LIBRARY AND TECHNOLOGY SERVICES P - AGE 6 - - ▶ RESEARCH COMPUTING The mission of the Research Computing Team at Lehigh is twofold. We strive to deliver the raw compute, storage, and connectivity necessary to support a diverse research portfolio. We also engage with Lehigh faculty and students to consult on how new and experi mental technology can drive their research forward. The foundation of our research computing services is our on-premises high-performance computing cluster. We provide an easy on-ramp for both large and small research projects to use high-performance computing thanks to a combination of NSF-funded resources and a faculty condominium (condo) program. In spring 2024 we expanded this cluster by nearly 50% to over 6,800 cores by consolidating faculty condo investments. Each of our new Intel Sapphire Rapids compute nodes provides a half-terabyte of memory and 64 cores that can be useful for prototyping and scaling both massively parallel and high-throughput computing projects. We have also added 48 new graphics processing units (GPUs) capable of supporting both the wide catalog of physics-based computations as well as the latest generation of artificial intelligence applications. We are currently hosting two pilot projects that use these GPUs for self-hosted open-source large language model (LLM) projects that act on highly-customized datasets. Our on-premises HPC resources continue to provide the easy access and large scale required by a large diversity of academic disciplines. In concert with the large demand for computing power, we have seen an even broader interest in new tools for managing research data. Following our pilot program from last year, we have developed a long-term plan to expand our high-speed storage system with solid-state drives. Researchers continue to develop intensive input-output (I/O) workflows that require higher bandwidth parallel filesystems. Research Computing has coordinated with our faculty steering committee to propose and implement new policies aimed at delivering a combination of both economical and high-speed storage options for our user base. We plan to continue expanding our storage options to meet increasing demand for data management and sharing plans mandated by many federal funding agencies. While we deepen and strengthen our traditional high-perfor mance computing and parallel storage infrastructure, we have also expanded to the cloud in two ways. First, we continue to operate science gateways that reside in our on-premises infrastructure. These offer the largest amount of control and security for exploratory projects. Second, we have continued to evangelize the use of the Secure Research Cloud (SRC), a Lehigh-curated computing environment built with Amazon Web Services (AWS) components that is compliant to host sensitive data. The SRC in particular is crucial to a research computing strategy that seeks to make scalable computing possible for all kinds of high-value data acquired by researchers funded by the Departments of Energy or Defense, or who build research projects that require protected health information. Our Research Computing Team cooperates with LTS infor mation security, systems engineering, and library teams to provide research support. We provide direct consultations that members of our community can request from a unified request system at lts.lehigh.edu/help. This system helps recruit the right set of experts from across LTS to guide researchers to the right technologies and services to move their work forward. Research Computing hosts a combination of office hours and cumulative HPC seminars to introduce new members of our WHO’S USING HPC 210 active users 54 active principal investigators (PIs)
RkJQdWJsaXNoZXIy MTA0OTQ5OA==