25-01-2013, 12:46 PM
Characteristics of WAP Traffic
1Characteristics of WAP.PDF (Size: 1.47 MB / Downloads: 29)
ABSTRACT
This paper considers the characteristics of WAP traffic.
We start by constructing a model of WAP traffic based
on a number of user scenarios. One traffic
characteristic which is of particular interest in network
dimensioning is the degree of self-similarity, so the
paper looks at the characteristics of aggregated traffic
with WAP, web and packet speech components to
estimate its self-similarity. The results indicated that,
while WAP traffic alone does not exhibit a significant
degree of self-similarity, a combined load from various
traffic sources retains almost the same degree of selfsimilarity
as the most self-similar individual source.
INTRODUCTION
Wireless Application Protocol (WAP) [1] is a protocol
for the transmission of web like content to mobile
phones and other (usually wireless) devices with lowbandwidth
links. Although commercial take-up has
been slower than originally hoped, WAP is still offers
considerable promise as more content tailored to small
devices becomes available.
The standard WAP configuration is to have a WAP
terminal communicating with a WAP gateway, which is
then connected to the web. The terminal sends its
request to the gateway, which then requests the content
from the web server. The web server replies to the
request with WML (Wireless Markup Language) files
which the gateway can then optimise for transmission
over the low bandwidth link to the terminal. A
modified approach which allows the browsing of
standard web pages is to use an HTML filter to convert
the HTML pages to WML, but on devices with small
displays this is unlikely to be successful, with content
having to be tailored to the device. Proposals for WAP
2.0 allow selection of information depending on the
capabilities of the device, but this paper deals with the
current WAP specification (1.2).
SCENARIOS
In order to investigate the traffic generated in a WAP
session, a number of scenarios were developed to
represent typical data transactions. When this was
done, two distinct scenario types developed
distinguished by the reading time or lack of it: look-up
or browse. In a look-up scenario, the user is attempted
to find a specific fact. This normally involves the
navigation of a number of pages, but the reading time of
these intermediate pages is very small, especially if the
user is familiar with the site. However, the nature of the
user interface on a phone means that there is a
significant navigation time. Examples include finding
the weather in a specific location, or a stock price.
Several facts can be found on a look-up session but
there will only be one meaningful reading time per
transaction. In a browse session, the information is
given to the user as the site defines it rather than the
user explicitly requesting each fact. An example would
be reading the news, or browsing a shopping site
(although a focussed visit to a shopping site which the
user was buying a specific product from would
probably be a ‘lookup’). In the browse case, WAP
traffic does exhibit reading times, although these are
smaller than for web scenarios. The distinction is not
clear cut – visiting a travel site costing a number of
different travel options could either be considered to be
a ‘browse’ or an number of different ‘look-up’s.
WAP MODEL
Look-up scenario: A number of different scenarios were
run based on users accessing stock price information
and weather forecasts.
No real distinction could be seen between the
loading time between objects and reading time for WAP
traffic. File interarrival time was modelled reasonably
well by an exponential distribution with a mean of
about 6 seconds (5.7 sec to 6.3 sec depending on the
scenario).
AGGREGATE TRAFFIC SIMULATOR
In order to evaluate the characteristic of WAP traffic,
first of all, a traffic model was built in the network
simulator ns to generate simulated WAP traffic. The
resultant traces were then studied through aggregation
over different time scales and the degree of selfsimilarity
and long range dependence was estimated.
A simple fixed network consisting of two wired
nodes and two base station nodes was created. The
nodes were inter-connected as shown in the network
topology diagram below in Figure 4.
CONCLUSIONS
Our results show WAP traffic varies significantly from
web traffic, particularly due to the way it is accessed.
This means that modifications to web based models are
unlike to model WAP traffic satisfactory. Of particular
note is that even though the page sizes are significantly
smaller, interarrival time is increased due to the reduced
user interface increasing reading time.
Note that the nature of the experiments means that
the results will be affected by the user interface, and so
would not be representative of WAP traffic on a device
with a different interface such as a pen based one.
Reading and scrolling times are also likely to be
dependent on the screen size, although the one used
here was fairly typical of a WAP-enabled GSM phone.
WAP traffic itself is self-similar, but not
significantly so. When aggregated with web traffic, it
does not greatly change the high degree of selfsimilarity
present in that traffic. A summary of the
simulation runs being conducted, as well as the
corresponding traffic characteristics in terms of the selfsimilarity
indicator, the Hurst parameter, is presented
below in Table 4.