water-filling)
• Better than utility-oblivious fair sharing

Equal-slope Pareto condition
uk
Vilfredo Pareto 1848-1923
rk
14
Distribution of IPTV over WLAN
5 Mbps
11 Mbps
2 Mbps
Home Media Gateway
courtesy van Beek, 2004
15
Video Streaming Over Shared Channel
Transcoder
Decoder
0
0
Transcoder
Decoder
1
1
Receiver
Transcoder
Decoder
(Multi-Channel)
2
2
Transcoder
Decoder
3
3
Controller
Kalman, van Beek, Girod 2005
16
Tx Backlog for 4 Video Streams 85 WLAN
Utilization
Kalman, van Beek, Girod 2005
17
Streaming of Stored Content
Media files are already compressed How can we
nevertheless adapt to network?
100s to 1000ssimultaneous streams
DSL
Cable
wireless
Server
Network
Client
18
Not All Packets are Equally Important
I
I
P
P
I
B
B
B
P
P
P
I
B
B
B
P



A
A

19
Not All Packets are Equally Important
I
I
P
B
P
P
I
B
B
P
P
I
B
B
B
P



A
A

20
Distortion-Aware Packet Dropping
Good Picture quality
Distortionaware
Packet dropping No retransmissions QCIF
Carphone I-P-P-P-P-P- . . .
Bad picture quality
Oblivious
Percentage of Packets Retained
Chakareski, Girod, ICME 2004
21
Rate-DistortionOptimized (RaDiO) Streaming

• Decide which packets to send (and when) to
  maximize picture quality while not exceeding an
  average rate 2001

Server
Client
Network
22
A Brief History of Media Streaming

• Media streaming w/o congestion avoidance
  reckless driving 1995
• TCP-friendly rate control Limit average rate
  for fair sharing with TCP 1997
• Rate-distortion optimized packet scheduling
  (RaDiO) Decide which packets to send (and
  when) to maximize picture quality while not
  exceeding an average rate 2001
• Congestion-distortion-optimized
  scheduling/routing (CoDiO) Decide which
  packets to send (and when) to maximize picture
  quality while minimizing network congestion.
  2004

23
Congestion vs. Rate

• Congestion queuing delay that packets experience
  
• weighted by size of the packet
• averaged over all packets in the network
• Congestion increases nonlinearly with link
  bit-rate

Rmax
24
Video Distortion with Self Congestion
Good Picture quality
Bad picture quality
Bit-Rate kbps
25
Streaming with Last Hop Bottleneck
Random cross traffic
High bandwidth links
Video traffic
Low bandwidth last hop
Acknowledgments
26
Delay distribution

• Overall delay distribution
• Queue length determines delay of last hop

pdf
delay
C
27
Comparison RaDiO vs. CoDiO
50
PSNR dB
PSNR dB
Rate kbps
End-to-end delay ms
Simulations using H.263 Rate 10 fps Sequence
Foreman (32kbps,32kbps) Sequence length
60s Playout deadline 600ms
28
How To Avoid Traffic Jams?

• Avoid congested times . . .
• ?Congestion-aware packet scheduling
• Avoid congested roads . . .
• ?Congestion-aware routing

29
Multipath Routing for Minimum Congestion

• Mesh network, fully connected
• Streaming 100 kbps from Node 1 to Node 5
• Random cross traffic

30
Multipath Video Streaming
Good Picture quality
6 dB
Sequence Foreman QCIF, 250
frames, 30 fps Codec H.26L TML 8.5 Playout
deadline 500 ms Packetization 1
frame/packet Traffic model CBR No. of
realizations 400
Bad picture quality
Bit-Rate kbps
31
Multipath Video Streaming
3 paths 187 kbps, PSNR 36.2 dB
1 path 80 kbps, PSNR 32.5 dB
32
Distribution of Live Streams via
Pseudo-Multicast

• Example
• AOL webcast of Live 8 concert
• July 2, 2005

Content delivery network
Media server
. . .
Splitter servers
1500 servers in 90 locations
50 Gbps
. . .
. . .
. . .
. . .
. . .
175,000 simultaneous viewers
8M unique viewers
33
Distribution of Live Streams via
Pseudo-Multicast

• Example
• AOL webcast of Live 8 concert
• July 2, 2005

300 kbps
175,000 simultaneous viewers
8M unique viewers
34
P2P Multicast over 1 Tree
35
P2P Multicast over 2 Trees
36
P2P Ungraceful Parent Leave
Parent leave is detected
Retransmissions requested
New parent is selected
Parent of yellow tree is down
Yellow tree is recovered
3 trees
Hello, Yellow Tree Parent?
37
Experimental Set-up

• Network/protocol simulation in ns-2
• 1000 nodes
• 300 active peers
• Random peer arrival/departure ON (5 min)/OFF
  (30 s)
• Over-provisioned backbone
• Typical access bandwidth distribution
• Delay 5 ms/link congestion
• Video streaming
• Compression H.264 at 220 kbps
• 15 minute live multicast

Setton, Noh, Girod, ACM MM 2005
38
Join and Rejoin Latencies
Setton, Noh, Girod, ACM MM 2005
39
Congestion-DistortionOptimized P2P Live Streaming
Without CoDiO
peersconnected to 4/4 trees
With CoDiO
peersconnected to 4/4 trees
Setton, Noh, Girod, ACM MM 2005
40
P2P Video Multicast 64 out of 300 Peers
Congestion-distortion optimized (CoDiO) streaming
Without CoDiO
H.264 _at_ 220 kbps2 sec latency for all streams
41
Concluding Remarks

• Over-provisioning makes QoS superfluous
• Elastic applications dont need QoS
• Joint rate control for access bottlenecks (e.g.
  IPTV, WLAN)
• Media-aware congestion control (e.g. CoDiO)
• Multipath routing to mitigate congestion
• P2P viable alternative for content delivery
  networks
• Client-server ? edge-based ? P2P

42
The End

• http//www.stanford.edu/bgirod/publications.html