A Department of Defense client taps Sirius Federal for a high-performance cloud solution.

A Department of Defense client and Colsa Corporation, an innovative provider of advanced systems and software engineering solutions to defense, intelligence and commercial clients.

A Department of Defense client needed a solution to improve the speed and responsiveness of computational workloads so they could quickly gain the most valuable and reliable insight from their data. They were looking for something that scaled with their mission and operated effectively and efficiently as they mined the data needed to make mission critical decisions. Performance was tightly linked to time, value and stakeholder satisfaction.

The client’s developers started their artificial intelligence (AI) journey without an IT-led strategy that offered solutions and that did not scale efficiently. The client needed a revolutionary new approach to delivering infrastructure quickly and easily with built-in AI tools, all without disrupting their day-to-day operations.

The Sirius Federal team partnered with Colsa Corporation, the client’s systems integrator, to ensure the client’s technology teams fully understood their current operating conditions. Sirius Federal’s experienced engineers worked with them to develop the exact requirements they needed to transform their infrastructure.
Through this discovery phase, Sirius Federal suggested a validated reference architecture as an all-inclusive strategy to improve the client’s infrastructure using a best-of-breed, white glove installation approach to deliver AI supercomputing with built-in storage capacity that scaled to enhance and simplify their daily mission operations.

Sirius Federal assembled the turnkey solution at one of our state-of-the-art integration facilities. Once configured and running, the client was able to perform remote benchmark tests on their new AI infrastructure to verify quality and inter-operability prior to the supercomputer being delivered to their secure facility. Sirius Federal’s ability to assemble and stage the fully functioning system and then enable the client to access it remotely to test and validate was a critical advantage in this project, one that gave the client the highest confidence in the solution.

The unique delivery approach dramatically reduced the time the client needed to wait before their new system entered production workloads from 12- 18 months to 90-120 days. When intelligence is on the line, saving a year in development is critical.

The fully configured and integrated system rapidly deploys operational resources that easily integrate with the client’s existing Cisco ACI fabric, while also delivering exceptional, predictable and reliable performance for a broad range of HPC workloads. The highly flexible, purpose-built infrastructure easily scales to support future AI workloads and eclipses legacy data center capabilities.

Sirius Federal leveraged multiple relationships and partnerships to ensure a truly customized and innovative AI solution for the DoD client — from the systems integrator and vendors to engineering software developer teams.

The state-of-the-art SuperPOD solution enabled the client to rapidly deploy revolutionary new infrastructure for AI workloads immediately upon delivery. It was a true “roll it down the hall, plug it into their data center, and start working” turnkey solution. The entire installation, configuration and deployment processes was seamless to ensure business continuity for the client’s mission.

The supercomputer is an evolution in supercomputing infrastructure, providing massively scalable data center infrastructure that delivers up to 40 PETA FLOPS of computing power. This marks a major milestone in the evolution of Sirius Federal’s supercomputing solution capabilities, offering massive computing power to propel the client into a new era of advanced computing.

Sirius Federal and Colsa Corporation proposed an AI supercomputer based on NVIDIA DGX A100 SuperPOD reference architecture. The fully autonomous cluster is powered by NetApp EF600 NVMe storage, Mellanox Quantum HDR InfiniBand, and managed using Bright Computing cluster manager software.
• NVIDIA DGX A100 systems
• NetApp EF600 NVMe storage
• Mellanox Quantum HDR fabrics
• HPE servers
• IBM spectrum scale
• APC equipment racks
• Bright computing software
• Cisco ethernet fabric