PhD Oral Exam - Hadeel El-Kassabi, Information and Systems Engineering
When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.
Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.
Cloud computing has emerged as a promising and powerful paradigm for delivering data-intensive, high performance computation, applications and services over the Internet. Cloud Computing has enabled the implementation and success of Big Data, a relatively recent phenomenon consisting of the generation and analysis of abundant data from various sources. Accordingly, to satisfy the growing demands of Big Data storage, processing, and analytics, a large market has emerged for Cloud Service Providers, offering a myriad of resources, platforms, and infrastructures. The proliferation of these services often makes it difficult for consumers to select the most suitable and trustworthy provider to fulfill the requirements of building complex workflows and applications in a relatively short time.
In this thesis, we first propose a quality specification model to support dual pre- and post-cloud workflow provisioning, consisting of service provider selection and workflow quality enforcement and adaptation. This model captures key properties of the quality of work at different stages of the Big Data value chain, enabling standardized quality specification, monitoring, and adaptation.
Subsequently, we propose a two-dimensional trust-enabled framework to facilitate end-to-end Quality of Service (QoS) enforcement that: 1) automates cloud service provider selection for Big Data workflow processing and 2) maintains the required QoS levels of Big Data workflows during runtime through dynamic orchestration using multi-model architecture-driven workflow monitoring, prediction, and adaptation.
The trust-based automatic service provider selection scheme we propose in this thesis is comprehensive and adaptive, as it relies on a dynamic trust model to evaluate the QoS of a cloud provider prior to taking any selection decisions. It is a multi-dimensional trust model for Big Data workflows over competing clouds that assesses the trustworthiness of cloud providers based on three trust levels: (1) presence of the most up-to-date cloud resource verified capabilities, (2) reputational evidence measured by neighboring users, and (3) a recorded personal history of experiences with the cloud provider.
The trust-based workflow orchestration scheme we propose aims to avoid performance degradation or cloud service interruption. Our workflow orchestration approach is not only based on automatic adaptation and reconfiguration supported by monitoring, but also on predicting cloud resource shortages, thus preventing performance degradation. We formalize the cloud resource orchestration process using a state machine that efficiently captures different dynamic properties of the cloud execution environment. In addition, we use a model checker to validate our monitoring model in terms of reachability, liveness and safety properties.
We evaluate both our automated service provider selection scheme and cloud workflow orchestration, monitoring and adaptation schemes on a workflow-enabled Big Data application. A set of scenarios were carefully chosen to evaluate the performance of the service provider selection, workflow monitoring, and the adaptation schemes we have implemented. The results demonstrate that our service selection outperforms other selection strategies and ensures trustworthy service provider selection. The results of evaluating automated workflow orchestration further show that our model is self-adapting, self-configuring, reacts efficiently to changes, and adapts accordingly while enforcing QoS of workflows.