Flexible Optical Systems and Reconfigurable Disaggregated Data Centre Networks

Abstract

Optical metro/core and data centre networks are under pressure to manage the exponential growth and dynamic nature of internet traffic. As a result, innovative technologies and approaches to improve the flexibility of the network infrastructure and operations need to be developed. However, the provisioning of networks with innovative technologies and resources to deliver flexibility without any design or allocation policy does not guarantee cost efficiency and optimum network performance when managing traffic demands. The introduction of flexibility as a measurable key performance indicator for networks has opened up potential prospects for the design of efficient optical networks. However, there is limited understanding on how quantitative levels of flexibility relates to other key performance indicators and design features. In addition, the combination of server resource disaggregation, electrical/optical technologies and network function programmability in data centres provides a promising solution to eliminate the limitations of conventional server-centric data centres. However, for this concept to materialize, resource allocation policies which ensure an optimum level of network performance must be developed.

The first part of this thesis investigates flexibility as a measurable key performance indicator for optical transmission and switching systems. Flexibility measurement models for different optical components and subsystems are derived and proposed based on maximum entropy. In addition, the flexibility of the examined optical components and subsystem are measured, and the relationship between flexibility, other measurable key performance indicators and design features are highlighted and discussed. The second part of this thesis investigates network strategies and algorithms for selecting and deploying electronic packet switching/optical circuit switching services and network resources to build Virtual Machines on reconfigurable disaggregated data centre resources. A comprehensive performance evaluation of the proposed network strategies and algorithms across different disaggregated data centre architectures is conducted.

Date of Award	23 Jan 2019
Original language	English
Awarding Institution	The University of Bristol
Supervisor	Georgios Zervas (Supervisor) & George Oikonomou (Supervisor)