A single-partition query retrieves data from exactly one ScyllaDB partition identified by a partition key. Querying this way minimizes the nodes serving a request and reduces database load, making it a basic performance target established during schema design. section 7.2.2 section 7.2.2
Use single-partition queries as a starting point, then check the remaining causes of latency:
- Limit each request to one partition key to reduce the number of nodes involved. section 7.2.2
- Remember that a single partition can still be slow when rows are large, the partition contains many rows, or the serving node lacks CPU, disk I/O, or memory. section 7.2.2
Restaurant-review data
A query against a small article_summaries_by_date partition illustrates the preferred shape. A single-partition query can still perform poorly if it reads a very large partition or unusually large review rows. section 7.2.2 section 7.2.2
