Edge Computing & IoT Language Exercises

An edge node is a computing resource located physically close to data sources — sensors, cameras, or industrial machines — rather than in a centralised cloud data centre. Engineers describe edge nodes as "pushing compute to the edge" to reduce latency, preserve bandwidth, and enable real-time decisions without a round trip to the cloud.

MQTT (Message Queuing Telemetry Transport) is described as a "lightweight publish-subscribe protocol" designed for constrained devices and low-bandwidth networks. Engineers say devices "publish" telemetry to topics on a broker, while subscribers "consume" those messages. Key vocabulary includes QoS levels (at most once, at least once, exactly once), retained messages, and last will and testament.

Edge computing discussions use terms like "round-trip latency," "processing latency," and "time to insight." Engineers contrast "cloud-processed" versus "locally processed" workloads, noting that edge inference can reduce latency from hundreds of milliseconds to single-digit milliseconds. Phrases like "latency-sensitive" or "latency-critical" signal workloads that cannot tolerate cloud round trips.

Device management vocabulary includes "fleet management," "over-the-air (OTA) updates," "device provisioning," and "certificate rotation." Engineers describe challenges like "managing thousands of endpoints," "remote attestation," and "zero-touch provisioning." The goal is articulated as keeping the fleet "patched and healthy" without physical access to each device.

Fog computing is an extension of edge computing that distributes processing across a hierarchy of nodes between IoT devices and the cloud — including gateways, aggregation points, and regional edge servers. Engineers use "fog layer" to describe intermediate processing nodes that aggregate, filter, and pre-process data before forwarding to the cloud, reducing upstream bandwidth.

IoT data ingestion is described with terms like "telemetry pipeline," "sensor stream," and "time-series ingestion." Engineers say sensors "emit" or "publish" readings at a given frequency (e.g., every 100ms), which are ingested by an edge gateway, optionally "down-sampled" or "filtered," and then forwarded to a time-series database or message broker like Kafka.

Edge AI vocabulary includes "model quantisation," "TinyML," "on-device inference," and "neural engine." Engineers describe deploying a "compressed model" to a microcontroller or edge GPU to run inference "at the edge" without a network call. Key phrases are "inference latency," "model footprint," and "accuracy-efficiency trade-off."

IoT security language covers "attack surface," "device identity," "mutual TLS," and "secure boot." Engineers warn about "rogue devices" joining a fleet and emphasise "certificate-based authentication" over shared secrets. Phrases like "defence in depth," "network segmentation," and "principle of least privilege" appear frequently in IoT security architecture discussions.

Real-time constraints are described using terms like "hard real-time" (missing a deadline causes system failure), "soft real-time" (occasional deadline misses are tolerable), and "deterministic scheduling." Engineers specify constraints with phrases like "maximum end-to-end latency of 10ms," "jitter budget," and "worst-case execution time (WCET)" to convey timing requirements precisely.

A digital twin is described as a "virtual representation" or "live replica" of a physical asset, continuously updated with real sensor data. Engineers say a digital twin "mirrors" the physical state of a machine, enabling predictive maintenance, simulation, and remote monitoring. Vocabulary includes "synchronisation lag," "twin fidelity," and "shadow state."