Question

A manufacturing company is building a new platform to ingest telemetry from 500,000 factory sensors, each sending a small reading every 5 seconds. Dashboards must display the last 24 hours of data with sub-second query latency, while readings older than 90 days are queried only occasionally for compliance audits and must be retained for 7 years at the lowest possible cost. The team wants a fully managed, purpose-built solution that minimizes operational overhead. Which architecture best meets these requirements?

Accepted Answer

Ingest readings into Amazon Timestream, which automatically tiers recent data to its memory store for fast queries and moves older data to its magnetic store; export data past the required window to Amazon S3 Glacier Deep Archive for 7-year retention.

Answer

Write all readings to a single Amazon DynamoDB table with on-demand capacity, use a TTL attribute to expire records after 90 days, and rely on DynamoDB point-in-time recovery for long-term compliance retention.

Answer

Stream readings into Amazon RDS for PostgreSQL on a large Multi-AZ instance, partition tables by day, and run a nightly job to VACUUM and archive partitions older than 90 days to Amazon S3.

Answer

Buffer readings in Amazon Kinesis Data Streams and store everything in Amazon OpenSearch Service with hot and warm data nodes, keeping all 7 years of data indexed for both dashboards and audits.