Edge Computing Explained: Processing Data Closer to the Source

Edge computing is a distributed computing paradigm that processes data at or near the physical location where it is generated, rather than transmitting it to a centralised data centre or public cloud region for processing. The "edge" refers to the boundary of the network — the point closest to the data source, which might be a retail store, a mining site, a hospital, a manufacturing plant, or a telecommunications tower. By performing computation locally, edge computing reduces latency, conserves bandwidth, and enables real-time decision-making that would be impossible if data had to make a round trip to a cloud region hundreds or thousands of kilometres away.

The rise of edge computing is driven by several converging trends. The explosion of IoT devices means vastly more data is being generated at the network periphery — Gartner estimates that by 2025, 75 percent of enterprise data will be created and processed outside the traditional data centre. Many emerging applications, from autonomous vehicles to industrial robotics, require sub-millisecond response times that no cloud region can deliver due to the physics of network latency. And in many Australian contexts — remote mining operations, rural healthcare facilities, offshore platforms — reliable high-bandwidth connectivity to the cloud simply is not available.

Understanding the relationship between edge, fog, and cloud computing requires thinking about a continuum rather than rigid categories. Cloud computing processes data in large, centralised data centres operated by providers like AWS, Azure, and Google Cloud. It offers virtually unlimited scale and a vast ecosystem of services but introduces network latency and bandwidth costs. Edge computing processes data at the outermost point of the network — on the device itself or on a local gateway or server within the same physical facility. It minimises latency and bandwidth usage but has limited compute and storage capacity.

Fog computing, a term popularised by Cisco, refers to an intermediate layer between the edge and the cloud. Fog nodes might be regional aggregation points — a server room in a distribution centre that collects and pre-processes data from dozens of IoT sensors before sending summarised results to the cloud. In practice, the distinction between edge and fog is often blurred, and many architects simply refer to a spectrum of "near-edge" and "far-edge" locations. The key design principle is to process data as close to the source as the use case demands and push only what is necessary to higher tiers for long-term storage, aggregation, or machine learning training.

The most compelling use cases for edge computing involve applications where even modest network latency is unacceptable. Industrial automation is a prime example: a robotic arm on a manufacturing line making quality control decisions based on camera vision needs to process images and respond within milliseconds. Sending each image to a cloud API for analysis would introduce 50-200 milliseconds of round-trip latency, far too slow for real-time control. By running the inference model on an edge GPU appliance co-located with the production line, the response time drops to single-digit milliseconds.

Autonomous vehicles represent another extreme case. A self-driving car generates approximately 4 terabytes of data per day from cameras, lidar, radar, and other sensors. Processing this data in the cloud is neither fast enough nor bandwidth-feasible. All real-time driving decisions must be made on-vehicle — the edge device in this scenario is the car itself, equipped with powerful GPU compute modules. Cloud connectivity is used only for map updates, fleet analytics, and model retraining, not for real-time driving decisions.

Edge computing requires purpose-built hardware designed for environments that lack the controlled conditions of a traditional data centre. Micro data centres — self-contained enclosures that include compute, storage, networking, power, and cooling in a single rack or cabinet — are increasingly popular for edge deployments. Vendors like Schneider Electric, Vertiv, and Rittal offer micro data centre solutions that can be deployed in retail backrooms, factory floors, or even outdoor locations. These units are designed to operate in environments with variable temperature, dust, and vibration, and they include integrated UPS systems and remote management capabilities.

Ruggedised servers from vendors like Dell, HPE, and Lenovo are built to withstand harsh conditions — extreme temperatures, humidity, shock, and vibration — making them suitable for deployment in mining operations, oil rigs, or outdoor telecommunications cabinets. For less demanding environments, compact form-factor servers and high-performance IoT gateways from vendors like Intel (NUC platform), Advantech, and Supermicro provide adequate compute power in a small, energy-efficient package. GPU-enabled edge appliances from NVIDIA (Jetson series) are popular for AI inference workloads at the edge, such as computer vision and natural language processing.

AWS Outposts extends AWS infrastructure and services to virtually any on-premises or edge location. Outposts is delivered as a fully managed rack of AWS-designed hardware installed in your data centre or facility, running the same AWS APIs, control plane, and tools (including EC2, EBS, ECS, and RDS) that you use in the cloud. This allows organisations to run workloads locally for low-latency or data residency reasons while maintaining a consistent operational model with their AWS cloud environment. Outposts is available in Australia and is particularly relevant for government, healthcare, and financial services clients with data sovereignty requirements.

Azure Stack Edge is Microsoft's equivalent offering, providing Azure services at the edge in a range of form factors from a portable ruggedised device (Azure Stack Edge Mini R) to a full rack-mount appliance. Azure Stack Edge is particularly strong for AI/ML inference at the edge, with integrated GPU support and the ability to deploy Azure IoT Edge modules, containerised workloads, and Azure Arc-managed Kubernetes clusters. For Australian organisations already invested in the Microsoft ecosystem, Azure Stack Edge offers a natural extension of their cloud strategy to edge locations.

In retail, edge computing enables real-time inventory tracking, in-store analytics, and personalised customer experiences. Smart cameras connected to an edge AI appliance can monitor shelf stock levels in real time and trigger replenishment alerts without sending video streams to the cloud — reducing bandwidth costs and ensuring the system works even if the store's internet connection drops. Point-of-sale systems that process transactions locally ensure business continuity during outages, syncing with cloud systems when connectivity is restored.

In manufacturing, edge computing powers predictive maintenance by analysing vibration, temperature, and acoustic sensor data from machinery in real time. An edge server running ML inference models can detect early signs of bearing failure or motor degradation and alert maintenance teams before a breakdown occurs. This is particularly valuable in Australian mining operations, where unplanned equipment downtime can cost hundreds of thousands of dollars per hour and where reliable cloud connectivity at remote mine sites is often unavailable.