topic How to handle large datasets: composite model or AAS model? in Developer

How to handle large datasets: composite model or AAS model?

PowerRon — Wed, 10 Jul 2024 14:54:37 GMT

Hi,

we have a few datasets containing each minimal 200GB of data.
We could use a composite model. But in my opinion composite model makes life a little more difficult and whne going to the detailed lverl by using direct query, perfromance might be an issue.

Azure Analysis Services model can handle large datasets and seems to perform excellent.

Does anybody has experience with AAS model ? Would it be a good choice to investigate or are there better options?

regards Ron

Re: How to handle large datasets: composite model or AAS model?

Anonymous — Thu, 11 Jul 2024 09:44:43 GMT

Hi @PowerRon

Regarding your description, it’s important to consider potential performance implications when using DirectQuery within a composite model. For detailed information, I recommend reviewing the official documentation provided below:

Performance implications

For Azure Analysis Services (AAS), it is specifically designed to efficiently handle large datasets, scaling up to manage several terabytes of data. It excels in performance, particularly for complex queries and extensive datasets. AAS processes data in-memory, resulting in fast response times. Additionally, its scalability allows flexibility in accommodating growing datasets. When integrated with Power BI, AAS empowers you to create intricate models while leveraging robust processing capabilities.

Considering your requirements, Azure Analysis Services appears to be an excellent choice for further investigation. However, I recommend conducting performance tests and evaluating costs to ensure alignment with your specific needs. For additional insights, you can explore the following resource:

Migrate from Azure Analysis Services to Power BI Premium or Power BI Embedded - Power BI | Microsoft Learn

Best Regards

Re: How to handle large datasets: composite model or AAS model?

PowerRon — Tue, 16 Jul 2024 15:14:01 GMT

Thnx for the extensive answer @Anonymous