The correct option is Time series database.

Time series database systems are designed for high write throughput and efficient queries over time ordered data which matches the probe workload. With 9,600 probes sending a reading every 40 seconds the system receives about 240 writes per second and about 20.7 million measurements per day, and Time series database engines handle this volume with features like compression, partitioning by time, and optimized time-range queries.

Time series database solutions also provide built in retention policies, downsampling and aggregation mechanisms which make long term storage and historical analysis practical without excessive cost. Those features let the analytics team quickly compute trends and windowed statistics over the sensor data.

Graph database is not a good fit because it is optimized for relationship traversal and complex connected queries rather than high volume sequential time stamped measurements.

BigQuery is a scalable analytical warehouse and it works well for large scale ad hoc analysis, but it is not ideal for very frequent raw inserts and time series primitives like automatic downsampling and retention management.

Relational database can store time stamped rows, but traditional relational systems struggle with sustained high ingest rates and they lack specialized compression and time based query optimizations that purpose built time series databases provide.

Full AWS Practitioner Certification Question