// 07 — iot & edge data

IoT & Edge Data — Roadmap

33 topics across devices & sensors, connectivity, edge computing, ingestion, time-series analytics, and security & operations.
Pairs with: Data Engineering
Total
33
Done
0
Progress
0%
Left
33
Est. time
~14 h

Devices & sensors

devices 0/5 ▾

Sensors, actuators & signal basics

Analog vs digital signals, sampling rate, Nyquist, noise and calibration drift.

Microcontrollers vs single-board computers

ESP32 vs Raspberry Pi: power, cost, OS vs bare metal, when each one fits.

Industrial I/O — PLCs, 4–20 mA, Modbus

How plants already expose data. Reading registers, polling vs events, legacy integration.

OPC UA

Information models, client/server vs pub/sub. The interoperability standard of industrial automation.

Power & battery budgeting

Duty cycles, deep sleep, wake-on-event. Estimating battery life before deploying 500 nodes.

Connectivity & protocols

connectivity 0/6 ▾

LoRaWAN

Long range, low power. Gateways, network server, spreading factor and duty-cycle limits.

Cellular IoT — NB-IoT & LTE-M

Coverage, power profile and cost per device compared with LoRaWAN.

Satellite IoT for remote sites

When there is no terrestrial coverage: latency, message size and cost trade-offs.

CoAP, HTTP & when not to use MQTT

Request/response for constrained devices. Choosing the protocol by traffic pattern.

Payload formats — JSON, Protobuf, Sparkplug B

Bytes matter on constrained links. Schemas and state management for industrial MQTT.

Edge computing

edge 0/6 ▾

Edge vs cloud

Latency, bandwidth, cost and autonomy. What must run on site and what can wait for the cloud.

Store-and-forward & intermittent connectivity

Local buffering, at-least-once delivery, idempotent writes and deduplication downstream.

Edge gateways — Node-RED, Telegraf

Protocol translation, filtering and aggregation before data leaves the site.

Containers at the edge — k3s

Lightweight Kubernetes on gateways. Deploying the same pipeline to many sites.

OTA updates & fleet management

Signed firmware, staged rollouts, rollback. Never brick a device you can't reach.

TinyML — inference on the device

Running small models on microcontrollers: keyword spotting, vibration classification.

Ingestion & streaming

ingestion 0/5 ▾

MQTT brokers — Mosquitto, EMQX

Single broker vs cluster, persistence, auth and bridging between sites.

Cloud IoT platforms — AWS IoT Core, Azure IoT Hub

Managed device registry, rules engine, routing into the data platform.

Bridging MQTT into Kafka

Decoupling device traffic from analytics. Topic mapping, keys and ordering per device.

Late, out-of-order & duplicated events

Event time vs processing time, watermarks, devices with wrong clocks.

Device identity & provisioning

Per-device X.509 certificates, just-in-time provisioning, revoking a stolen device.

Time-series & analytics

timeseries 0/7 ▾

Time-series data modeling

Tags vs fields, wide vs narrow tables, cardinality explosions.

Downsampling, retention & rollups

Raw for days, aggregates for years. Continuous aggregates and tiered storage.

Resampling, interpolation & gaps

Irregular sampling, forward-fill vs interpolation, marking missing data explicitly.

Anomaly detection on sensor data

Thresholds, rolling statistics, isolation forest. Fewer false alarms beats more alarms.

Predictive maintenance

Vibration features, remaining useful life, labeling failures you rarely see.

Dashboards & alerting — Grafana

Operational dashboards on time-series, alert rules and on-call routing.

Security & operations

security 0/4 ▾

IoT security basics

Default credentials, insecure updates, exposed services. The usual ways devices get owned.

OT/IT segmentation — IEC 62443

Zones and conduits, keeping plant networks off the internet.

Monitoring device health

Heartbeats, battery, signal strength (RSSI/SNR), silent devices as an alert.

Digital twins — beyond the buzzword

State model of an asset fed by telemetry. What is real value and what is marketing.