[Olivier Crepin-Leblond]

Crepin-Leblond, Olivier M.J., Digital Communications Section, Electrical and Electronic Engineering Department, Imperial College of Science, Technology & Medicine, London, UK.
Supervisor: Prof. Laurie F. Turner (now retired)
Publication Date: April 1997
Page(s): 243 (inc. appendices)
References Cited: 71
Location: Imperial College EEE Library.
Copy: Senate House Library, University of London.


The underlying technology of B-ISDN, Asynchronous Transfer Mode (ATM) makes use of end to end packet switching technology, where delays are introduced at multiplexers during high traffic load conditions.  This end to end delay, which is variable and is directly related to the congestion in the packet switched network, causes errors in the transmission, which affect the quality of service (QoS).

In the thesis, a new approach is put forward for reducing the end to end delay of voice and video cells by dividing traffic into two main traffic classes and giving priority to voice and video traffic over data traffic.

A review of the properties and characteristics of ATM is provided.  Methods for flow control and reduction of delay in ATM traffic are reviewed.  It is concluded that Markov Modulated Poisson Processes (MMPPs), provide a more accurate characterisation of ATM traffic than other methods.

A simulation method based on Discrete Time Event simulation and using the "event float" approach is presented for simulating the behaviour of an ATM multiplexer.  After being tested as a single queue system with a Poisson arrival process, modulated by burst and silent periods, the tool is extended to incorporate two inputs and one output.

Markov Modulated Poisson Processes (MMPPs) are then used in the traffic characterisation of ATM traffic and to analyse the unprioritised ATM system.  Analytical results are compared with results obtained using the stimulator tool

The simulator tool is then used to test two novel priority multiplexing algorithms.  The use of the algorithms leads to a reduction in delay, and in delay jitter for voice and video traffic.  The "Maximum data Wait" algorithm was effective at all traffic loads.  The "Maximum voice/video Wait" algorithm had effects which were highly dependent on the input traffic mix.  Both algorithms were compact enough for a use in a real-life multiplexer.