Nayun Cho, Mino Ku, Rui Xuhua, and Dugki Min, Konkuk University, Korea

The rapid growth of networking and storage capacity allows collecting and analyzing massive amount of data by relying increasingly on scalable, flexible, and on-demand provisioned largescale computing resources. Virtualization is one of the feasible solution to provide large amounts of computational power with dynamic provisioning of underlying computing resources. Typically, distributed scientific applications for analyzing data run on cluster nodes to perform the same task in parallel. However, on-demand virtual disk provisioning for a set of virtual machines, called virtual cluster, is not a trivial task. This paper presents a feature model-based commonality and variability analysis system for virtual cluster disk provisioning to categorize types of virtual disks that should be provisioned. Also, we present an applicable case study to analyze common and variant software features between two different subgroups of the big data processing virtual cluster. Consequently, by using the analysis system, it is possible to provide an ability to accelerate the virtual disk creation process by reducing duplicate software installation activities on a set of virtual disks that need to be provisioned in the same virtual cluster.

Virtual Cluster Disk Provisioning, Feature Model-based Virtual Cluster Commonality and Variability Analysis