Description of Beocat for Proposals or Information
Below is a current description of Beocat Compute resources and availability that can be used within proposals or to provide information. If you have any questions regarding this, please reach out to us at beocat@cs.ksu.edu
Compute Resources Description
Beocat, the K-State research computing cluster, is currently the largest academic supercomputer in Kansas. Its hardware includes nearly 400 researcher-funded computers, approximately 3.3PB of storage, ~10,000 processor cores on machines ranging from dual-processor Xeon e5 nodes with 128GB RAM and 100GbE to 128 core AMD nodes with 2TB RAM connected by 40-100 Gbps networks, and a total of 170 different GPUs ranging from GTX 1080ti's to NVIDIA L40S's. Beocat and its staff have provided tours demonstrating the value of K-State research and a high-tech look at our research facilities for over 3,000 participants, including USD383 StarBase, current and prospective students, funding agencies, faculty recruitment, and outreach activities. Classes supported include topics such as bioinformatics, business analytics, cybersecurity, data science, deep learning, economics, chemistry, and genetics. Beocat is supported by many NSF and university grants, and it acts as the central computing resource for multiple departments across campus. Beocat staff includes one full-time system administrators, a full-time applications scientist with a PhD in Physics and 35 years’ experience optimizing parallel programs and assisting researchers, and a part-time director.
Beocat is available to any academic researcher in Kansas and their partners under the statewide KanShare MOU. Under current policy, heavy users are expected to buy in through adding computational or personnel resources for the cluster (condo computing). Their jobs, then, are given guaranteed priority on any contributed machines, and they have access to other resources in the cluster on an as-available basis. Thus, projects can preserve a guaranteed base level of computation while utilizing the larger cluster for major computations. Users can also purchase archival data storage as needed. Dr. Daniel Andresen is the K-State XSEDE Campus Champion in the event national-class computational resources are required.
