Thanks to modern communications infrastructure we can bring the compute power from the likes of Amazon, Microsoft, and Google to just about any point in the developed world. But what do your customers do when they need data-heavy analytics in remote, rural, or undeveloped locations? They can’t always afford to build a data center where they need them. The poor quality of infrastructure in these regions prevents the use of high bandwidth solutions. But there must be a solution, right?
Think of an on-premise data center as the ground and a remote data center as the cloud. Now there needs to be an intermediate layer, something on the Edge where the data resides. Thinner than a cloud but looks the same up close, this is where Fog computing fits in.
Fog computing, or Edge processing, is the answer for more scenarios than just rural and remote locations. This newly defined compute layer can lower costs in any application where data communications is metered. But what is it? It’s simple really. Fog computing is adding a compute resource, usually at the point of demarcation. This is where your customers’ gear connects to their internet provider’s network. The secret sauce in Fog computing is to filter, analyze, compress, or pre-process their data before it hits that metered, slow, or unreliable connection. Of course we’ve always done this when needed in enterprise data centers with remote and branch offices. What’s new is all the new and powerful ways that it can be done in the field.
The big guys in networking have allowed us to run a VM on their switches for a while now. This is an ideal place for Edge processing or Fog computing as some vendors call it, because a router is the literal device that connects your customers’ networks to their service provider’s network or point of demarcation. This VM can be an IoT gateway, firewall, or a custom number cruncher developed for a specific solution.
Thanks to new capabilities in hardware, we can add a neural network to any compute resource via a USB stick. Movidius, a recent acquisition by Intel, developed and is now shipping a self-contained deep learning AI accelerator in a USB stick form factor for less than $80. Now we can do complex behavior-based processing, such as 100 GFLOPs locally instead of relying on a one-bar cell connection to the cloud. As for power requirements, the Movidius Neural Compute Stick (NVS) only draws one Watt of power. This product is for developers and will take some time to make an impact. If you want to learn more about Neural Compute Sticks, ask your Tech Data representative to contact the Tech Data Intel team.
The Universal IoT Gateway (UIG) is another offering from Cisco that’s being combined with HD cameras from the likes of Axis Communications and ruggedized servers from any of our OEMs to bring to market some incredible security solutions. Built around two-way communications, these remote surveillance units can be solar or wind powered "crime prevention" security systems utilizing real-time video analytics of vehicular and human traffic. The UIG can be attached to just about anything from the side of a building to light poles to add security anywhere it's needed.
When your customers are using these Edge devices, they’re keeping the computations close to the data-saving bandwidth, reducing latency, and in many cases, saving money. Keeping the data on premise also reduces their online exposure to internet-based threats. Ideally the only data, if any, that needs to be transmitted are dashboard updates and alerts.
Smart IoT Solutions by Tech Data was created to help our customers and vendors make IoT simple, so they can seize IoT opportunities much faster. Contact the Tech Data Smart IoT Solutions team at email@example.com for more information on any of the solutions mentioned above, or to set up a consultation.