Category II: ACES - Accelerating Computing for Emerging Sciences

类别 II：ACES - 加速新兴科学的计算

• 批准号: 2112356
• 负责人: Honggao Liu
• 金额: $ 500万
• 依托单位: Texas A&M University
• 依托单位国家: 美国
• 项目类别: Cooperative Agreement
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-10-01 至 2026-09-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2112356&HistoricalAwards=false
• 关键词: Category; II; ACES; Accelerating; Computing

The ever-growing complexity of Science and Engineering (S&E) workflows and expectations of Open Science have encouraged researchers to adopt new technologies, such as containerization, virtualization and composability, that enable them to respond to an increasingly complex cyberinfrastructure (CI) landscape while producing shareable, and reproducible results. ACES (Accelerating Computing for Emerging Sciences), an innovative advanced computational prototype to be developed by Texas A&M University, tries to answer a fundamental question: how does one effectively offer a holistic computing platform that can simultaneously meet the needs of a continuum of users in diverse research communities with varying levels of computing adoption? The project will allow researchers to creatively develop new programming models and workflows that utilize these architectures while simultaneously advancing HPC (High Performance Computing) and data science projects.The ACES platform removes significant bottlenecks in advanced computing by introducing the flexibility to aggregate various components (i.e., processors, accelerators and memory) on an as-needed basis to solve problems that were previously not addressable. By letting researchers switch and run on accelerators best suited for their workflows, ACES will benefit many research and development projects in the fields of artificial intelligence and machine learning (AI/ML), cybersecurity, health population informatics, genomics and bioinformatics, human and agricultural life sciences, oil & gas simulations, de novo materials design, climate modeling, molecular dynamics, quantum computing architectures, imaging, smart and connected societies, geosciences, and quantum chemistry. Toward facilitating researcher use, ACES will offer avenues for interactive computing, portals, and cloud connectivity. ACES will support the national research community through coordination systems supported by the National Science Foundation (NSF). Finally, ACES will also leverage existing efforts that promote science and broaden participation in computing at the K-12, collegiate, and professional levels to have a transformative impact nationally by focusing on training, education and outreach. ACES activities are designed to expand the participation of traditionally underrepresented groups in computing and STEM (Science, Technology, Engineering and Mathematics), particularly at minority-serving institutions. ACES will offer fellowships to students, continue efforts to support teacher programs, and offer a number of formal and informal courses, whose materials will be offered to the national community for use free-of-charge.This project funds the development of a dynamically composable high-performance data analysis and computing platform, named ACES. AI and ML are integrated with traditional simulation and modeling approaches in the pursuit of innovation. Edge-computing and instrumental probes have pushed the need to verify, process, store, analyze, and query vast amounts of unstructured data in real time. The coupling of analytics with closely-situated data on highly-usable web-based technologies connected to a compute backend have led to a paradigm shift in expectations from research computing environments. The ACES innovative composable hardware platform helps accelerate transformative changes in research areas that can leverage novel High Bandwidth Memory (HBM) processors and accelerators for analytics and computing. ACES leverages Liqid’s composable framework via PCIe (Peripheral Component Interconnect express) Gen5 on Intel’s HBM Sapphire Rapid processors to offer a rich accelerator testbed consisting of Intel Ponte Vecchio GPUs (Graphics Processing Units), Intel FPGAs (Field Programmable Gate Arrays), NEC Vector Engines, NextSilicon co-processors, Graphcore IPUs (Intelligence Processing Units). The accelerators are coupled with Intel Optane memory and DDN Lustre storage interconnected with Mellanox NDR 400Gbps (gigabit-per-second) InfiniBand to support workflows that benefit from optimized devices. ACES will enable applications and workflows to dynamically integrate the different accelerators, memory, and in-network computing protocols to glean new insights by rapidly processing large volumes of data, and provide researchers with a unique platform to produce complex hybrid programming models that effectively supports calculations that were not feasible before.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.

科学与工程（SE）工作流程的日益复杂性和对开放科学的期望鼓励研究人员采用新技术，如容器化，虚拟化和可组合性，使他们能够响应日益复杂的网络基础设施（CI）环境，同时产生可共享和可复制的结果。ACES（Accelerating Computing for Emerging Sciences）是由德克萨斯州A M大学开发的一种创新的先进计算原型，它试图回答一个基本问题：如何有效地提供一个整体计算平台，同时满足不同研究社区中具有不同计算采用水平的用户的需求？该项目将允许研究人员创造性地开发利用这些架构的新编程模型和工作流程，同时推进HPC（高性能计算）和数据科学项目。ACES平台通过引入聚合各种组件（即，处理器、加速器和存储器），以解决以前无法解决的问题。通过让研究人员在最适合其工作流程的加速器上切换和运行，ACES将使人工智能和机器学习（AI/ML），网络安全，健康人口信息学，基因组学和生物信息学，人类和农业生命科学，&油气模拟，从头材料设计，气候建模，分子动力学，量子计算架构，成像，智能互联社会、地球科学和量子化学。为了方便研究人员使用，ACES将提供交互式计算、门户和云连接的途径。ACES将通过国家科学基金会（NSF）支持的协调系统支持国家研究界。 最后，ACES还将利用现有的努力，促进科学和扩大K-12，大学和专业水平的计算参与，通过专注于培训，教育和推广，在全国范围内产生变革性的影响。ACES的活动旨在扩大传统上代表性不足的群体在计算和STEM（科学，技术，工程和数学），特别是在少数民族服务机构的参与。ACES将为学生提供奖学金，继续努力支持教师计划，并提供一些正式和非正式的课程，其材料将提供给全国社区免费使用。该项目资助开发一个动态组合的高性能数据分析和计算平台，命名为ACES。 AI和ML与传统的仿真和建模方法相结合，以追求创新。边缘计算和工具探测器推动了对真实的实时验证、处理、存储、分析和查询大量非结构化数据的需求。分析与连接到计算后端的高度可用的基于Web的技术上的紧密定位的数据的耦合导致了对研究计算环境的期望的范式转变。ACES创新的可组合硬件平台有助于加速研究领域的变革，这些领域可以利用新型高带宽内存（HBM）处理器和加速器进行分析和计算。ACES通过PCIe（Peripheral Component Interconnect express）Gen 5在英特尔HBM Sapphire Rapid处理器上利用Liqid的可组合框架，提供由英特尔Ponte Vecchio GPU（图形处理单元），英特尔FPGA（现场可编程门阵列），NEC矢量引擎，NextSilicon协处理器，Graphcore IPU（智能处理单元）组成的丰富的加速器测试平台。这些加速器与英特尔Optane内存和DDN Lustre存储相结合，并与Mellanox NDR 400 Gbps（千兆比特每秒）InfiniBand互连，以支持从优化设备中受益的工作流程。ACES将使应用程序和工作流程能够动态集成不同的加速器，内存和网络计算协议，通过快速处理大量数据来收集新的见解，并为研究人员提供一个独特的平台，以产生复杂的混合编程模型，有效地支持以前不可行的计算。该奖项反映了NSF的法定使命，并通过评估被认为值得支持使用基金会的知识价值和更广泛的影响审查标准。

项目成果

Scaling Study of Flow Simulations on Composable Cyberinfrastructure

可组合网络基础设施流模拟的规模化研究

• DOI: 10.1145/3569951.3597565
• 发表时间: 2023
• 期刊: USA
• 作者: Mishra, Sambit;Witherden, Freddie;Chakravorty, Dhruva;Perez, Lisa;Dang, Francis
• 通讯作者: Dang, Francis

Porting AI/ML Models to Intelligence Processing Units (IPUs)

• DOI: 10.1145/3569951.3603632
• 发表时间: 2023-07
• 期刊: Practice and Experience in Advanced Research Computing
• 作者: Abhinand Nasari;Lujun Zhai;Zhenhua He;Hieu Hanh Le;S. Cui;Dhruva K. Chakravorty;Jian Tao;Honggao Liu

Benchmarking the Performance of Accelerators on National Cyberinfrastructure Resources for Artificial Intelligence / Machine Learning Workloads

• DOI: 10.1145/3491418.3530772
• 发表时间: 2022-07
• 期刊: Practice and Experience in Advanced Research Computing
• 作者: Abhinand Nasari;Hieu Hanh Le;Richard Lawrence;Zhenhua He;Xin Yang;Mario Krell;A. Tsyplikhin;M. Tatineni;Tim Cockerill;Lisa M. Perez;Dhruva K. Chakravorty;Honggao Liu

Performance of Distributed Deep Learning Workloads on a Composable Cyberinfrastructure

可组合网络基础设施上分布式深度学习工作负载的性能

• DOI: 10.1145/3569951.3593601
• 发表时间: 2023
• 期刊: USA.
• 作者: He, Zhenhua;Saluja, Aditi;Lawrence, Richard;Chakravorty, Dhruva;Dang, Francis;Perez, Lisa;Liu, Honggao
• 通讯作者: Liu, Honggao

Extending Functionalities on a Web-based Portal for Research Computing

扩展基于 Web 的研究计算门户的功能

• 发表时间: 2022
• 期刊: PEARC '22: Practice and Experience in Advanced Research Computing
• 作者: Duy Pham, Kyle Hsu