SI2-SSE: Open OnDemand: Transforming Computational Science through Omnidisciplinary Software Cyberinfrastructure

Ohio State University, The, Columbus OH

Investigators

Abstract

Supercomputing, or High-Performance Computing (HPC), has the power to advance work in all fields of science and engineering. Unfortunately, the impact of HPC is often limited simply because the computers themselves are difficult to use and scientists and engineers would rather spend their time advancing their disciplines than learn HPC. Learning esoteric commands is a hurdle to many students and researchers when they first begin to work with traditional HPC systems, which has contributed to the relatively small proportions of women, minorities, and persons with disabilities within related STEM fields. The Open OnDemand project addresses this problem through innovative software that makes HPC no more difficult than using a desktop computer or a web site, hence reducing that initial learning curve. Open OnDemand will provide an enhanced infrastructure for research and education, in that students and educators will be exposed to the same tools and techniques on smaller departmental and classroom systems that they'll eventually utilize at larger HPC centers. This streamlines the pipeline for developing a more globally competitive STEM workforce that is prepared to dive right into computational problems and resources once they graduate from academia. Open OnDemand also allows scientists and engineers to make specialized domain-specific workflows and expertise available to more collaborators and users, both from academia and industry. This should result in increased partnerships, and the transfer of more technologies from the public to private sectors. The web has become the dominant access mechanism for remote compute services in every computing area except high-performance computing (HPC). Accessing HPC cyberinfrastructure (CI) resources today, either at the campus or national level, typically requires advanced knowledge of UNIX, familiarity with command-line interfaces and installation and configuration of custom client software. Web applications in HPC today do exist in the form of science gateways. However, gateways have not proliferated in part due to the development and administrative overheads required for each individual gateway. These factors demonstrate an accessibility gap for HPC. Open OnDemand is an open source platform for HPC and remote computing access that addresses the accessibility gap. Open OnDemand will be a public release based on the successful OSC OnDemand platform. OSC OnDemand is a web platform providing Ohio Supercomputer Center (OSC) users integrated access to HPC systems, web applications, and VNC services. OSC OnDemand has been in production since January 2013, has over 800 distinct users from 27 different NSF fields of science, and its apps have been launched over 70,000 times. In addition to easing access to HPC services, Open OnDemand centralizes web app overheads, easing support for custom visualization and science gateway apps. Open OnDemand has the ability to transform cyberinfrastructure by providing a platform to enable a new delivery method for scientific web tools like HubZero apps, XSEDE Science Gateways, iPython notebooks and workflow tools like Pegasus. Under this project, the Open OnDemand platform will be created by (a) transitioning the existing OSC OnDemand software to a community developed project hosted on GitHub, (b) extending the per-user web server to serve Rails apps, (c) replacing the custom-developed Proxy with an existing open source project like NGINX and (d) replacing the existing Java-based VNC client with an HTML5 solution. File usage will be improved by updating or replacing AjaXplorer and integrating high performance file transfer functions (sftp and Globus Online). The accessibly apps will be updated by by (a) upgrading or replacing AnyTerm, (b) updating Job Constructor and (c) integrating Open XDMoD and SUPREMM for job and cluster performance metrics. In addition, community infrastructure will be created including (a) system administrator documentation and discussion forums and (b) training materials based on existing OSC OnDemand materials. A beta program will be conducted including (a) assisting beta sites with installs, (b) updating training materials to include new functionality, (c) beginning metric reporting and (d) documenting a Galaxy case study for app integration. HPC center staff will be engaged by through the "boot camps" and "train the trainer" sessions. Finally, Open OnDemand will be proposed as a project to the XSEDE Campus Bridging Technology Insertion Service. These activities will help meet the following objectives: (1) Transition OSC OnDemand to a community-developed open source software package called Open OnDemand, (2) Improve the interface capabilities of Open OnDemand by updating and expanding the accessibility apps, including integration of Globus Online and Open XDMoD projects. (3) Conduct a program to engage departmental, campus and national HPC users and administrators on enhancing HPC inclusivity through Open OnDemand and (4) Leverage Open OnDemand as a platform to support existing web-based applications such XSEDE Science Gateways and HubZero applications based on our experience supporting a data-intensive biomedical package (the Galaxy Project).