Table of Contents
FogHorn aims at providing an edge computing platform as close to the data generation point as possible. Their platform, called FogHorn Lightning™ Edge AI, is intended to process live data in real-time and publish insights to public or private clouds or an on-premises data center. The platform can generate real-time alerts to operators for the derived actionable insights. Closed loop control is also supported through a polyglot SDK. The Lightning Edge AI platform is managed by the FogHorn Manager, which can run in the cloud or a data center.
FogHorn Manager provides the orchestration and management capabilities, including configuration, deployment and security protections, that govern the functionality of Lightning Edge AI deployments.
FogHorn does not produce its hardware to run FogHorn Lightning Edge AI, the software running on an edge device managed by FogHorn Manager. They have partnerships with multiple hardware vendors to pre-load the software on their systems. They can also support existing devices on various form factors, including programmable logic controllers (PLCs), Raspberry Pis, and Edge Gateways using Intel, Arm or NVIDIA architectures. All of FogHorn’s software is containerized, providing a modular deployment model.
Based on the taxonomy established in the Key Criteria for Evaluating Edge Infrastructure report, FogHorn’s solution falls under the category of Platform Systems.
Figure 1: Edge Applications Depend Upon Physical, Core & Platform Systems
FogHorn Lightning Edge AI processes streaming data in memory through multiple stages (figure 2):
Figure 2: FogHorn Lightning Edge AI Process
The data leaving the edge device can be sent as an alert to operators, control signals to other devices in the environment, or the public and private cloud.
FogHorn uses many common protocols to communicate with devices downstream, including Modbus, OPC-UA/DA, and Ethernet, to name a few. FogHorn will utilize the existing network infrastructure on the hardware for upstream communication. Upstream communication always initiates using an outbound connection instead of FogHorn Manager making inbound calls.