Master of Science in Computer Security and Information Assurance from Rochester Institute of Technology.
Title: Differential Virtualization for Large-Scale System Modeling
Abstract: Today’s computer networks become more complex than ever with a vast number of connected host systems running a variety of different operating systems and services. Academia and industry alike realize that education in managing such complex systems is extremely important for computer professionals because, with computers, there are many levels of detailed configuration. Configuration points can occur during all facets of computer systems including system design, implementation, and maintenance stages. In order to explore various hypotheses regarding configurations, system modeling is employed – computer professionals and researchers build test environments. Modeling environments require observable systems that are easily configurable at an accelerated rate. Observation abilities increase through re-use and preservation of models. Historical modeling solutions do not efficiently utilize computing resources and require high preservation or restoration cost as the number of modeled systems increases. This research compares a workstation-oriented, virtualization modeling solution using system differences to a workstation-oriented, imaging modeling solution using full system states. The solutions are compared based on computing resource utilization and administrative cost with respect to the number of modeled systems. Our experiments have shown that upon increasing the number of models from 30 to 60, the imaging solution requires an additional 75 minutes; whereas, the difference-based virtualization solution requires an additional three (3) minutes. The imaging solution requires 151 minutes to prepare 60 models, while the difference-based, virtualization solution requires 7 minutes to prepare 60 models. Therefore, the cost for model archival and restoration in the difference-based virtualization modeling solution is lower than that in the full system imaging-based modeling solution. In addition, by using a virtualization solution, multiple systems can be modeled on a single workstation, thus increasing workstation resource utilization. Since virtualization abstracts hardware, virtualized models are less dependent on physical hardware. Thus, by lowering hardware dependency, a virtualized model is further re-usable than a traditional system image. If an organization must perform system modeling and the organization has sufficient workstation resources, using a differential virtualization approach will decrease the time required for model preservation, increase resource utilization, and therefore provide an efficient, scalable, and modular modeling solution.