An Automated Virtualization Performance Analysis Platform

Abstract

This research compares three traditional categories of virtualization to a technique known as hybrid virtualization. Each technique is evaluated in terms of both capability and performance. The traditional methods of platform virtualization such as full virtualization, paravirtualization and operating system virtualization each comes with its own set of capabilities and engineering trade-offs. Hybrid virtualization attempts to leverage the benefits of full and operating system virtualization by allowing virtual machines of each type of virtualization to run simultaneously on the same host machine. This research measures the time required for each virtualization technique to perform a workload inside virtual machines as the number of virtual machines running the workload scales. This performance data will help determine the usefulness of the hybrid technique in building a military cyber warfare training simulation environment based on virtualization. The goal is to determine which technique is capable of supporting large-scale environments required by realistic network training scenarios. The capability evaluation results indicate that hybrid virtualization successfully leverages the benefits of its two virtualization components while minimizing the trade-offs of each individual technique. The performance results indicate that the performance of each virtualization technique differs significantly relative to the workload applied. Some workloads saw no significant differences in performance between techniques. However, in the workloads that did show significant difference, the hybrid technique performed as well as or better than full virtualization or operating system virtualization alone. This leads to the conclusion that hybrid virtualization is a viable candidate as the basis for a military cyber warfare simulation and training environment.