Azure Analysis Services on top of Azure Data Lake?
Recently I have seen implementations of Analysis Services that go directly to the Data Lake. In order to speed things up and to remove redundancy, there is no relational data storage even if the data is relational in nature. Data is persisted raw in the lake and goes directly to a user friendly data model in Analysis Services. How to set this up can be found at https://blogs.msdn.microsoft.com/analysisservices/2017/09/05/using-azure-analysis-services-on-top-of-azure-data-lake-storage/
I wonder if this is very practical. Let me ask you: If you are a BI consultant or a BI developer, just how often do you have to answer questions about data? Quite a lot of your time? And what do you use to figure the complex cases? Good old SQL if you ask me. 75% of my time as a consultant is often writing queries to prepare data sets for Power BI or SSAS, figuring stuff out, analyzing the data. Imagine doing that on a data lake with U-SQl. This is very, very time consuming. That is a hidden cost for the business that they may not be aware off. Of course you can use the semantic SSAS model to analyze the data. But there are always situations where you want to dive really deep and a SQL query just gives you more control.
Writing queries is still fastest with the SQL Management Studio on a relational database. Let me clarify: it's not the speed of the writing itself but the speed at which you can get results back. U-SQL on a data lake is slow and cumbersome.
My personal take on BI architecture has always been to use each tool to it's strength. Data Lake is great in handling flat file, schema-less big data. SQL Server is best in relational data. Do not let Data lake or Hadoop be your relational database. It's a different beast.