The Pervasiv1e Technology Institute (PTI) at Indiana University selected proposals for the second phase of the Humans Advancing Research in the Cloud (HARC) program. PTI will invest nearly $180K in the work of Cloud Research Support Engineers (CRSEs) at four U.S. institutions: Georgia State University, University of Notre Dame, University of Pittsburgh, and Arizona State University.
The HARC project, with initial funding from the Microsoft Corporation, is an independent effort by PTI to advance the use of cloud-based cyberinfrastructure (CI) for research through increasing the understanding of the important role that people play in supporting that use.
We are excited to add four new partners to our efforts to illuminate the understanding the critical role that people play in supporting the use of cloud-based cyberinfrastructure to advance research.
Craig A. Stewart, PTI executive director and HARC executive investigator
The overall goal of the HARC project is to provide illumination for institutions of the need for investment in people in enabling research using cloud CI. Too often, institutions and funding agencies focus on costs of hardware, software and other physical resources involved in use of CI to advance research. Without the role that people play, those investments do not reach their full potential of impact. This is even more critical a factor as research moves away from on-premise CI and into the cloud, where the complexities of those environments, the variations in offerings from commercial cloud providers, and interacting with provider service models designed more for administrative use increase the challenge to researchers.
The HARC project consists of several elements. In 2019, the first group of eight CRSEs was engaged as individual contributors, providing insights into how their work enabled the use of cloud CI rather than traditional premise-based resources across several disciplines. Their experiences were captured and shared with the community via presentations delivered and papers published in the broader academic record. As HARC advanced past this first phase, it has broadened its deliverables to include continued raising of awareness of the value of people moving to the cloud, examining the return on investment for cloud alternatives, and creating resources to be used by the national community.
“We are excited to add four new partners to our efforts to illuminate the understanding the critical role that people play in supporting the use of cloud-based cyberinfrastructure to advance research,” said Craig A. Stewart, PTI executive director and executive investigator on the HARC project. “These individuals, working closely with Indiana University, will also expand the resources offered by HARC to support the broader community of individuals in these critical roles across the U.S. and globally.”
The successful proposals are:
Georgia State University—Cloud computing for research on secure privacy-preserving machine learning. This project contributes to HARC’s goal of fully achieving meaningful results from use of commercial vendor cloud-based cyberinfrastructure to demonstrate the value of these environments in research. Machine learning algorithms based on deep neural networks have achieved remarkable results in a variety of domains. However, machine learning solutions require access to the raw data, which creates potential security and privacy risks, and machine learning based on neural networks can be fragile and is vulnerable to several attack vectors. To address these issues, the project seeks to develop new techniques to provide solutions to run deep learning algorithms over encrypted data and enable running the algorithms without parties having to reveal their sensitive data. While protecting privacy is the primary focus, combining advanced encryption techniques with deep learning algorithms could potentially help defend against some adversarial attacks.
“Our inclusion in the HARC project will demonstrate benefits of using commercial cloud platforms in developing trustworthy AI systems and secure privacy-preserving machine learning and provide evidence on how human resources support and contribute to realizing these benefits, showcasing potential benefits of using commercial cloud platforms in other domains since machine learning services are used in a variety of domains,” said Daniel Takabi, founding director of the INSPIRE Center at GSU. “Our work with HARC will document the advantages of using commercial cloud platforms in academic research and highlight their potential for higher education institutions.”
University of Notre Dame—Accelerating machine learning across many research areas using a hybrid cloud approach. The Cyberinfrastructure for Accelerating Machine Learning (CAML) project will use cloud-based computational acceleration resources to target two specific use cases: (1) providing researchers with access to a broader variety of heterogeneous accelerator architectures and (2) providing on-demand, short-term burst access for student teaching and learning objectives such as machine learning coursework and workshops. The multiple teams working with CAML will share priority goals to (1) develop training materials for faculty, postdoctoral researchers and students, (2) make the CAML infrastructure accessible to the classroom with integration into a number of courses, and (3) integrate the CAML-supported research into multiple existing programs for high school students and undergraduate researchers.
“The HARC support will help us empower students and research support personnel to have the knowledge and tools to use the cloud efficiently and effectively,” said Scott Hampton, assistant director, Center for Research Computing at Notre Dame. “As funding agencies increasingly award grants emphasizing both cloud and local components, we think that a well-documented experience and the presence of tools in the community will assist other researchers making the transition.”
The winning proposals are from Georgia State University, University of Notre Dame, University of Pittsburgh, and Arizona State University.
University of Pittsburgh—Leveraging existing humanware for research in the cloud through judicious bursting. The Center for Research Computing at the University of Pittsburgh (CRC) operates an on-premise HPC cluster, providing compute and storage resources supporting various computational models, including traditional distributed computing, shared-memory processing, high-throughput computing, and accelerator-based computing. Missing from CRC’s premise-based resource profile are nodes with extreme amounts of system memory, nodes supporting accelerating inferencing, and analytics environments that conform to HIPAA requirements. These and other missing capabilities can be filled using the public cloud. Additionally, the public cloud can serve as a target for cloud bursting when demands for resources exceed on-premise capacity. With these use cases in mind, the project aims to transform static, fixed-size, on-premise HPC cluster into an elastic computing environment by integrating local resources with Azure, Google Cloud, and AWS.
“The Center for Research Computing at the University of Pittsburgh is delighted to be working with the Pervasive Technology Institute to investigate and publicize the effort required to get actual applications migrated from on-premises resources to the cloud,” said Ralph Roskies, vice chancellor for research computing at the University of Pittsburgh. Principal Investigator Kim Wong, research associate professor, CRC, added: “I am excited to be joining the HARC initiative and given the opportunity to contribute, both from the technical as well as the end-user support perspectives, to the broader discussion on the return on investment of public clouds for supporting research. The tools and new knowledge that we develop here will have a lasting impact on our community.”
Arizona State University—Computational materials design in the cloud. Through the HARC project, Arizona State University (ASU) will explore computationally assisted materials design with direct applications in environmental stewardship. Specifically, the proposed work aims to advance applied research in the development of plastic materials to increase their useful lifespan and simplify recycling efforts by obviating the need to pre-sort plastics. Moreover, by contributing code and pre-built tools, documentation, and best practices for the set-up and maintenance of ASU’s cloud instance, the funding will support ASU contributions to help build resources for the broader cloud research support engineer community.
“Due to the effects of environmental degradation, wear, and the prevalence of single-use plastics, there is a large degree of plastic waste produced that is difficult to recycle and not biodegradable, negatively impacting our environment,” said Douglas Jennewein, senior director of research technology at ASU. “The HARC funding will be a great asset to our efforts to advance the use of computationally aided materials design to address this critical challenge and allow us to contribute value to the broader research support community.”
The four project participants will help advance the overall goals of HARC in several ways:
- CRSEs and investigators will make contributions to the academic record through publications of their results related to the important role people play in supporting cloud CI during the course of their research.
- CRSEs and investigators will be featured in a workshop to be presented at the now-virtual PEARC20 Conference later in the summer. Related, a call for participation has been made, inviting others in the community to contribute papers to the workshop, which is co-sponsored by the National Institutes of Health Science and Technology Research Infrastructure for Discovery, Experimentation, and Sustainability.
- CRSEs, along with HARC/PTI staff at Indiana University, will contribute to the further development of an online GitHub repository, featuring a collection of papers, presentations, documentation of experiences and best practices, and of particular value, pre-built tools for a wide variety of commercial and community clouds used by researchers. The latter will allow people engaged in CI enablement across the community to build on the experiences of colleagues, saving them time by providing reusable valuable code.
About the Pervasive Technology Institute
The Pervasive Technology Institute at Indiana University is a collaborative organization designed to marshal IU’s computational experts and resources quickly in response to societal, research and educational needs. In partnership with UITS, the Pervasive Technology Institute also led the original Jetstream award. The institute was established in 1999 by a generous grant from the Lilly Endowment and has continued to lead productive uses and applications of research technologies for over 20 years.