Title page for ETD etd-0429102-123525

Document Type thesis

Author Name Staake, Thorsten R

URN etd-0429102-123525

Title IP Traffic Statistics - A Markovian Approach

Degree MS

Department Electrical & Computer Engineering

Advisors
Matthew C. Bromberg, Advisor
Donald R. Brown, Committee Member
Mark L. Claypool, Committee Member

Keywords
fit parallel Erlang-K distributions to time series
performance of channel assignment procedure

Date of Presentation/Defense 2002-04-24

Availability unrestricted

Abstract
Data originating from non-voice sources is expected to play an increasingly
important role in the next generation mobile communication services. To plan
these networks, a detailed understanding of their traffic load is essential.
Recent experimental studies have shown that network traffic originating from
data applications can be self-similar, leading to a different queueing behavior than
predicted by conventional traffic models. Heavy tailed probability distributions are
appropriate for capturing this property, but including those random processes in
a performance analysis makes it difficult and often impossible to find numerical
results.
In this thesis three related topics are addressed: It is shown that Markovian
models with a large state space can be used to describe traffic which is self-similar
over a large time scale, a Maximum Likelihood approach to fit parallel Erlang-k
distributions directly to time series is developed, and the performance of a channel
assignment procedure in a wireless communication network is evaluated using the
above mentioned techniques to set up a Markovian model. Outcomes of the performance
analysis are blocking probabilities and latency due to restrictions of the
channel assignment procedure as well as estimations of the overall bandwidth that
the system is required to offer in order to support a given number of users.

Files
staake.pdf

Browse by Author | Browse by Department | Search all available ETDs

Questions? Email etd-questions@wpi.edu
Maintained by webmaster@wpi.edu

Document Type	thesis
Author Name	Staake, Thorsten R
URN	etd-0429102-123525
Title	IP Traffic Statistics - A Markovian Approach
Degree	MS
Department	Electrical & Computer Engineering
Advisors	Matthew C. Bromberg, Advisor Donald R. Brown, Committee Member Mark L. Claypool, Committee Member
Keywords	fit parallel Erlang-K distributions to time series performance of channel assignment procedure
Date of Presentation/Defense	2002-04-24
Availability	unrestricted
Abstract Data originating from non-voice sources is expected to play an increasingly important role in the next generation mobile communication services. To plan these networks, a detailed understanding of their traffic load is essential. Recent experimental studies have shown that network traffic originating from data applications can be self-similar, leading to a different queueing behavior than predicted by conventional traffic models. Heavy tailed probability distributions are appropriate for capturing this property, but including those random processes in a performance analysis makes it difficult and often impossible to find numerical results. In this thesis three related topics are addressed: It is shown that Markovian models with a large state space can be used to describe traffic which is self-similar over a large time scale, a Maximum Likelihood approach to fit parallel Erlang-k distributions directly to time series is developed, and the performance of a channel assignment procedure in a wireless communication network is evaluated using the above mentioned techniques to set up a Markovian model. Outcomes of the performance analysis are blocking probabilities and latency due to restrictions of the channel assignment procedure as well as estimations of the overall bandwidth that the system is required to offer in order to support a given number of users.
Files	staake.pdf