If you’re in the IoT space, you’ve no doubt heard the terms “fog computing” and “edge computing” being thrown around a lot recently. What exactly are fog and edge computing, and what do they have to do with the IoT?
Fog computing and edge computing are not just buzzwords, but real system designs being used today in IoT deployments. As the IoT makes its way into businesses across the world, many industries are finding that fog and edge computing can fill in the gaps that traditional cloud-centered IoT design misses.
Traditional IoT Architecture
In the traditional IoT model, large numbers of low power sensors and actuators interface with the physical world. These devices are small, low power, wireless, and either output or take as input, small amounts of state information, such as temperature, pressure or humidity.
Those data are gathered by a gateway device and delivered to the cloud for processing. The analyzed data are then used to guide business processes or used to interact with the physical world through actuator devices.
This architecture is simple to understand and makes sense in theory, but in practice, this style of deployment has been very limited. Most industries do not have the need, the resources, or the technical expertise to deploy the hundreds, thousands, or hundreds of thousands of low powered devices necessary for these deployments to be effective.
Businesses that have experimented with this model have found that the raw data delivered by simple, low powered devices can be overwhelming, using up lots of network bandwidth, with a low signal to noise ratio.
In fog computing, data processing happens not only in the cloud data centers but at the local area network layer. In the strictest sense of fog computing, a gateway device will not only route data to the cloud but perform analysis, compression and filtering to reduce the amount of data sent, making more efficient use of network bandwidth. End devices may still be simple nodes like the traditional cloud-centered IoT model, but only usable data are sent to the cloud, while the rest are filtered out.
Besides reducing upstream network usage, fog computing also allows for more immediate feedback - essential for manufacturing, oil and gas, and many other applications that interact with physical processes. Data are not only sent to the cloud but also analyzed and used at the LAN layer to drive real-time processes.
Compared to the traditional cloud-based IoT model, fog computing provides better network latency for real-time processes as the processing happens on the local area network. It also makes better use of cloud computing network bandwidth and processing capabilities and is thus becoming an increasingly popular way to deploy IoT.
Edge computing and fog computing are sometimes used interchangeably, as they both push data processing away from the data center; however, in a stricter sense, they are different system architectures. While fog computing focuses on data processing at the gateway and LAN level, edge computing pushes that data processing all the way to the network edge.
In a sense, consumer electronics, which are relatively powerful devices that interface with us, can be said to be a form of edge computing. As we use smartphone apps, access our Gmail accounts, or check our Facebook feeds, we’re in a sense performing edge computing. We are sending and requesting useful data that are processed on our own device and sent in a compressed form over the internet to cloud data centers.
However, while consumer cloud does reflect edge computing from a network architecture perspective, the term edge computing generally refers to industrial applications. Adding intelligence to devices like smart valves, automation controllers, or IP cameras, and connecting them to the internet, allows valuable real-time processing such as pressure control, manufacturing processes, or facial recognition to happen, while valuable, and relevant business data are still passed through to the cloud.
Edge computing in its strictest sense - where processing power happens at the edge - also has the benefit of having the most straightforward network architecture. Rather than data from edge devices being gathered and processed at a central gateway device, edge devices process what they need and send the rest directly to the cloud.
Finally, edge computing often has lower device counts, often not much more than the company is already using, which helps make it a popular and practical IoT deployment approach.
As businesses experiment with the IoT, many are discovering that the idealized, cloud-centric models of what IoT computing is, aren’t a good fit for their technical or business requirements. Instead, pushing data processing closer to the network edge can improve the use of network bandwidth and cloud data processing power.
Edge and fog computing provide two system architectures that reduce network latency, allow for real-time feedback, and often are much more straightforward for today’s businesses to deploy than traditional cloud-centric IoT architecture.