Enabling Network Stability in a Scalable Software-Defined Data Center Network
Loading...
Date
2022
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Zululand
Abstract
In the recent past, Data Centre Network (DCN) has received a lot of research attention because of
the indispensable role that DCNs play in modern-day network communication as; the data storage
container for on-demand service provisioning. One of the major responsibilities of the DCN is the
management of the network’s need for optimization, thereby addressing request services. One
consequence of the increasing requests is network flow interference that incurs network instability.
Therefore, the task of addressing flow interference within a network becomes very important
because it plays a vital role in restoring network stability.
Network scalability continues to be the concern of many networks management engineers while
less consideration is given to the effectiveness and efficiency of their adopted network stability
solutions. Furthermore, very few approaches to network stability gave significant attention to
achieving an optimal solution. Several related works notably still left unaddressed slow
convergence, computational complexity, complex memory requirements and lack of adaptive
stability mechanisms among others. Therefore, the need to enhance network stability in a scalable
software-defined and data-oriented DCN remains an open issue for the research community to
contend with.
This research work addressed this knowledge gap by conducting a comprehensive literature review
on relevant existing approaches used by other researchers to address network stability. This
provided useful insight and motivation to launch a search for a more robust and appropriate
approach to fast network adaptation, and economical utilization of network resources with no
negative impairment on the expected optimal performance. The literature review revealed that the
concept of Adaptive Rendering Technique (ART) in Computer Graphics was generic and
analytically simple enough to adopt for high-level applicability to network flow.
The methodology for this study emerged from using ART as the analogy for modelling the traffic
flow interference which causes network instability. A Multi-Objective Optimization Crosspoint
Queue (MOCQ) was modelled after the primary stages of the ART’s adaptive pipeline processes.
The MOCQ process consisted of a control flow list, sorting entry matching, matching flow table
and matching operation. This process enabled the achievement of stable controller-switch states.
vi
The efficacy of the proposed ART-based MOCQ approach has been extensively validated using
numerical analysis, miniature testbed and computer simulations.
The approach resulted in improved performance over existing approaches as demonstrated by
faster stability, improved response time and network throughput, and speedy rate of convergence.
Furthermore, a balanced perspective of individual interests on the sides of both service providers
and end-users was attained. The emerging facts were confirmed by a measure of a 42.9% reduction
in the rate of network response delays, and over 150% point improvement in correcting the existing
rate of switch failures when compared to the ordinary Crosspoint Queue approach. The approach
also assisted in determining the level of stability attained and in the case of the setup in this thesis,
a level of 0.7 was achieved for the hierarchical data centre network we are considering. This
determination is an improvement over existing systems as they do not go all the way to achieve
this goal.
Description
A thesis submitted to the Faculty of Science, Agriculture and Engineering in fulfilment of the requirements for the Doctor of Philosophy in the Department of Computer Science at the University of Zululand, South Africa, 2021.
Keywords
Data Centre Network, Network scalability, Adaptive Rendering Technique