4

As I can see, the cube issues some queries to the db when processing, like "select xxx from sss", and then build index.

But the doc said, cube will "preaggregate" the measures by dimensions so that it's faster than query the database directly.

Really? Group all measures by all dimension members when processing? If that why still "Design Aggregation" matters in the cube desgin(there is a tab in cube project)?

And the terms in ssas is quite confusing. I want to understand what cube really do on "processing" rather than the "different processing way", like "process full/diff/update", I want to know what "pre-aggreation" means rather than "Aggregation Design", the tab in cube project.

Try hard to be on the same page with MSDN guys...

Does there a book like "SSAS internal", just like "SQL Server Internal" exists?

L_G
  • 209
  • 2
  • 10
  • 1
    "Design Aggregation" - decide _how much_ data is pre-aggregated. (all? none? some?). Processing: load the cube data structure including pre aggregations defined in "Design Aggregation". Do you really need to know how it works internally? Lots of pre-aggregation = better performance but larger cube. Less pre-aggregation = worsse performance but smaller cube – Nick.Mc Jan 17 '18 at 06:21
  • if there is no "Design Aggregation" at all, does cube do any "pre aggregations"? Or just query database and load raw data into cube files? – L_G Jan 18 '18 at 15:34
  • 2
    @L_G if you design no aggs it will not do any pre aggregation during processing. But it will build indexes. When you say "the doc said" it would be helpful if you can include a link. – GregGalloway Jan 19 '18 at 07:41
  • After days of reading books (SQL Server 2008 Analysis Serverices Unleashed, chapter 20 and 23), I think you are right. thanks. – L_G Jan 22 '18 at 09:57

0 Answers0