Liqid, provider of the world's most-comprehensive composable disaggregated infrastructure (CDI) platform, announced today a new contract with the Department of Defense (DoD) to provide the world's-largest composable supercomputer. The $32 million contract, shared with groundbreaking-technology provider Intel, was awarded by the United States Corps of Engineers and will be deployed at the U.S. Army Combat Capabilities Development Command's Army Research Laboratory's DoD Supercomputing Resource Center at Aberdeen Proving Ground in Maryland. The two supercomputers, named "Jean" and "Kay," provide a total of 15 petaflops of performance. They are named in recognition of Jean Bartik and Kay McNulty, two of the women responsible for developing the first public sector supercomputer in the United States, most of whom never received sufficient recognition for their invaluable contributions to computing in their lifetimes. Day-to-day operations of the two systems will be overseen by the High Performance Computing Modernization Program (HPCMP), a technology-led, innovation-focused program committed to extending high performance computing (HPC) and propelled by artificial intelligence (AI) to address some of the world's most significant challenges.

Composing Adaptive Public Sector Computing with the DoD

The integration of Liqid's Composable Infrastructure platform with the groundbreaking Intel Xeon Platinum 9200 CPUs featuring Intel DL Boost technology and further acceleration from NVIDIA A100 Tensor Core GPUs will deliver the unprecedented agility and performance required to meet the high demands of HPCMP's AI-infused workloads, especially during data ingest and preparation. The innovative Intel Xeon Scalable processor series sets a new level of platform convergence and capabilities across compute, storage, memory, network, storage-class memory technology, and hardware-based security. The feature-rich, highly versatile platform will enable the HPCMP to accelerate critical research and development and other matters of national security.

With Liqid's Composable Fabric, the HPCMP has the powerful ability to compose the exact amount of NVIDIA A100 GPU performance into the platform to meet the data performance requirements during the training phase of the AI process. Users can dynamically orchestrate any CPU to GPU ratio, along with composing other accelerators across PCI-Express, Infiniband, and Ethernet fabrics. Liqid's orchestration software dynamically composes CPUs, GPUs, NVMe SSDs, networking, and storage-class memory to create software-defined bare metal servers on demand. This enables unparalleled resource utilization to deliver previously impossible performance for AI-driven data analytics operations.

"As HPC and AI workloads continue to converge, a new class of hardware and software solutions are emerging to address the data-intensive applications. Intel Xeon Scalable Processors with built-in AI acceleration address the need for faster, more-adaptive computing platforms to support these powerful applications," said Trish Damkroger, vice president and general manager of Intel's high-performance computing organization. "We are pleased to work with the team at Liqid to deliver groundbreaking HPC innovation with the world's largest composable supercomputer, capable of meeting the evolving threats to national security, whether related to climate research, virology, national security, physics-based modeling, and other pressing needs that require cutting-edge performance."

The two composable systems are comprised of the following:

Jean

• The Liqid CDI platform
• 1,202 Intel Xeon Platinum 9200 CPUs
• 280 NVIDIA A100 GPUs
• 323 terabytes of memory
• 12.5 PB of Liqid all-flash NVMe-oF parallel file system storage
• NVIDIA Mellanox HDR 200Gb/s InfiniBand Smart Networking

Kay

• The Liqid CDI platform
• 1,010 Intel Xeon Platinum 9200 CPUs
• 76 NVIDIA A100 GPUs
• 240 terabytes of memory
• 10 PB of Liqid all-flash NVMe-oF parallel file system storage
• NVIDIA Mellanox HDR 200Gb/s InfiniBand Smart Networking

Honoring Women's Pivotal Role in Public Sector HPC

Kay McNulty, Jean Bartik, Betty Holberton, Ruth Teitebaum, Frances Spence, and Marlyn Metzer were the first programmers of the Electronic Numerical Integrator and Computer (ENIAC). The first digital public sector supercomputer in the United States, the general-purpose ENIAC system was first deployed in 1946 to aid in the war effort and ran continuously from 1947 to 1955.

Sometimes referred to dismissively as "refrigerator girls" who simply "modeled" the hardware much like an ad for a kitchen appliance, these pioneers of digital computing were largely not given their due. By naming new, innovative systems after McNulty and Bartik, the HPCMP seeks to honor their role and positive impacts in the history of computing. These women made significant contributions that codified many of the fundamentals of HPC. Future deployments will bear the names of the other women who developed the groundbreaking technologies.

"Liqid is honored to collaborate with Intel and be chosen for this historic contract with the DoD, competing against a field of Silicon Valley's most legendary technology providers, to deliver the world's largest fully composable supercomputer -- a system that will be utilized to conduct some of the world's most critical data analytics," said Sumit Puri, CEO & Cofounder, Liqid. "We look forward to working with these teams to collaborate on AI-centric HPC systems to solve the country's most pressing problems, and we are fully driven to live up to the precedents set by their namesakes, Kay McNulty and Jean Bartik."

Go to www.liqid.com to schedule a demo and discuss the latest Liqid composable solutions and services and download a free white paper on the benefits of composable infrastructure.