Weighing an accelerated shortcut against a standard one for query speed

In Real-Time Intelligence, OneLake shortcuts let a KQL database reference Delta tables in OneLake without copying data, but querying that data directly can be slow because it isn't stored in the KQL engine's native indexed format. Query acceleration for OneLake shortcuts builds and maintains a cached, indexed copy of the shortcut data behind the scenes, giving near-native KQL query performance at the cost of extra storage and a caching delay. Choosing between them is a trade-off between query speed/complexity versus storage cost and data freshness requirements.

Must-know

• Standard OneLake shortcuts read source Delta/Parquet files on demand, so queries run without extra storage cost but with lower performance for complex, high-cardinality, or frequent analytical queries.
• Query acceleration adds a background caching/indexing process that transforms shortcut data into the KQL engine's optimized storage format, improving query speed similar to native ingested tables.
• Because acceleration runs asynchronously, there is a lag between new data landing in the source and it becoming available in the accelerated cache, so very fresh data may briefly be served from the unaccelerated path.
• Enabling query acceleration incurs additional storage consumption in the Fabric capacity since it maintains a cached copy alongside the original OneLake data.
• Choose standard shortcuts when data volume/query frequency is low, near-real-time freshness is critical, or minimizing storage cost matters; choose query acceleration for large datasets with frequent, complex, or performance-sensitive analytical queries.
• Query acceleration is configured per shortcut (not automatically applied to all shortcuts), so it must be explicitly enabled where performance gains are needed.

Coverage checked against the published exam guide on Aug 11, 2026.

These are independent practice questions, written against this certification's published exam guide. They are not the certification vendor's own questions, and not the real exam.