When using XMLA endpoints to manage and update semantic models in Microsoft Fabric, the performance of write operations (such as processing, structural changes, or metadata deployments from Tabular Editor) is directly influenced by the storage format and how the model is persisted.

Why Option A is Correct

By default, Fabric semantic models use the Small semantic model storage format.

To improve write operations performance through XMLA, you must change the workspace setting to use the Large semantic model storage format.

The large format uses more efficient storage techniques, supports partitioning, and handles larger models with optimized write capabilities.

This setting is applied at the workspace level and impacts all semantic models within that workspace, including Model1.

This is explicitly documented in Microsoft's guidance: Large semantic model storage format is required when using XMLA write operations for large or frequently updated models.

Why the Other Options Are Incorrect

B . Configure Model1 to use the Direct Lake storage format.

Direct Lake mode is designed for query performance (reading data directly from OneLake in delta format without import/duplication).

It improves query latency and freshness but does not improve XMLA write operations, which deal with model metadata and structural updates.

C . Delete any unused semantic models from Workspace1.

Deleting unused semantic models helps manage capacity and storage but does not increase the speed of XMLA endpoint write operations.

Workspace storage overhead does not directly impact the write throughput of XMLA operations.

D . Delete any unused columns from Model1.

Removing unused columns reduces the memory footprint and can improve query performance.

However, it does not directly improve the speed of XMLA write operations. The bottleneck in XMLA writes is tied to the storage format, not the model size alone.

Summary

To increase the speed of XMLA write operations on semantic models, you must enable the Large semantic model storage format at the workspace level. This setting ensures better handling of writes and metadata operations via the XMLA endpoint.

Reference

Large models in Power BI and Microsoft Fabric

Use the XMLA endpoint in Microsoft Fabric