HPC Auto-Assistant: Making HPC Software and Tools Easier to Use for the Experienced (Not Expert) User

Our proposal will make high performance computing (HPC) software easier to install, configure, stage data for, and run. This will reduce the need to invest a substantial amount in learning these tools and services and the insurmountable barrier for many organizations, as requested in Topic 2b. Our solution is an intelligent system called HPC Auto-Assistant (HAA). HAA enables the automated creation of the following items to assist a manufacturing or engineering firms staff use HPC: simple, dialog-like user interfaces that hide the complexity of many manual steps, and executable scripts and codes that are based on knowledge of how to perform complex operations with HPC software. HAA helps people develop simple tool interfaces that hide the complexity rather than learn how to make right decisions. We are exploring a new cognitive computing approach to the development and use of HPC. Furthermore, we design HAA to be invoked from our current DOE-funded system, CodeBook, a web application and distributed repository of information on how to install, configure, and run computational algorithms for the DOE community. Thus HAA can gain exposure to DOE labs wherever CodeBook is used. Commercial Applications and Other Benefits: We seek to transform what is now a slow, human-to-human process of information transfer when installing, configuring, or using HPC codes to a human-to-machine process. HAA can yield a substantial intangible return on investment for DOE in the bolstering of the supply chain by lowering the barriers for tier 2 and 3 suppliers to use HPC.