What is edge computing?
Edge computing is the practice of processing data close to where it is produced, at the edge of the network, rather than shipping all of it to a central cloud or data center. The edge can be a gateway box in a truck, a server in a store’s back room, an industrial PC beside a production line, or the sensor itself. The central system still exists, but it receives summaries, exceptions, and selected records instead of every raw reading.
Edge computing grew with the Internet of Things (IoT). A single plant can generate millions of sensor readings an hour, and a fleet of vehicles moves through places with no reliable connection.
Why edge computing matters
Three forces push processing to the edge. The first is latency, since a machine guard or a vehicle alert must act in milliseconds and a round trip to a cloud region takes longer. The second is bandwidth, because streaming high-resolution video or vibration data from every site costs more than the insight is worth, so the edge filters first. The third is resilience and privacy: a store must keep selling during an internet outage, and a hospital may prefer that patient monitoring data never leave the ward. Edge AI, the subset that runs machine learning models on local devices, adds the ability to make judgments on site rather than only collecting data.
Edge computing in practice
A refrigerated trucking company fits each trailer with a gateway that reads temperature, door, and location sensors every few seconds. The gateway checks readings against the load’s limits, alerts the driver immediately if a door is left open or the temperature drifts, and stores the full log locally. Once an hour, or when a cellular connection is available, it uploads a summary and any exceptions to the central fleet system. Dispatch sees every excursion within minutes, the company holds a complete cold-chain record for each shipment, and cellular data use is a small fraction of what streaming every reading would require.
What to watch for
- Plan how software on hundreds of edge devices will be updated and monitored, since fleet management is the hard part.
- Decide which data is kept at the edge and which is sent back, and document the retention rules for both.
- Secure the devices physically and on the network, because an edge box in a store or a truck is far easier to reach than a data center server.
- Design for the connection being down, including how long the device can buffer and what happens when the buffer fills.