Dynamic topologies for sustainable and energy efficient traffic engineering in communication networks

Aldraho, Abdelnour Mohamed Alnour (2013) Dynamic topologies for sustainable and energy efficient traffic engineering in communication networks. [Thesis (PhD/Research)]

Text (Whole Thesis)

Download (7Mb) | Preview


Energy consumption and related emissions have been in the public focus for some time. Contributions of the Information and Communication Technol- ogy (ICT) sector to increase the Greenhouse Gas (GHG) emissions are growing. Networks are responsible of a significant portion of the ICT energy foot- print and are generally dimensioned for peak loads. For extended off-peak periods, resources continue to consume power, but are lightly used or un- used. The goal of this project is to reduce power consumption in commu- nication networks through network management techniques. This research investigates the concept of dynamic topologies, i.e. networks that adapt their topology according to traffic volume.

In contrast to related work, this thesis addresses networks where nodes are both emanating and consuming traffic. This requires power models for routers and a reduced functionality power-state is proposed that bridges local de- mands to a single interface.

The key aim of this study was to investigate power reductions that can be achieved by dynamic topologies. It proposes a novel network transformation
and introduces mathematical programming models that result in energy-optimal topologies for given traffic loads. This part focuses on the optimisation prob- lems and studies gains in static environments. Numerical results are pre-
sented for example networks using a large set of traffic matrices.

Efficient heuristics are necessary for larger networks as mathematical pro- gramming models cannot be solved in practical time frames. Two sets of al- gorithms are proposed to find minimal network topologies. These rely either on link utilisation or node gravity to decide whether active devices can be switched off. To avoid hot spots and link overloads, shortest path weight set- ting techniques are implemented.

Network resilience to failure is an important requirement of network oper- ators. To account for resilience constraints, two additional programming models are formulated; one that protects individual links and one that pro- tects traffic demands. Both models are studied and energy savings are com- pared to the original models.

To demonstrate the feasibility of the approach a potential implementation of dynamic topologies using Multi Protocol Label Switching (MPLS) networks is introduced. Most MPLS functions and nodes are not affected by the proposal. A flow tracking and topology tracking mechanism is required at the network ingress; and all nodes have to include a power management function that controls the power state of routers. The impact of changes in routing patterns on active UDP and TCP flows has been investigated and found to be minimal. Aggregated flow-based performance has been analysed and the results show that there is no discernable impact on network performance.

Adapting topologies of computer networks dynamically to traffic volumes is feasible and can lead to significant reductions in energy footprints. For the test networks, dynamic topologies reduce the average network power con- sumption, by 12-52 per cent depending on network load.

Statistics for USQ ePrint 28782
Statistics for this ePrint Item
Item Type: Thesis (PhD/Research)
Item Status: Live Archive
Additional Information: Doctor of Philosophy (PhD) thesis.
Faculty/School / Institute/Centre: Current - Faculty of Health, Engineering and Sciences - School of Mechanical and Electrical Engineering
Supervisors: Kist, Alexander; Maxwell, Andrew
Date Deposited: 16 Feb 2016 23:48
Last Modified: 17 Feb 2016 01:01
Uncontrolled Keywords: energy efficiency; computer networks; dynamic topologies
Fields of Research : 10 Technology > 1005 Communications Technologies > 100503 Computer Communications Networks
URI: http://eprints.usq.edu.au/id/eprint/28782

Actions (login required)

View Item Archive Repository Staff Only