Amazon Quick Sight has introduced a new capability called Multi-Dataset Relationships, which allows users to define logical relationships between datasets and perform runtime joins at query time. This change eliminates the need to pre-join all data into a single wide dataset before analysis can begin, thereby reducing upfront data preparation and improving analysis flexibility. Users can now maintain each dataset at its native level of detail and let the system handle the necessary joins when needed for visuals, calculated fields, or natural-language queries.
The new feature addresses common challenges in traditional single-dataset models, such as the costs of upfront preparation, measure duplication, and dataset sprawl. By modeling tables once and keeping them at their native granularity, users can avoid the need for repeated data transformations and ensure consistent data access across different analytical scenarios. The system also supports row-level security at runtime, ensuring data-access policies apply consistently across datasets. Additionally, datasets can have independent refresh schedules, allowing data to be ingested at different cadences based on volatility.
The post explains how Multi-Dataset Relationships work by using two distinct modeling layers: a physical layer within each dataset and a logical layer across datasets in a Topic. The physical layer combines tables with joins, unions, and transforms, while the logical layer defines relationships between datasets through matching key columns. Quick Sight only combines datasets when a visual or calculation requires fields from multiple sources, ensuring efficient and targeted data retrieval.
Source: awsml
