The Edge and The Cloud for Industrial Applications
It keeps getting cheaper and easier to fit devices with sensors. But sensors without analytics to turn their data into understanding and decisions are of limited use.
ABI Research finds that in 2014, IoT-connected devices captured 233 exabytes of data, and only managed to transmit less than 10% of it. This data load will only get greater.
Data or Insights?
Data is only as useful as the decisions it lets you make. Sometimes the decisions are small, local, and immediate, like detecting an open valve and closing it. Sometimes they are large, wide-ranging, and have a long time horizon, like investing in a new manufacturing facility.
Insisting on processing all the data centrally is:
- Slow: in fast-changing situations the latency from send to receive can result in instructions outdated before they are received.
- Expensive: those exabytes of data require bandwidth and storage
- Insecure: all the data is vulnerable in motion. Local processing will often be more secure.
Industrial Use Cases
These considerations of cost, latency, and security have led to a lot of activity around increasing the analytical abilities of the edge—at the sensors and devices. But how much capability to put at the edge, and how much to retain centrally, in the cloud? Different industries face different challenges and requirements—and must make weigh considerations according to these requirements.
Wind turbines are located in remote areas. Being able to respond immediately to any change in wind direction, velocity, or turbulence increases power generation efficiency significantly.
Oil and Gas
Modern oil rigs generate vast amounts of time-sensitive data, over 1 TB a day. Malfunction and failure can occur quickly and unexpectedly, resulting in dangerous working conditions, expensive downtime, and environmental damage if not responded to promptly. It makes sense to do a significant amount of analysis onboard for quick decisions that minimize and contain problems.
Edge computing is most important when decisions need to be fast, and the consequences of a slow decision are significant. Self-driving cars and trucks, with their need for split second responses to protect their passengers, and the slightly longer decisions about fuel use and pollution control, will be a key area of edge analytics.
Each Edge Reports to a Cloud
Of course, cleaned and edited data goes back to the cloud to make larger-scale business decisions in each of these cases. The wind farm makes business decisions about overall power generation and equipment maintenance, the oilfield operator makes production and utilization decisions, and the traffic control system makes decisions about congestion and light timing.
What to Keep, What to Send
What will result are various hierarchies of analytics and decision. The edge will mostly execute pre-built analytic models, using as little memory and processing as possible. Gateways linking various sensors will monitor wider areas, while the cloud will analyze and update the models used by the edge, as well as doing the deep analysis of large amounts of data over time.
As processing speeds, transmission capacities, deployment costs, and business requirements change, the balance of edge and cloud will change as well.