Date of Award

12-2013

Document Type

Thesis

Degree Name

Master of Science (MS)

Legacy Department

Computer Science

Advisor

Hallstrom, Jason O

Committee Member

Apon , Amy

Committee Member

Malloy , Brian A

Abstract

Emerging cloud computing infrastructure models facilitate a modern and efficient way of utilizing computing resources, enabling applications to scale with varying demands over time. The core enabler of the cloud computing paradigm is virtualization. The concept is not new; virtualization has garnered significant research attention, fostering a race to achieve the lowest possible virtualization overhead. The evolution has led to three primary virtualization approaches: full-virtualization, para-virtualization, and container-based virtualization, each with a unique set of strengths and weaknesses. Thus it becomes important to study and evaluate their quantitative and qualitative differences. The operational requirements of the Intelligent River® middleware system motivated us to compare the choices beyond the standard benchmarks to bring out the unique benefits and limitations of the three virtualization approaches. This thesis evaluates representative implementations of each approach: (i) full-virtualization - KVM, (ii) para-virtualization - Xen, and (iii) container-based virtualization - Linux Containers. First, this thesis discusses the design principles behind the chosen virtualization solutions. Second, this thesis evaluates the solutions based on the overhead they impose to virtualize system resources. Finally, this thesis assesses the benefits and limitations of each solution based on their operational flexibility, and resource entitlement and isolation facilities. The study presented in this thesis provides an improved understanding of available virtualization technologies. The results will be useful to system architects in selecting the best virtualization platform for a given application.

Share

COinS