New paths for edge devices: distributed intelligence
When you think of how to collect and store data traditionally, you might envision a data center with big compute racks. Think bulky form factors. But your form factors don’t always need to have a large footprint to embrace intelligent edge. It all depends on your business use case. You may be able to distribute your form factors, leveraging smaller hardware with an aggregation point at the edge itself, and that can have a bigger compute component.
Remember: An edge device is any piece of hardware that controls and processes the data and flow between where the data is generated and the enterprise network.
It’s an empowering definition, because you can still run your software on a very small device like an Intel NUC, Lenovo, HPE or Dell Gateway, all of which could run on a CPU — then take your stack and move it to a bigger compute. That compute could have dedicated GPUs, IGPUs or multi-processors.
Businesses can even use their existing hardware and bring that intelligence closer to it. In manufacturing and retail, we’re seeing the use of existing (not specialty) cameras doing analysis of how space is being used. You have options. It’s just a matter of knowing what those options are (which many businesses today do not) and determining which works best for your use case and objectives.
Navigating edge applications
Of course, we can’t look at edge devices in a vacuum. The applications that power them are equally important. Today, there are software stacks that are built with specialization. For instance, software stacks can be given an AI model, and that stack will know exactly how to run on a GPU, an IGPU or an FGPA. The AI model can even be accelerated with a deep learning toolkit like Intel OpenVINO or NVIDIA SDKs.
This flexibility is a gamechanger. Form factors will keep changing at the edge, but you won’t need to keep developing separate, “snowflake” solutions for every new hardware that comes to the forefront.