Doctor of Philosophy
Electrical and Computer Engineering
Cloud computing has become a business reality that impacts technology users around the world. It has become a cornerstone for emerging technologies and an enabler of future Internet services as it provides on-demand IT services delivery via geographically distributed data centers. At the core of cloud computing, virtualization technology has played a crucial role by allowing resource sharing, which in turn allows cloud service providers to offer computing services without discrepancies in platform compatibility.
At the same time, a trend has emerged in which enterprises are adopting a software-based network infrastructure with paradigms, such as software-defined networking, gaining further attention for large-scale networks. This trend is due to the flexibility and agility offered to networks by such paradigms. Software-defined networks allow for network resource sharing by facilitating network virtualization. Hence, combining cloud computing with a software-defined network architecture promises to enhance the quality of services that are delivered to clients and reduces the operational costs to service providers. However, this combined architecture introduces several challenges to cloud service providers, including resource management, energy efficiency, virtual network provisioning, and controller placement.
This thesis tackles these challenges by proposing innovative resource provisioning techniques and developing novel frameworks to improve resource utilization, power efficiency, and quality of service performance. These metrics have a direct impact on the capital and operational expenditure of service providers.
In this thesis, the problem of virtual computing and network provisioning in geographically distributed software-defined network-enabled cloud data centers is modeled and formulated. It proposes and evaluates optimal and sub-optimal heuristic solutions to validate their efficiency. To address the energy efficiency of cloud environments that are enabled for software-defined networks, this thesis presents an innovative architecture and develops a comprehensive power consumption model that accurately describes the power consumption behavior of such environments. To address the challenge of the number of software-defined network controllers and locations, a sub-optimal solution is proposed that combines unsupervised hierarchical clustering. Finally, betweenness centrality is proposed as an efficient solution to the controller placement problem.
Alhazmi, Khaled M., "Optimization and Management Techniques for Geo-distributed SDN-enabled Cloud Datacenters' Provisioning" (2018). Electronic Thesis and Dissertation Repository. 5587.
Available for download on Thursday, December 31, 2020