combination with stochastical traffic generators for the regarded
applications and models for the radio channel was
chosen as the methodology in [9, 6, 8].
Evaluating the traffic performance accurately by measurement
in the existing GPRS network is not possible,
since a scenario with a well-defined traffic load is hard to
set-up, the evaluation of the performance based on measurements
is very difficult, and the analysis of the behavior
of different protocol options is not possible in an existing
radio network. Therefore, the measurement results in this
paper are not suitable for accurate GPRS dimensioning and
can only serve as an overview of the GPRS deployment status.
Additionally the results are used to validate the simulation/emulation
tool GPRSIM [6] that was the basis for the
dimensioning rules presented in [9, 8].
After the introduction of the measurement setup in Section
II several measurement series are presented. In Section
III measurement results for the round-trip time that is
one limiting factor for both transaction oriented and TCPbased
applications are presented. After a measurement series
for FTP applications in Section IV, Section V and Section
VI comprise measurement results for WWW applications
typical for laptop users and WWW and e-mail applications
with small files typical for PDA users, respectively