Mercury’s New Rugged Distributed Processing Solutions Tackle the Most Challenging AI Workloads at the Edge
First rugged network-attached GPU edge server to incorporate NVIDIA BlueField DPU technology
Mercury Systems, Inc., a leader in trusted, secure mission-critical technologies for aerospace and defense, announced the new rugged distributed processing (RDP) rackmount server series, revolutionary data center-class servers designed to deliver GPU parallel computing resources over high-speed Ethernet networks. These high-performance computing (HPC) servers are optimized for size, weight, and power (SWaP)- constrained, compute-intensive, low-latency workloads at the edge such as sensor processing, artificial intelligence (AI) and data analytics.
Accelerated by NVIDIA® BlueField® data processing unit (DPU) technology, the new servers deliver powerful GPU computing resources to distributed sensors or clients connected over the network. With a high-speed PCIe Gen 4 fabric, multiple 100G Ethernet interfaces and integrated management processor, RDP servers enable GPU processing without the SWaP and cost disadvantages of a separate server to host and manage the GPU.
Recommended AI News: Rackspace Technology Awarded Amazon QuickSight Service Delivery Designation for AWS
Why It Matters:
The latest in data center-class compute performance is now readily available for aerospace and defense application deployment, particularly in space-constrained platforms such as aircraft, ships, submarines, and ground radar stations. This new technology dramatically accelerates AI for threat assessment, augmenting mission-critical decision making, and ultimately making those platforms and their crews safer.
“Customers are seeking the highest-performance solutions packaged in rugged and SWaP-reduced form factors to enable compute-intensive AI and deep learning applications at the edge,” said Dusty Kramer, vice president and general manager, Mercury Edge. “Our new purpose-built RDP server series is the first in the industry to pair NVIDIA DPU and GPU technologies for use in vehicle, shipboard and aircraft applications at the edge. It’s another example of how we collaborate with key Silicon Valley tech companies to make commercial technology profoundly more accessible to aerospace and defense.”
Typically, GPU management is performed by a separate host computer, increasing cost and complexity. Mercury’s RDP server, on the other hand, incorporates all GPU and DPU components into a ruggedized 1U chassis, making GPU resources accessible anywhere. RDP 1U servers can also be installed as a companion unit with other Mercury servers and data storage solutions for enhanced computing power, while multiple RDP systems can be deployed as a GPU cluster for massive data processing tasks. For embedded VPX system applications, Mercury offers a module with dual NVIDIA Turing GPUs. Future RDP products may support additional NVIDIA DPU and GPU modules.
Recommended AI News: Backblaze Doubles Down on Security With Open Bug Bounty Program
“Innovations in breakthrough technologies such as AI, 5G and simulated environments are supercharged by next-generation data center infrastructure,” said Kevin Deierling, senior vice president of networking at NVIDIA. “NVIDIA’s accelerated computing platform equips innovators with extreme performance as well as the security and SWaP-optimized efficiency to push the boundaries of state-of-the-art technologies, making it the ideal choice for Mercury to integrate into their solutions.”
Powerful Edge AI Capabilities with a Small, Ruggedized Footprint
- NVIDIA A100 Tensor Core GPU and BlueField-2 DPU technology in a standard 20” depth, 1U ruggedized form factor with rear I/O connectors
- Integrated host management using BlueField-2 Arm-core processors running Linux
- Low-latency PCIe Gen 4 backplane to dual 100G Ethernet ports for high-speed data transfer
- Eliminates need for x86 host server for management, reducing size, weight, and power footprint
Recommended AI News: AE Studio and Edge Of NFT Launch Edge of AE Studio at NFT LA
[To share your insights with us, please write to sghosh@martechseries.com]
Comments are closed.