Project Proposed: This project, acquiring a hybrid computing cluster for high-end applications in science and engineering, services, among others, bioinformatics, ecological modeling, and physics. GPU-based computing support for distributed memory parallel applications constitutes a specific novel feature of the cluster. The work aims to support the following activities: - Modeling genomes, hyper-extractive economies, and ecological forecasting (by bringing much greater computational power) and, - Enabling new science utilizing the instrument to develop new algorithms in physics modeling and genomics. Broader Impacts: Mainly, the multidisciplinary nature of the proposed research provides the broader impacts. Expected are large broader impacts on basic physics research. It is expected to also impact on medicine through the development of better molecular models to view how our proteins and membranes interact. Planned are training activities of a new generation of researchers in tools and techniques for high-performance computing. The PIs would build on the broad past experience in preparing widely-used undergraduate and graduate educational materials, allowing students to perform real-world projects and impacting K-12 and STEM education. Hence, the project aims to significantly enhance and integrate our educational efforts at the K-12, undergraduate and graduate levels in bioinformatics.
(particularly bioinformatics, ecological modeling, and physics), targeting research and research training for Kansas State University and the state of Kansas. Our goal was to dramatically expand existing research activities in a number of disciplines and multidisciplinary areas. The system provides faculty and students with a critical resource to expand research training, foster the development of the next generation of research opportunities and exploration in several science and engineering departments, provide an "on ramp" to XSEDE, and promote partnerships between researchers both internally and externally to K-State. A major activities over the course of the event have been the acquisition of multiple systems to effectively create our cluster as a viable computing tool. Our initial purchases included 64 dual-socket Xeon e5-2690 R620 compute nodes from Dell, 6 1TB RAM/80-core X3850 X5 servers from IBM, and a 540TB parallel filesystem with two fileservers from IBM. We also purchased sufficient QLogic QDR Infiniband cards and switches for fast connectivity. During the project, we have more than doubled the number of computing cores available (over 2,000), and our number of users went up three-fold (over 750 now, including 125 faculty members). On the research front, our faculty regularly make comments like: - The KSU Beocat computing cluster was critical for the success of our research project aimed at analyzing a large number of wheat accessions. - The work made use of Beocat, and without a doubt, the hardware, software, and support made the usual trial-and-error of a new graduate student go from O(months) to O(weeks). - We have multiple externally funded projects with national and international impact that rely on Beocat as a critical resource… leveraging the power of Beocat for storage and analysis of over 10TB of genomic sequence data that is being used to develop whole-genome prediction models for wheat. The Innovation Lab aims to develop heat-tolerant, high-yielding, and farmer-accepted varieties for South Asia. - We achieved 50x speedup in an environmental modeling and simulation framework on Beocat and compatibility with XSEDE allowed us to utilize both resources. - It would not be possible for us to work on any of [our] projects … without access to Beocat. As of our outreach and workforce training, we estimate we have given tours to over 1,500 people in the past year alone! Examples of our activities include: - Beocat training sessions, which typically have 30-50 attendees. We record the sessions and make them available through our YouTube channel (www.youtube.com/channel/UCfRY7ZCiAf-EzEqJXEXcPrw). Our demonstration of using MPI on Beocat has over 2,200 views to date. We have also presented to the KSU Bioinformatics journal club, helped teach various graduate-level bioinformatics courses, and presented in dozens of classes. We also present more focused sessions to individual labs. - A particularly fun outreach/training activity has been our work with the summer GROWE/KAWSE camps for junior girls, where we have been working with 30 girls per camp, and using a wind simulation written in MIT's Scratch to teach parallel programming concepts. We begin with a talk on HPC and parallel computing, give a Beocat tour, then start the exercise. The reaction has been phenomenally positive from the girls. We also trained two groups of high-school teachers participating in our summer GK12 institutes to use these lessons in their own classrooms. - We have also introduced parallel computing concepts into our freshmen computing concepts course, which has about 400 students per year. Our philosophy has been to introduce them to parallel thinking and the limits of parallel computing early in their programming career, and hopefully make this a consideration as they design algorithms in industry and academia. - Another outreach/training effort has been through STARBASE, a program for middle school students in the Manhattan, KS area (USD383), where we introduce them to programming in Scratch, and give them a brief talk and tour about supercomputing. This has involved 600 students so far this year. Both students and teachers have been enthusiastic about supercomputing and their introduction to large-scale computing similar to that used by companies they already know, like Google and Facebook. Beocat has had a massive, statewide impact. Academic users in the state of Kansas are allowed to use Beocat for free – including researchers and students from Bethany College, Benedictine College, Pittsburg State University, and Washburn University – as well as supporting thousands of students, hundreds of papers, and dozens of research projects within KSU. Researchers have contributed their own resources to augment those purchased through the MRI, doubling our computing power again over the last few years. Broader impacts include thousands of people introduced to supercomputing, increasing the safety and security of our food supply, and helping enhance the economic security of agribusiness.
