Are Cubes Dead?

  • Comments posted to this topic are about the item Are Cubes Dead?

  • This was removed by the editor as SPAM

  • When I first started working in BI and  data warehousing (10-ish years ago), cubes in SSAS were absolutely the goal. But the effort and expense of creating and maintaining them has always been a showstopper for the orgs I've worked in.

    Then PowerBI came along and management went to some data conferences to learn about it. Now they all see that as a silver bullet and that "citizen scientists" around the business can replace SSAS cubes with PowerBI desktop and a 2 day training course. 😃

  • We have an on premises MDX cube I'm trying to retire that has been 95% recreated as an Azure Analysis Services cube. But we do have some capable citizen developers that are comfortable building Power BI solutions against the Data Mart tables that populate the AAS, and they can use the mashup tools of Power Query to develop things faster than I can add them to the AAS cube.  The measures and reports we create do not need real-time data, so the overnight ETL process still works for us.


  • Considering that:

    1. Microsoft has been touting the SSAS tabular model over the multidimensional model
    2. There haven't been any real upgrades to the multidimensional model in quite some time
    3. Azure Analysis Services doesn't support multidimensional at all...

    Yup, cubes are dead.

  • I'm just entering the BI space , and so is the company that employs me. We are planning on implementing a DW for the first time and using MS BI(SSAS,SSIS,SSRS) services.

    If you are starting from scratch would you try and go straight to the cloud if you want to use a data lake? I'm not seeing anything about on-premise data lake from microsoft.

    I heard Thomas LaRock joke recently on his BI podcast that building a DW is near impossible.

    Any advice in this realm?

  • Are you creating a data LAKE or a data WAREHOUSE?  They are two vastly different things.

  • Sorry, not trying to hijack the thread! We will try to build a data warehouse for sure. I think we may also end up implementing a data lake, though I've only heard that idea in passing.

  • Okay, that makes sense.

    If you are trying to build a data warehouse "from scratch" (also known as "greenfield"), I would definitely do cloud first.  Also, I would start small, perhaps with a PoC on Azure SQL Database, before going straight to Azure Synapse.  Azure Synapse Analytics can get pricey.

  • It depends. Some users browse the cube directly as a decision support tool, so they can quickly slice and dice measurements across dimensions, so the cube itself is the product. But if the cube is really just an intermediate step for generating SSRS reports - just another cog in a complex pipeline - then it probably isn't needed. There are other technologies like Azure Synapse (SQL Warehouse), Snowflake, Amazon Redshift, and other elastic MPP database platforms that can crunch TB of star-schema or raw transactional data in real-time within seconds.

    "Do not seek to follow in the footsteps of the wise. Instead, seek what they sought." - Matsuo Basho

  • Cubes are a legacy technology.  I would also consider SSRS to be somewhat legacy.  If dealing with a greenfield situation, I would go Azure.  And ironically, In-Memory OLTP can help with some star schemas without needing to pay the expensive cost of Azure Synapse, Databricks, etc.

    Again, I recommend creating a Proof of Concept first.

  • Any suggestions on where to look for modern training(book or course)? I'm guessing older training is going to focus on SSAS,SSRS etc.

  • They seem to be on their way out. Some shops I've worked at or known through colleagues had no SSAS (but plenty of databases and a DW), and of the two that did have it, just one instance out of hundreds, used by one team. Though of course it seems like they can still be important to teams who need them - I'm currently on a gig migrating one to 2019. - My technical blog

  • This was removed by the editor as SPAM

  • Azure Synapse has a 1TB minimum.  If you are greenfield, then it will be much cheaper to go with an Azure Database in the meantime.

Viewing 15 posts - 1 through 15 (of 19 total)

You must be logged in to reply to this topic. Login to reply