| The ongoing rush for cloud-based services by small, medium, and large-scale organizations in other to reduce operational cost overhead, and have more flexibility in the deployment of business applications continues unabated. Cloud computing environment allows sharing of configurable computing resources (e.g., CPU, memory, networks, servers, storage, applications, and services) among cloud tenants. Thus, the cloud computing environments has significantly simplified the capacity provisioning process to support massive migration of enterprise applications to the cloud. However, there exist several challenges facing cloud services that are preventing cloud subscribers; individuals, private, and public organizations from achieving total satisfaction. These challenges include CPU sharing, contention for hardware, and lack of fairness in resource management. This dissertation aims at supporting ongoing research efforts in academia and private industry in addressing these challenges.;The contributions of this dissertation start with measurement and interference profiling of application performance in a multi-tenant cloud environment. Through series of quantitative and qualitative experimental research methods in public and private clouds, this dissertation reveals how an imbalance in the allocation of shared cloud computing resources affect in-cloud application performance. The dissertation examined the complex interplay among cloud tenants as they compete for CPU time, and shared hardware resources by identifying the relationship between low-level CPU multiplexing and high-level application performance in multi-tenant cloud.;The dissertation continues by establishing a unique linear mathematical relationship between CPU steal time and overall application performance. Our approach is to monitor the progress of submitted applications at runtime, tracks the slowdown of individual application and applies adjustment until convergence. Thus, when an application suffered more slowdown, we allocate more CPU to reduce unfairness. In establishing system support for fine-grained profiling, we report system level activities at sub-second granularity. We predicted application performance degradation by creating a mathematical relationship between high-level application performance and low-level machine events (i.e., CPU steal time and Misses per thousand instructions). We validate the added value of our approach by comparing application performance slowdowns (average) with various datasets. Based on our experimental results, our approach helps mitigate co-tenant interference and reduces unfairness by minimizing the overall application slowdowns. Lastly, the dissertation concludes by proposing Adaptive CPU Sharing (ACS) in multi-tenants' clouds. ACS leverages the Xen CPU scheduler capability to reduce unfairness in resource sharing among cloud tenants and reduce application performance degradation. |
