Transcript Research Rugby

Streaming Video over a
Wireless Network
So what is the problem!!
WPI CS Research Rugby
Bob Kinicki
November 30, 2004
CC-Wireless Rugby Team
prop
Feng Li
wireless measurement
hooker
Mingzhe Li
wireless bandwidth estimation
flanker
Jae Chung
congestion control
flanker
Yubing Wang
retransmissions
scrumhalf
Robert Kinicki
protocols
flyhalf
Huahui Wu
FEC
weak side wing
Emmanuel Agu
low-level wireless
fullback
Mark Claypool
multimedia
Research Rugby November 30, 2004
2
Outline



Motivation – the Future
Streaming Video over Wireless What Problems?
Background
–
–
–
–
–


Layered network architecture
Networking ‘religious’ tenets
Measurement and Tools
Streaming multimedia primer
Introduction to wireless networks
Focus: Wireless Bandwidth Estimation
Discussion ??
Research Rugby November 30, 2004
3
Streaming Video
Video Frames
Network Cloud
Server
Research Rugby November 30, 2004
Client
4
Streaming Video over Wireless
Video Frames
Wired network
AP
Client
Server
Research Rugby November 30, 2004
5
Networking Protocol Layers
Application Layer: Real Player,Windows
Media Player, MPEG
Transport Layer:
Network Layer:
TCP, UDP
IP
Data Link Layer (MAC): 802.3, 802.11b,
802.11g
Research Rugby November 30, 2004
6
Measurement and Tools
Wireless Streaming Multimedia Lab Topology





WM Tracker
Real Tracker
Ethereal
Ethereal
Monitor Mode
Wireless sniffer
(seine)
Hub
7x
8x
9x
1x
2x
3x
10x
11x
12x
7x
8x
9x
4x
5x
6x
1x
2x
3x
10x
11x
12x
4x
5x
6x
7x
8x
9x
1x
2x
3x
10x
11x
12x
7x
8x
9x
4x
5x
6x
1x
2x
3x
10x
11x
12x
4x
5x
6x
C
7 8 9 101112
A
Ethernet
Access point
Ethernet
Mobile station
(yangtze)
12 34 56
A
B
Switch
NFS Server
(york)
WPI Campus Network
C
7 8 9 101112
A
12 34 56
A
B
Video Camera
Video capture
Emulated Internet
VCR

IDC



WM Tracker
Real Tracker
Ethereal
Media Client
(???)
Packet Sniffer
(merlot)
NAT Router
(streamit)





DVD
DVD Player
Windows media server
Real streaming Server
Quick time Darwin
Windows IIS server
Ethereal
Windows 2003 server
(saco)
Slide courtesy of Mingzhe’s PEDS presentation
Monday October 25,2005
Research Rugby November 30, 2004
8
Wireless Measurement Study
Research Rugby November 30, 2004
9
Preliminary Wireless Measurements
Application Layer
Network Layer
Data Link Layer
Research Rugby November 30, 2004
10
Video Streaming Choices



Media Player or Real Player
TCP or UDP
Single layer encoding or multilayer
encoding
Research Rugby November 30, 2004
11
MediaPlayer vs.
™
RealPlayer A Comparison
™
of Network Turbulence
Real vs WMP Playout Rate
RealPlayer
buffers at
a higher
rate
Bandwidth (Kbits/sec.)
700
Real Player (284K)
600
Real Player (36K)
500
Windows Media Player (323K)
400
Windows Media Player (49K)
300
200
100
0
0
50
100
150
200
250
Time (seconds)
Research Rugby November 30, 2004
13
Video Compression Standard
I0

B00 B01
P1
B10
P2
I0
MPEG
– Popular compression standard
– Intra-compression and inter-compression
– Three types of frames: I, P and B
– Group Of Pictures (GOP)
Research Rugby November 30, 2004
14
340 Kbps Clip - Bottleneck Capacity 725 Kbps
TCP Friendly?
Research Rugby November 30, 2004
< 0.001 packet loss
After 15 seconds
15
548 Kbps Clip - Bottleneck Capacity 725 Kbps
Not TCP- Friendly!
Research Rugby November 30, 2004
~ 0.003 packet loss for WSM
~ 0.006 packet loss for TCP
after 15 seconds
16
1128 Kbps Clip - Bottleneck Capacity 725 Kbps
Responsive to Capacity?
Research Rugby November 30, 2004
17
Wireless Networks
Wireless Issues
Higher error rates - Bursty?
 Hidden terminal problem
 Variable bandwidth
 Mobile versus stationary
 MAC layer saturation

Research Rugby November 30, 2004
19
Hidden Terminal Problem
1

2
3
802.11 Solution to the Hidden Terminal Problem
– Use a four-way handshake: RTS-CTS-DATA-ACK where the
RTS and CTS packets are significantly smaller than the
average data packet.
– The maximum number of RTS retransmissions is set to 7.

However, the 802.11 protocol will still have problems
if the MAC layer becomes saturated!!
Research Rugby November 30, 2004
20
Rate Constrained TFRC



A seven-hop chain network was simulated.
The TFRC sending rate is manually constrained.
The MAC layer saturates at 300Kbps.
Research Rugby November 30, 2004
21
Wireless Bandwidth
Estimation
Performance Definitions



The capacity, Ci , of hop i is the
maximum possible IP layer transfer at
that hop. Namely, the bit rate for
transferring MTU-sized IP packets.
The available bandwidth, Ai , of a link
relates to the unused capacity of a link
during a time period.
The Bulk-Transfer-Capacity (BTC) is
the maximum throughput obtainable by
a single TCP connection.
Research Rugby November 30, 2004
23
Bandwidth Estimation Techniques

Variable Packet Size (VPS) probing: used to measure
minimum RTT’s to all hops on path.
• pathchar, pchar, ...




Packet pair probing: each packet pair consists of two
packets of the same size sent back-to-back. The goal
is to measure the dispersion of a packet pair.
Packet train probing: extends packet-pair probing by
using multiple back-to-back packets.
Self-Loading Periodic Streams (SLoPS): send a train
of packets (k=100) equal-sized packets at a rate R.
Sender continues to vary R to get it close to A.
Trains of Packet Pairs (TOPP): send many packet
pairs at gradually increasing rates from source to
sink.
Research Rugby November 30, 2004
24
Packet Dispersion
Research Rugby November 30, 2004
25
Bandwidth Estimation Techniques


pathchirp: uses an exponentially spaced
chirp probing train and launches m
packet chirps where each chirp is made
up of N UDP probe packets each of size
P bytes.
pathchirp discards all chirps with
dropped packets!
Research Rugby November 30, 2004
26
Bandwidth Estimation and Video

Currently, bandwidth estimation used in
Windows Media Player
– Capacity estimation before streaming
starts
– Uses packet pair

Mingzhe’s Research : new approach to
bandwidth estimation to improve
performance of video over wireless
networks.
Research Rugby November 30, 2004
27
Discussion ?
Research Rugby November 30, 2004
28