TY  - UNPB
TI  - Maximising Achievable Throughput in Optical Network Design
AV  - public
Y1  - 2025/01/28/
EP  - 188
N1  - Copyright © The Author 2025. Original content in this thesis is licensed under the terms of the Creative Commons Attribution 4.0 International (CC BY 4.0) Licence (https://creativecommons.org/licenses/by/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms.
N2  - This thesis is an investigation into maximising throughput for optical core network
design. To calculate the maximum achievable throughput that an optical network can
sustain, an NP-hard routing optimisation problem needs to be solved and therefore
designing the network to maximise this property is computationally difficult.
Both structural and physical properties impact the maximum achievable throughput
of optical networks. Therefore, the SNR-BA generative graph model is proposed in
Chapter 3, to investigate how structural and physical properties affect the maximum
achievable throughput of optical networks. The results showed that the networks with
better connectivity had on average 40% lower wavelength requirements and allocated
between 8-11% more lightpaths than the SNR-BA model. However, with path lengths
between 95 and 215% longer than the SNR-BA networks, they achieved 30-32% less
maximum achievable throughput. This demonstrated why including physical properties
within the design of optical networks is important.
More computationally efficient methods for calculating maximum achievable
throughput were needed for it to be included in physical topology design. Chapter 4
explores several strategies to reduce this computational complexity, including linear
programming, geometric deep learning and graph theoretical metric correlation.
Results in Chapter 4 show that the proposed graph theoretical metric, demand
weighted cost, had a high inverse-linear correlation to maximum achievable
throughput and thus was chosen to be embedded within the optimisation problem as a
proxy for maximum achievable throughput.
Chapter 5 investigates whether a proxy such as demand weighted cost can maximise
the maximum achievable throughput of optical networks. Compared to a control-set the
demand weighted cost showed to increase maximum achievable throughput of networks
by up to 63% compared to the control-set. However, the lowest demand weighted cost
did not always lead to the highest maximum achievable throughput. This showed that
the objective pushes networks in the right direction, however cannot directly optimise
maximum achievable throughput. To achieve this, limiting cut theory was employed,
achieving a 106% increase in maximum achievable throughput compared to the control-
set and thus directly optimising maximum achievable throughput of optical networks.
The results of this work can be applied to future network design and to ensure intelligent
access to achievable capacity.
ID  - discovery10204124
UR  - https://discovery.ucl.ac.uk/id/eprint/10204124/
PB  - UCL (University College London)
M1  - Doctoral
A1  - Matzner, Robin Michael
ER  -