The principle of parsimonious modelling of Internet traffic states that a minimal
number of descriptors should be used for its characterization. Until early 1990s,
the conventional Markovian models for voice traffic had been considered suitable
and parsimonious for data traffic as well. Later with the discovery of strong
correlations and increased burstiness in Internet traffic, various self-similar count
models have been proposed. But, in fact, such models are strictly mono-fractal
and applicable at coarse time scales, whereas Internet traffic modelling is about
modelling traffic at fine and coarse time scales; modelling traffic which can be
mono and multi-fractal; modelling traffic at interarrival time and count levels;
modelling traffic at access and core tiers; and modelling all the three structural
components of Internet traffic, that is, packets, flows and sessions.
The philosophy of this thesis can be described as: “the renewal of renewal theory
in Internet traffic modelling”. Renewal theory has a great potential in modelling
statistical characteristics of Internet traffic belonging to individual users, access
and core networks. In this thesis, we develop an Internet traffic modelling
framework based on fractal renewal processes, that is, renewal processes with
underlying distribution of interarrival times being heavy-tailed. The proposed
renewal framework covers packets, flows and sessions as structural components
of Internet traffic and is applicable for modelling the traffic at fine and coarse
time scales. The properties of superposition of renewal processes can be used
to model traffic in higher tiers of the Internet hierarchy. As the framework is
based on renewal processes, therefore, Internet traffic can be modelled at both
interarrival times and count levels.
| Identifer | oai:union.ndltd.org:canterbury.ac.nz/oai:ir.canterbury.ac.nz:10092/10194 |
| Date | January 2014 |
| Creators | Arfeen, Muhammad Asad |
| Publisher | University of Canterbury. Department of Computer Science & Software Engineering |
| Source Sets | University of Canterbury |
| Language | English |
| Detected Language | English |
| Type | Electronic thesis or dissertation, Text |
| Rights | Copyright Muhammad Asad Arfeen, http://library.canterbury.ac.nz/thesis/etheses_copyright.shtml |
| Relation | NZCU |
Page generated in 0.0019 seconds