Elastic Optical Networks and Probablistic Design

Lead Research Organisation: University of Cambridge
Department Name: Engineering


The UK optical communication network requires constant improvement to route ever increasing demands at suitable QoS and latencies. Such an improvement could be found from increasing the efficiency of bandwidth allocation and from improved modelling of networks at a large scale.
This project seeks to model and design an elastic optical network, wherein WDM channels can be grouped together, providing larger bandwidth for larger demands, or split into several separate channels to reduce overhead on smaller demands, thus increasing network efficiency.
Given the scale of the UK communications grid, modelling this requires novel approaches to representing this type of system to shrink the computing requirements to reasonable levels. Statistical physics is purpose built for this task and by using concepts such as entropy and temperature to represent states of network activity we seek to reduce the computational complexity of the problem. This will allow modelling beyond the network backbone and could even show emergent 'group' properties of such systems.
This representation will be paired with probabilistic design methods to provide either efficient ways to reinforce the current network or an entirely new network design which could replace the current network when economically viable.
This will improve the performance and reliability of the UK communications infrastructure, ensuring its ability to cope with future increases in internet usage such as the 'Internet of Things' or high bitrate video streams.


10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/N509103/1 01/10/2015 31/03/2021
1775341 Studentship EP/N509103/1 01/10/2016 31/12/2020 Robert James Vincent
Description I showed in a recent paper (10.1109/JLT.2019.2942710) that the capacity of a nonlinear optical network can be estimated accurately and quickly by a new routing and wavelength algorithm. I verified these results for a range of topologies representative of both national- and continental-scale networks.
In another paper ( 10.1364/JOCN.11.000C76), I employed the above approach to compare the effect of system margin on network capacity. It was found that metropolitan-scale networks are more resilient to system margin due to the logarithmic association between signal-to-noise ratio (SNR) and spectral efficiency. Equivalently it implies that large networks, with longer transmissions and hence lower SNR, need to be designed more carefully as network capacity is more sensitive to system margin.
At the Optical Fibre Communications Conference 2020 in San Diego I shall compare the effect of amplifier procurement on link capacity. This subject has been mostly ignored in core networks because reconfigurable optical add-drop multiplexers (ROADMs) can remove power imbalances between channels of different wavelengths. In a space-division multiplexing regime, there could be fewer ROADMs meaning power discrepancies propagate through the network. I show that a small inventory of amplifiers would put operators within 10% of optimal performance whilst introducing minimal increased complexity.
Exploitation Route The routing and wavelength algorithm I developed is not only high performance but it is deceptively simple to implement. This makes it ideal for automated network control which will be essential if the information economy continues to demand ever higher data rates. The networks I generated to test this algorithm have already been shared with researchers at UCL under the TRANSNET project EP/R035342/1.
There is further work to be done in understanding what is an 'optimal' topology. This is especially true in the context of uncertain regulatory frameworks.
Sectors Digital/Communication/Information Technologies (including Software)