This preliminary investigation
of Logitech QuickCam Web Camera network traffic aims to report upon the inter-packet arrival
times and packet size distributions of one and two way data
rates. Traffic characteristics of web cameras are interesting to compare to
previously collected data on Half-Life
and Quake3 network games in identifying
how such applications utilise network resources. The results of this
investigation will form the basis for further research into the effect of
jitter and latency on video interactions over the web.
Generating and Capturing Web Camera Traffic
Web camera traffic was
captured in a LAN environment with one camera connected to each
computer separated by a bridge computer running FreeBSD in order to collect the traffic
flow. This paper looks at the traffic generated from Web Camera 1 on Computer
1 transmitting to Web Camera 2 on Computer 2 and both computers transmitting simultaneously.
Each test was conducted over the period of one hour, although longer tests could be done with largely similar results.
Web Camera Computer Configurations
||Pentium(R) 4 2.66GHz
||Pentium(R) 4 2.00GHz|
||Microsoft Windows XP SP1
||Microsoft Windows XP SP1|
Equipment used for generating/capturing Web Camera traffic
Figure 1.1: Web Camera Equipment Setup
Computer 1 was connected via crossover cable directly
to a bridge computer running FreeBSD. A 5-Port 10/100Mbps Nway Switch used to provide
access to the Internet via the Swinburne University ITS network.
Each transmitting web camera resulted in two flows of traffic, the data packets and the ACK replies from the computer receiving the transmission. When simulteneous transmission from Web Camera 1 and Web Camera 2 occured, four flows of traffic were identified. In this report we refer to the data (and corresponding ACK replies) from Web Camera 1 to Computer 2 as Connection A and the data (and corresponding ACK replies) from Web Camera 2 to Computer 1 as Connection B. Figures 1.2 and 1.3 show a simlified network topology for Connection A and Connection B.
Figure 1.2: Connection A
Figure 1.3: Connection B
Bridge/Packet Sniffing Computer (136.186.299.64)
Web Camera 1 and 2
CPU: Pentium 3 800MHz
OS: FreeBSD 4.9
Model: QuickCam Express
Application: MSN Messenger Version 6.1
Tcpdump was running on the Bridge/Packet Sniffing Computer and
was used to generate packet traffic histograms. Pkthisto recorded a traffic flow once it saw 50 packets
and packets were required to arrive more frequently than 1000ms for the flow to be considered active. Each inter-packet
arrival time and packet length histogram produced by Pkthisto represents a maximum of 500 packets.
The data set used for analysis was collected over a one day period. 1 hour sessions were observed, where one or two
web cameras were used to transmit video data using MSN Messenger 6.1.
Single Idle Web Camera Traffic Results
- Connection A time: 1hr 24sec
Single Active Web Camera Traffic Results
- Connection A time: 1hr 10sec
Two Idle Web Camera Traffic Results
- Connection A time: 1hr 22sec, Connection B time: 1hr 11sec
Two Active Web Camera Traffic Results
- Connection A time: 1hr 39sec, Connection B time: 1hr 41sec
Summary and Further Analysis
The table below is a summary of the inter-packet arrival times and packet size distributions observed
from the web camera network traffic.
|Idle Web Camera Traffic
||Active Web Camera Traffic|
|To Computer with Web Camera Transmitting
||From Computer with Web Camera Transmitting
||To Computer with Web Camera Transmitting
||From Computer With Web Camera Transmitting|
|Inter-packet Arrival Times
3-9% at <0.25ms
67-79% at 201ms
|10-18% at <0.25ms
46-57% at 201ms
11-14% at 201.25ms
|15-27% at <0.25ms
60-65% between 35 and 100ms
|49-54% at <0.25ms|
4-11% at 2ms
27-35% between 35 and 100ms
89-<100% at 46 bytes
>0-11% at 52 bytes
|3-5% at 64 bytes
11-20% at 1500 bytes
Other significant peaks at 856, 584, 1128 and between 924 and 1024 bytes
92-<100% at 46 bytes
>0-8% at 52 bytes
7-12% at 64 bytes
49-53% at 1500 bytes
As we can see form the table above, transmissions where there is no movement in front of the web camera
result in inter-packet arrival times of 201ms between 46% or more packets. We can also see that a computer receiving
a transmission sends mostly 46 byte ACK messages back to the computer sending the web camera transmission. The most
common packet size sent by a computer transmitting video data is 1500 bytes.
When there is activity in front of the camera, there is a large percentage of packets with a 1usec to 249usec inter-packet
arrival time (and 2ms inter-packet arrival time from the computer transmitting). There is also a significant percentage
of packet sizes with inter-packet arrival times between 35 to 100ms. Again, the computer receiving a web
camera transmission replies with mostly 46 byte ACK messages and the most common packet size sent by a computer transmitting
data is 1500 bytes.
From the above statistical results, we can assume that when there is no movement in front of a web camera, packets
are sent between the transmitting and receiving computers every 201ms. The receiving computer sends 46 byte ACK packets,
while the transmitting computer sends a variety of packet sizes. When there is movement in front of the web camera a large percentage of packets arrive with an inter-arrival time of less than 0.25ms and between 35 and 100ms. Again, the computer receiving
the web camera transmission generally only sends 46 byte ACK packets, while the transmitting computer sends a larger variety
of packet sizes, most commonly 1500 byte packets.
Further research will include modelling web camera traffic in order to make statistical comparisons to models created
by game traffic. It will aid in the development of software that uses packet statistics to identify and differentiate between
applications producing traffic across a network.
Grenville Armitage, Director of Centre for Advanced Internet Architectures.