Performance of refresh of a semantic model is inconsistent and fails some times.

[AlexEncuble]

Hello Community. 

We are switching from an Azure Analysis Cube to a Power BI Semantic Model. The transfer was working good, but the refresh of the Semantic Model is sometimes unusually long, runs into a Timeout we never set or fails because of internal errors... We can't figure out why. 

Do you have any tips or tricks to make sure it works better so we can finally get rid of the AS Model? 

Thanks in advance for any tip. 

[v-dineshya · Accepted solution]

Hi @AlexEncuble ,

Thank you for reaching out to the Microsoft Community Forum.

Please check the below workarounds to fix the issue.

Query Folding : Avoid unnecessary joins or transformations. Use query folding wherever possible to push processing to the source.

Incremental Refresh: Configure incremental refresh to avoid full dataset reloads. This is helpful for large datasets.

Partition Management: Partitions can reduce refresh time and isolate failures.

Capacity and Resource Monitoring: Your Power BI workspace is on a Premium capacity and monitor resource usage. Overloaded capacities can cause timeouts.
Use the Refresh History and Performance Analyzer to identify bottlenecks.

Gateway Configuration: If you are using an on-premises data gateway, ensure it's updated and has sufficient resources. Consider scaling out the gateway cluster if multiple refreshes are running concurrently.

Optimize the SQL code: Optimize your SQL code and then refresh.

Model Size and Complexity: Remove unused columns and tables. Avoid complex calculated columns and measures during refresh.

Increase SQL Timeout: Power BI defaults to a 10-minute SQL timeout in many cases. Set a higher timeout in the Data source settings via Power BI Desktop or using advanced options in the Power Query connection string, Command Timeout = 1800; (1800 seconds = 30 minutes).

Please refer Microsoft articles.

Query folding guidance in Power BI Desktop - Power BI | Microsoft Learn

Incremental refresh for semantic models in Power BI - Power BI | Microsoft Learn

If my response has resolved your query, please mark it as the "Accepted Solution" to assist others. Additionally, a "Kudos" would be appreciated if you found my response helpful.

Thank you

[AlexEncuble]

Thanks, i will try that

[burakkaragoz]

Hi @AlexEncuble ,

Yes, it is possible to use both Microsoft Fabric Link and Azure Synapse Link in parallel from the same Dataverse environment. Microsoft supports having multiple export profiles (such as Synapse Link and Fabric Link) running simultaneously on Dataverse.

Things to consider:

1. Performance Impact:
   Running both links in parallel will add extra load to your Dataverse environment. If you have large tables or frequent data changes, you might notice some slowdown or increased resource usage. However, both Synapse Link and Fabric Link are designed to work efficiently in production environments.

2. Capacity and Limits:
   Be sure to check Microsoft’s limitations for table sizes, row counts, and export service limits.
   More info: Dataverse export service limits and performance

3. Monitoring and Testing:
   After enabling both links, closely monitor your environment for any slowdowns or errors. If possible, test in a non-production environment before going live.

4. Best Practices:

• Only export the tables and columns you actually need.
• Optimize the refresh schedules and avoid unnecessary frequency.
• Regularly review resource usage and export logs.

Summary:
Yes, you can run both Synapse Link and Fabric Link in parallel from the same Dataverse environment. Just keep an eye on performance and follow Microsoft’s best practices to minimize any potential impact.

Good luck!
translation and formatting supported by AI

[AlexEncuble]

That is not really an answer to my question...