It is a model of computing infrastructure that decentralizes a portion of the architecture dedicated to data processing. In practice, an infrastructure based on edge computing is composed of a nerve center for data storage and processing, which is associated with a number of peripheral nodes that process information locally. The need to delocalize thus arises from the need to reduce the huge amount of data flowing to the Data Center and to lighten the network traffic.
IoT produces huge amounts of data, which must be processed and analyzed before being used. Edge computing brings computing services closer to the end user or data source, such as an IoT device, allowing IoT data to be collected and processed at the edge, at the location where the device is, before sending it to a data center or cloud, in order to anticipate for identifying patterns that give rise to actions, such as anomaly detection for predictive maintenance.
The goal is to keep traffic and processing local to reduce latency, but make the most of the edge's capabilities by being able to allow it more autonomy. Much of the current focus on edge computing stems from the need for IoT systems to provide disconnected or distributed capabilities for specific industries such as manufacturing or retail. It is certainly worth keeping in mind that edge computing will increasingly become a dominant factor in nearly every industry and usage type, as edge systems can be empowered by increasingly sophisticated and specialized computing resources and increased data storage capacity.
Without edge computing, IoT would be completely dependent on network connectivity and computing services offered by the cloud or datacenter, and the continuous exchange of data between IoT devices and the cloud would slow down response times, to the detriment of operational efficiency.
Edge computing also helps solve other problems, such as the network bandwidth required to exchange massive amounts of data over a slow cellular or satellite connection, plus the ability to continue working offline in the event of a network outage.