When working on an IoT application, the first architecture decision to make is:
where to process the data?
The standard answer is "edge or cloud."
That's not a very interesting distinction.
A much better approach is to classify each workload according to its latency, connectivity needs, data volume, reliability requirements, and CPU/memory needs.
This article is about a practical framework for determining where an IoT workload should be processed.
Start With the Data Flow
An industrial grade IoT system will often resemble this:
Sensors / Devices
↓
Local Gateway
↓
Edge Processing Layer
↓
Network / Message Broker
↓
Cloud Platform
↓
Analytics / AI / Storage
↓
Dashboards / Applications
The important thing is that not every data point has to go all the way up the stack.
For example:
Temperature Sensor
↓
Edge Rule Engine
↓
Temperature Normal?
↙ ↘
Yes No
↓ ↓
Aggregate Send Alert
↓
Top comments (0)