Dynamic Resource Management using Operating System-Level Virtualization

Abstract: This thesis expands upon an existing system called Golondrina that performs autonomic workload management among a cluster of hardware nodes running operating system-level virtualization. Golondrina works by identifying localized resource stress situations and attempting to dissipate them by reallocating system resources and, if necessary, migrating or replicating virtual machines. It is predicted that, using Golondrina, efficiency of similar systems can be further improved by achieving greater resource utilization on the hardware nodes while maintaining resource availability for each virtual machine. The following topics are discussed: virtualization technologies and associated challenges relating to resource management, the architecture and design of Golondrina, intelligent resource reallocation based on predefined policies, and preliminary results demonstrating the effects of a memory resource management policy on the performance of a web application hosted in a virtual environment. This research makes a significant contribution to the study of virtualized data centres since currently no other system considers virtual machine replication and dynamic memory reallocation as an approach to workload management.