Variable Telemetry Volume
Factory telemetry volume varies dramatically — high during production shifts, near-zero during weekends and shutdowns. Traditional server-based architectures waste money on idle capacity.
Serverless & Lambda Consulting
Serverless Architecture for Industrial IoT on AWS
We design serverless architectures for industrial IoT workloads — AWS IoT Core, Lambda, Amazon Data Firehose, DynamoDB, and Amazon Timestream — that scale automatically with production volume and eliminate idle infrastructure costs for variable-load factory operations.
- An event arrives — API, queue, or schedule.
- A function handles one unit of work.
- Integration is the bus and the store, not a hidden hop.
- Traces and alarms are in the same change as the function.
- Deploys stop on the pipeline check, not after customers see it.
- Capacity follows load; idle time is not a fleet you pay for.
Key Challenges We Solve
Real-Time Alert Latency
Equipment fault alerts must reach maintenance teams within seconds of detection — not after batch processing cycles. Event-driven architectures are required for safety-critical alerting.
Multi-Protocol Telemetry Ingestion
Industrial sensors publish on different protocols (MQTT, HTTP, OPC-UA). A unified ingestion layer must normalize and route messages from heterogeneous devices without per-device infrastructure.
Time-Series Storage Cost
High-frequency sensor data (100ms samples from hundreds of sensors) generates massive data volumes. Choosing the right time-series storage — Timestream vs. DynamoDB vs. IoT SiteWise — significantly impacts cost.
Our Approach
Event-Driven IoT Pipeline
AWS IoT Core rules engine routes telemetry to Lambda for real-time processing, Amazon Data Firehose for S3 archival, and Amazon Timestream for time-series queries — a fully serverless pipeline with no servers to manage.
Serverless OEE Alerting
Lambda functions triggered by IoT Core rule conditions evaluate OEE thresholds and dispatch SNS alerts to maintenance teams via SMS, email, or CMMS API integrations — sub-second latency from fault detection to notification.
Amazon Timestream for High-Frequency Telemetry
Amazon Timestream purpose-built time-series database stores high-frequency sensor data with automatic data tiering (memory store for recent data, magnetic store for historical) — dramatically lower cost than DynamoDB for time-series workloads.
Frequently Asked Questions
Which AWS services form the core of a serverless manufacturing IoT stack?
The core serverless IoT stack for manufacturing: AWS IoT Core (device connectivity and message broker), AWS IoT Greengrass v2 (edge processing and protocol translation), Lambda (real-time telemetry processing and alert logic), Amazon Data Firehose (streaming delivery to S3), Amazon Timestream (time-series storage), and Amazon QuickSight (operational dashboards).
Should we use Amazon Timestream or DynamoDB for IoT sensor data?
Amazon Timestream for time-series queries (trending, anomaly detection, OEE calculations) and high-frequency sensor data — it is purpose-built for this workload and significantly cheaper per GB for time-series data. DynamoDB for operational state (device shadow, current readings, alert status) where point-lookup latency matters. Most manufacturing architectures use both.
How does Amazon Data Firehose fit into a manufacturing IoT pipeline?
Amazon Data Firehose (formerly Kinesis Data Firehose, renamed Feb 2024) delivers streaming IoT telemetry to S3 for long-term archival and batch analytics. It buffers data, applies optional Lambda transformation, converts to Parquet or ORC format, and partitions by date/site — creating a queryable data lake without managing Kafka or stream processors.
How do we handle offline-first operation with a serverless architecture?
AWS IoT Greengrass v2 handles offline operation at the edge — buffering telemetry locally when AWS connectivity is interrupted and replaying to IoT Core when connectivity resumes. Greengrass also runs local Lambda functions and ML inference, ensuring critical plant-floor logic continues operating without cloud connectivity.
What does a serverless IoT architecture cost for a typical factory?
A facility with 500 sensors publishing at 1-second intervals generates ~43 million messages/day. At AWS IoT Core pricing ($1/million messages) plus Timestream storage and query costs, typical monthly infrastructure cost is $200-800 — versus $2,000-5,000/month for equivalent EC2-based infrastructure. Savings grow during planned shutdowns when serverless cost drops to near zero.
Ready to Get Started?
Talk to our AWS experts about serverless architecture for industrial iot on aws.