Edge computing moves processing capabilities to the devices at the network’s edge. It’s a distributed computing paradigm, as opposed to cloud computing, where computations occur in the cloud.
While the cloud is flexible and cost-effective, moving data from the edge to the cloud can strain networks, especially with the forecasted growth in IoT. By moving processing to the edge, we eliminate some of the unnecessary back and forth to the cloud.
CB Insights expects the global edge computing market to reach $6.72B by 2022, as industries begin to take advantage of edge computing efficiencies.
Why do we need edge computing in IoT?
There are 4 main reasons to move processing power to the edge.
- Cost SavingsWhen we process information on the edge, we reduce the amount of data we send to the cloud, sending only critical information. This reduces connectivity costs and extends the life of the device, as well as its battery. In a real-life project developed on Axonize, a fleet manager installed GPS tracking devices in their trucks. GPS systems are only precise to a 1-2 meter range, and the cloud application kept getting movement notifications from trucks that were standing still because of the GPS inaccuracies. When we programmed the device to only send data for movements above 20 meters, the fleet manager got sufficient tracking data and was able to reduce the cost of tracking.
- Faster, Real-Time ResponseThe time it takes data to traverse the network to the cloud and back, while even if just a few seconds, can be too slow for some applications. Your autonomous car, for example, won’t be able to wait 3 seconds to hear from the cloud whether that is a person on the road ahead, so he knows to stop immediately. Bringing it back to the present, in a recent production line customer implementation, the line manager needed an immediate alert when the station was full, so he knew to move to the next point on the line. Even a delay of a few seconds was too costly for that specific manufacturer.
- Resilience to Deteriorating Network ConditionsThere are some IoT applications that must continue to operate even if the network is down. An example of this is a door sensor that is part of your security network. If an unauthorized person is trying to open it, you want to get an alarm even if the network is down.
- PrivacyRegulations in certain countries prohibit uploads of private information or certain sensitive data to the cloud. In this case, we want to process it locally.
What kind of processing do we typically perform on the edge?
- Rule EngineIf this occurs, then do that. In our door sensor example, it could call the security guard.
- AI on the Edge DataFor example, a camera that needs to have image recognition will run an AI instance on the edge, rather than sending it back and forth to the cloud.
- AnalyticsRunning analytics on the edge is a common way to reduce costs. For example, you can send the cloud the daily temperature average, rather than every single reading.
How do I implement edge computing?
All of the major IoT providers have an edge offering, Azure IoT Edge, Amazon Greengrass, Cisco Fog Computing and Google IoT Edge. These allow you to deploy your code from the cloud to a compatible edge device, or to use a marketplace of pre-written code.
Typically implementing edge capabilities would go something like this:
- Buy a device with edge capabilities
- Connect to a cloud provider
- Write your own code or use the marketplace
We don’t want to make it sound as easy as 1-2-3, it can actually become quite complex. You don’t know how your code will impact the device, you could be trying to run heavy logic on a light device. Or you could run into device compatibility issues if you have many devices, where your code will work on one kind but not on another.
Axonize’s disruptive architecture was purposely designed to enable deployment of complete and fully customized solutions across all applications and device types in mere days. Schedule your demo today.