Enabling High Productivity Computing through Virtualization

The inherent complexity in utilizing and programming high performance computing (HPC) systems is the main obstacle to widespread exploitation of HPC resources and technologies in the Department of Defense (DoD). Consequently, there is the persistent need to simplify the interface for the generic user. This need is particularly acute in the Signal/Image Processing (SIP), Integrated Modeling and Test Environments (IMT), and related DoD communities where typical users have heterogeneous unconsolidated needs. Typically, in these communities, most users work from a Windows PC and just the overhead involved in learning UNIX well enough to accomplish HPC tasks is itself a significant barrier. Virtualization is a technique for hiding the physical characteristics of computing resources from the way in which other systems, applications, or end users interact with those resources. In this proof-of-concept study, we report on our experience with SIP applications running on top of a virtual Windows platform -guest OS- that itself runs on top of High Performance Computing Modernization Program (HPCMP) HPC platforms running Linux -host OS-. Productivity is strongly influenced by the workflow of the user (e.g., time spent running vs. time spent programming). Therefore, we study performance and productivity aspects of this novel approach using the realistic and prototypical SIP algorithms embedded in the DARPApsilas High Productivity Computing Systems (HPCS) Challenge Scalable Synthetic Compact Application Benchmark # 3 (SSCA # 3). SSCA # 3 is also part of the High Performance Embedded Computing (HPEC) benchmark challenge and contains sensor processing (SAR) and knowledge formation (ATR) algorithms. Our conclusion is that virtualization technology is a promising approach to preserve desktop PC user productivity while, transparently, allowing the exploitation of HPC systems. This may ease the migration of SIP and other non-traditional computation technology areas (CTAs) codes to HPC platforms fostering the utilization of HPCMP resources among these communities.