IoT Architecture

•Sensors collect data from the environment or object under measurement and turn it into useful data
•The sensing/actuating stage covers everything from legacy industrial devices to robotic camera systems, water-level detectors, air quality sensors, accelerometers, and heart rate monitors. The scope of the IoT is expanding rapidly, thanks in part to low-power wireless sensor network technologies and Power over Ethernet, which enable devices on a wired LAN to operate without the need for an A/C power source.
•In an IoT architecture, some data processing can occur in each of the four stages. However, while you can process data at the sensor, what you can do is limited by the processing power available on each IoT device. 
For deeper insights that require more extensive processing, you'll need to move the data into a cloud- or data center-based system that can bring several sources of data together. But some decisions simply can’t wait for deep processing. Did the robotic arm performing the surgery cut an artery? Will the car crash? Is the aircraft approaching the threat detection system a friend or a foe? You don't have time to send that data to your core IT assets. You must process the data right at the sensor— at the very edge of the edge network—for the fastest response
Stage 2
•The data from the sensors starts in analog form. That data needs to be aggregated and converted into digital streams for further processing downstream. Data acquisition systems (DAS) connects to the sensor network, aggregates outputs, and performs the analog-to-digital conversion.
• The Internet gateway receives the aggregated and digitized data and routes it over Wi-Fi, wired LANs, or the Internet, to Stage 3 systems for further processing.•an IoT system is always on, providing continuous connectivity and data feeds
•That's a lot of data to transport into the data center. It's best to preprocess it.
analog data has specific timing and structural characteristics that require specialized software to process. It's best to convert the data into digital form first, and that's what happens in Stage 2. 
Intelligent gateways can build on additional, basic gateway functionality by adding such capabilities as analytics, malware protection, and data management services. These systems enable the analysis of data streams in real time
Stage 3
•edge IT systems, which perform more analysis,. Edge IT processing systems may be located in remote offices or other edge locations, but generally these sit in the facility or location where the sensors reside closer to the sensors, such as in a wiring closet.
•Because IoT data can easily eat up network bandwidth and swamp your data center resources, it's best to have systems at the edge capable of performing analytics as a way to lessen the burden on core IT infrastructure. 
• For example, rather than passing all aggregated and converted data, analyze it, and send only projections as to when each device will fail or need service.
•You might use machine learning at the edge to scan for anomalies that identify impending maintenance problems that require immediate attention.
Stage 4
•Data that needs more in-depth processing, and where feedback doesn't have to be immediate, gets forwarded to physical data center or cloud-based systems, where more powerful IT systems can analyze, manage, and securely store the data.
•It takes longer to get results when you wait until data reaches Stage 4, but you can execute a more in-depth analysis, as well as combine your sensor data with data from other sources for deeper insights.
 Stage 4 processing may take place on-premises, in the cloud, or in a hybrid cloud system, but the type of processing executed in this stage remains the same, regardless of the platform