RCCI News August 2017
The Research Computing and Cyberinfrastructure group is in the process of making significant upgrades to our high performance computing resources. We are working on a new cluster, as yet unnamed, that will be based on RedHat Enterprise Linux (RHEL) 7 with an increase in internode bandwidth from 10 Gbps to 25 Gbps. As a part of this larger effort and our annual refresh of aging hardware, installation is underway of additional standard, GPU, and high-memory compute nodes, all of which will be equipped with 25 Gbps network adapters. The high-memory nodes will have 768 GB of memory (approximately 760 GB available for user jobs). Each of the new GPU nodes will have two NVIDIA Pascal P100 GPUs. We will announce the availability of these nodes in the coming weeks. Our target for beginning migration of users to the new cluster is December 1. We hope to complete the transition to the new cluster by June 1, 2018.
As with our previous migrations to new clusters, we really need the help of HPC users in testing. We’ll need to limit the first testers to just a few and then invite more as time passes and the new cluster emerges from its growing pains. If you would like to test the new cluster, please email firstname.lastname@example.org and include information about the software and specific node characteristics you would be using. At present, we have configured 34 compute nodes (approximately 540 cores) in the new cluster and these include three GPU nodes.
A Science Gateway is a web-hosted set of community-developed tools that give researchers, educators, and students easy access to specialized resources that are specific to a science or engineering discipline. Such resources are meant to reduce the computing skills threshold for researchers to apply advanced computing infrastructure to research problems. Through Science Gateways, researchers concentrate on the questions, while the science gateway interface serves to organize computing with data, models and methods to expand research capabilities and improve reproducibility of analyses.
Science Gateways are distinguished from web portals by the integration of high-performance computing capabilities, including application of machine learning and other data science techniques, available from a browser-interface. Science Gateways enable access to shared data, software, instruments, and other discipline-specific resources. For example, they may connect to or between instruments (such as weather sensors or other instruments such as telescopes), data collections or specialized software. Researchers may also participate as collaborators, and make contributions to feed back into the community, through providing unique data sets or enhancements to the methods employed in the Science Gateway itself.
Science Gateways are sponsored and hosted by funding agencies such as the National Science Foundation (NSF), as well as individual research groups and universities. Examples of institutionally supported Science Gateways include the NeuroScience Gateway (www.nsgportal.org) and the Materials Project (https://www.materialsproject.org/). XSEDE (Extreme Science and Engineering Discovery Environment) supported Science Gateways span about two dozen fields of science are cataloged at: https://www.xsede.org/web/guest/gateways-listing.
In support of the CWRU research community, for those who may want to establish their own Science Gateway, RCCI participates in the Science Gateway Community Institute (SGCI), a funded 5-year NSF grant to support Science Gateway development. In its second year, the aim of the SGCI (https://sciencegateways.org/) is to serve the science gateway community with both online and in-person resources and services. SGCI provides support to projects through dedicated developers and staff who may partner with research group members and university staff. SGCI has evolved from experiences developing projects within XSEDE, and has experience with many Science Gateway platforms and other tools, so that you can focus on the features that are unique to your user community.
If you are interested in this topic please contact RCCI for more information. We are planning a workshop for Fall semester on the use of web-hosted resources such as a Science Gateway for research involving large-scale computation and complex problem solving.
Title: HPC Bootcamp
Time: 9:00 A.M. - 12:00 P.M.
Date: Thursday, September 14, 2017
Location: Toepfer Room
The RCCI team will offer a High-Performance Computing Bootcamp on Thursday, September 14th. This event is geared towards new users to the HPC cluster, presenting introductory materials in lecture with demonstrations. Our goal is to provide a grounding in the concepts necessary to conduct work in the HPC environment, and to make clear the methods for connecting to the HPC from personal and lab computers. The event will take place between 9 a.m. and noon in the Toepfer Room of Adelbert Hall. Please register at the link provided.