This thesis is a study of the theory of scalability and its application in an infrastructure as a service (IaaS) cloud context. The cloud based utility computing paradigm is presented along with how scalability principles are applied in the cloud. The differences of scalability in general and the cloud concept of elasticity are discussed.
A quality of elasticity (QoE) metric is developed to facilitate factual discussion and comparison of different cloud platforms' elasticity capabilities and the effectiveness of elastic scaling strategies. The metric is based on business requirements expressed as preference functions over a set of lower level metrics. Multi-criteria analysis is applied to these possibly conflicting preferences to arrive at a unified value of utility based on weighting the sum of the preferences. QoE reflects the utility of the system over time.
The concept of an elasticity controller application is presented and a prototype implementation described in order to exercise the QoE metric. Two load testing scenarios are executed against a simple test application whose deployment is managed by the prototype controller. The elastic scaling behavior of the system is analyzed in terms of the QoE results to confirm the prototype is functional and to find areas of improvement.
