CC* Planning: High-Performance GPU Cluster for Computational Intensive Interdisciplinary Research

University Of California-Los Angeles, Los Angeles CA

Investigators

Abstract

With recent technical advances in research methodologies, there is an ever-increasing generation of large datasets in biomedical research. This coincides with advances in analytical methodologies such as deep learning techniques. These analytical tools heavily utilize the power of graphical processing units (GPUs) as the main computational hardware. However, the current infrastructure of many research institutes is not sufficient to provide solutions for this kind of special computational needs of the biomedical fields. This results in attempts by individual research laboratories to solve these computational needs temporarily, slowing down the research and putting the difficult-to-obtain data at risk of loss. To address these issues and find a more efficient and sustainable solution, this project aims to systematically go through several steps at different levels, create solutions that will expedite the biomedical research and provide an institutional example for setting up infrastructure to meet these computational needs for biomedical research institutes. More specifically, the goal is to create a detailed plan of such an infrastructure. These include: (1) identification of science use cases and representative users, (2) determining the standards for computing, networking, storage and security of such a system, (3) creating a high-performance GPU architecture plan and (4) developing a support model to maintain and improve this infrastructure. This project aims to create a model that is efficient, useful, sustainable and that will expedite the data processing and analytics in biomedical research. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.