Microsoft Certified: Fabric Analytics Engineer AssociateImplement and manage semantic models (30-35%)Medium
A data engineer is configuring a semantic model in Microsoft Fabric. The model sources data from a data warehouse that is updated every 15 minutes. The business users require the reports based on this semantic model to reflect the latest data with minimal latency. However, a full refresh of the entire model takes over an hour. Which refresh strategy should the data engineer implement to meet the latency requirement?
- AStandard scheduled refresh every hour
- BIncremental refresh configured for daily updates
- CEvent-driven refresh triggered by a data pipeline completion
- DReal-time refresh using DirectQuery or Direct Lake mode for critical tables
Show answer & explanationAnswer & explanation
Correct answer: D. Real-time refresh using DirectQuery or Direct Lake mode for critical tables
To meet the 'minimal latency' and 'reflect the latest data' requirement from a source updated every 15 minutes, 'Real-time refresh' which leverages DirectQuery or Direct Lake mode for critical tables is the most suitable strategy. This avoids the long full refresh time by querying data directly.
Why the other options are wrong
- A. An hourly refresh would not meet a 15-minute latency requirement.
- B. Incremental refresh optimizes refresh time but still involves loading data; daily updates won't meet 15-minute latency.
- C. Event-driven refresh is good for automation but doesn't inherently reduce the refresh duration if the model is in Import mode and takes an hour to refresh.
Real-time Refresh
A refresh strategy that provides the freshest possible data by leveraging DirectQuery or Direct Lake modes, where queries are executed directly against the source data, minimizing latency.
- Achieved primarily through DirectQuery or Direct Lake storage modes.
- Data is always current, reflecting source changes immediately.
- Suitable for scenarios requiring minimal data latency, but can impact query performance.
Memory trick: Real-time data: no wait, just query straight.