The University of Kentucky (UK) Center for Computational Sciences (CCS) has supported high performance computing (HPC) for scientific research and teaching since 1987. The CCS supercomputing facilities are available to the faculty at UK and to research institutions throughout the Commonwealth of Kentucky. CCS supports more than 75 research groups in a variety of disciplines. Historically support has centered around the natural sciences (e.g., physics, chemistry, and biology) and engineering. However, in recent years, UK’s computational research landscape has undergone several significant changes:
- A Shifting Researcher Population: Researchers no longer hail (primarily) from Arts & Sciences and Engineering, but now include researchers from essentially all areas on campus (e.g., Agriculture, Pharmacy, Business, Design, Medicine, Linguistics, Fine Arts);
- New Computational Requirements: Computational requirements now extend beyond conventional HPC systems, and are being driven by new research demands involving big data, massive/elastic parallelization, and custom/special-purpose applications;
- New Computational Environments: Researchers must now choose from a complex set of possible computational environments and solutions including choices of hardware, software, physical vs. virtual environments, and private (on-prem) vs public (remote) cloud solutions.
- Increasing Expertise Requirements: Access to hardware alone is not sufficient. New demands for domain expertise/knowledge/training (i.e. Humanware) now dominate.
- New Organizational Relationships: The need for infrastructure maintainers (ITS) to work seamlessly with user support (CCS) is critical for today’s complex computational solutions.
As a result of the changing computational demands, CCS has been expanding its traditional HPC systems and support to include new and emerging computational solutions. Examples include new loosely-coupled, reconfigurable, virtualized high memory systems designed to support computation on big data sets, new GPU clusters designed to support machine learning and deep learning, new object storage systems designed to provide fast, cost-effective, access to big data, and on-prem private cloud services that allow users to seamless scale their research problem size from desktop to on-prem resources to public cloud and national labs (See additional information below). In addition to providing multiple systems and technologies relevant to varying application challenges, CCS also provides consulting and technical documentation as well as workshops, seminars, lectures, on-boarding to national labs, and other training opportunities to assist researchers in using these resources effectively.
In addition, CCS/ITS are developing relationships with the three main Cloud Service providers: Google’s Cloud Platform, Microsoft’s Azure and Amazon Web Services (AWS). CCS has offered faculty and staff meetings and workshops related to use of these systems and will continue to provide information and support as need to interested faculty. It is CCS’s intention to develop our own Ky Research Informatics Cloud whereby UK researchers can develop their Docker Contrainers (tailored for their research application) that can run our own UK “cloud” resources or those of our major Cloud Partners (Amazon, Google, etc.) as resources, priorities and policies dictate.