AI/HPC Edge Reference Architectures
Accelerate workloads despite form factor and field limitations
Edge AI/HPC Reference Architectures
The plethora of data available today – and the ever-increasing rate of new data being created – and improvements in DevOps has led to the creation of many powerful new algorithms to extract more value from data.
But implementing HPC and AI on the edge requires a strong understanding of both computing and form factor. That’s why Silicon Mechanics created a series of reference architectures for specific types of edge deployment and workloads.
Each one is the result of hours of engineering, testing, and optimization for power, latency, and related concerns of resource-hungry applications as well as space, size, ruggedization, and other issues facing in-field deployments.
That’s because we want to save clients time and focus on customizing the design for your specific workload and organizational needs –not redoing the basic elements with each new engagement.
Each reflects all our past work designing edge devices to meet the unique demands AI and HPC places on hardware. And they are a great starting place for your customized edge deployment.
Learn more about our edge HPC and AI reference architectures:
Argos Ruggedized Edge System
Devices that support AI on the edge (i.e., such as in-vehicle or MIL-SPEC systems) are complex to design and usually only support a specific, brand-name edge cluster. The Ruggedized Edge System includes commodity hardware nodes designed to resist environmental conditions and powerful enough to support accelerated computing, without the costs of brand-name devices.View
Some organizations need cloud-like services (i.e., storage, compute, containers, GPU) in edge environments. But brand-name solutions like the AWS Snowball are expensive and have vendor-specific configurations. The customizable Edge System combines ruggedized, enterprise-grade components and a private cloud software stack suitable, all designed for use in the field, for the same performance but lower cost than Snowball.View
CDI Edge Cluster
Running a cluster outside of a datacenter can limit the workloads you can support. You have limited footprint, environmental concerns (vibration, heat, dust, moisture), power envelope limitations and other issues. The CDI Edge Cluster gives you the power to reconfigure resources on the edge and still get bare-metal performance. At the same time, it addresses form factor and deployment limitations you find at the edge.View
HCI Edge Cluster
A secure enclave/cloud environment is complex to design. Putting that capability at the edge makes it even more challenging and costly. The Edge HCI Cluster solves the problem with a dense, scalable central cloud with a smaller, localized cloud-in-a-box at the edge, all using commodity components to keep costs down.View