Experimental measurement campaigns
have demonstrated that the classical Poissonian
Model does not represent adequately the real characteristics
of data traffic in local, metropolitan and geographical
networks. In particular, in experimental results,
self-similarity and long-range dependence have
been observed on various protocol stack layers
and in different network contexts. In order to
model this experimental evidence, mathematical
traffic models based on fractal, multi-fractal
and and generically fractionary noise have been
It is impressive the impact that
these novel findings have on network engineering
and on the classical paradigms for network design
and performance evaluation. It is worth noting
the scientific and engineering importance of these
models that, by modelling the real traffic features
in today multi-service networks, allow to estimate
more accurately the system behavior and consequently
achieve more efficient network dimensioning.
In short, this project aims at significantly contributing
to new scientific achievement in this research
field, achieving, by one side, a deeper knowledge
of multi-service data traffic characteristics,
and, on the other side, providing the research
novel analytical instruments and experimental
data, available in a public domain archive. Secondarily,
we will evaluate how to exploit these traffic
measurements as central element of new and improved
network control and management procedures.
In the following we give some further detail about
1) Collection and analysis of experimental data
in multi-service packet networks, on backbone
links (STM-16 and Gigabit Ethernet) and in WiFi
radio access network, both in laboratory and public
structures and making a differentiation by protocols
(TCP, IP and MAC) and services (www, peer-to-peer
and real-time streaming video and voice).
2) Realization of a public domain archive containing
experimental data (www server), to make results
available for a long time and shared also among
3) Study and development of novel mathematical
tools for traffic analysis and parametric synthesis
of pseudo-random traces.
4) Development of traffic measurement hardware
and software tools, in the network (TCP/IP traffic),
access (MAC) and physical layers.
5) Development of modules for data traffic collection
in order to provide automated access control and
resource allocation functions in an open-source
Experimental measurements will be carried on in
collaboration with GARR, the consortium for the
Italian University and Scientific Research Network,
and FASTWEB, a private operator offering advanced
data, video and voice services on IP network.
POLIMI tasks will be mainly the coordination of
project activities, the traffic measurements on
backbone links, the study of novel analytical
methods, the identification of models suitable
to experimental results.
POLITO will be in charge to characterize TCP and
real-time multimedia (VoIP, MPEG) traffic, both
in wired and wireless WiFi networks. POLITO will
also realize and maintain the project website,
where experimental results and new tools will
be made available.
UNIPI will evaluate the application of traffic
results to advanced network control and management
procedures, developing data collection modules
for automatic access control and resource allocation
on a open-source platform.
UNIPA will take in charge the experimental measurements
in 802.11 mobile wireless networks, stating from
data acquisition to data analysis, investigating
problem of unfairness among stations and interaction
problems among protocols in high layers and in
MAC and physical layers.