Millions of IoT devices stream sensor data to BigQuery. Queries focus on recent data filtered by create_date, location_id, and device_version. How should you partition and cluster the table for cost and performance?

Partitioning by createdate is crucial because queries frequently filter by recent data. This significantly reduces the amount of data scanned, directly impacting cost and performance. Clustering by locationid and deviceversion further refines data organization within each partition. When queries filter by these columns, BigQuery can efficiently locate relevant data blocks, minimizing scan time and improving query speed. Incorrect options: Partitioning by all three columns would be inefficient. BigQuery supports partitioning by a single column (or an ingestion-time column or integer range), and while it allows partitioning by an additional column for date/timestamp/datetime, adding more would not be supported in this manner. Clustering by all three columns without partitioning would not offer the same performance benefits for time-based filtering, as BigQuery would still scan all partitions. Clustering by createdate and partitioning by the other two is incorrect because partitioning by createdate is the primary optimization for time-series data, and BigQuery's partitioning capabilities are not designed to partition by multiple non-date/timestamp columns in this way.