SQL Server BI / DW Scalability

One of the common FUD* techniques deployed against SQL Server is to raise questions about its ability to scale. Rather than blather on about technical reasons why this is bunkum, i’ll just hit you up with some numbers from this springs SQL PASS:

Category Metric
Largest single database (DW) 80 TB
Largest table 20 TB
Biggest total data 1 customer 2.5 PB
Highest transactions per second 1 db – OLTP 36,000
Fastest I/O subsystem in production 18 GB/sec
Data load for 1TB 20 minutes
Largest cube 4.2 TB

So, seriously, unless you have volumes of data on a par with MySpace… Scalability is not an issue.

2 thoughts on “SQL Server BI / DW Scalability

  1. Sadly, I have to disagree. Scaling *remains* an issue for SQL Server, even if it’s *much less* of an issue than it once was. The problem with your table above is that it gives no indication of the resources required to achieve those figures, e.g., how much hardware, software, licensing, and professional services cost was incurred.

    Having worked on large-ish SQL Server DWs, I can surmise that the answer is *a lot*. Regardless of my experience, Microsoft’s huge investment in the PDW product verifies the existence of the issue. PDW and Vertipaq, along with other Denali improvements, are promising but not the ‘promised land’. Vertipaq in particular will move things forward for many SS shops, but the Denali version falls short of offering a true row/column hybrid.

    SQL Server continues to represent an excellent middle ground, however it’s use at extreme scales is very limited and will continue to be so unless Microsoft fully commits to a radical new architecture in the way Oracle has done with Exadata, IBM has done with the Netezza acquisition, etc.


  2. Thanks Joe. My point is that in the *vast majority* of cases, scaling isn’t an issue. I’m sure plenty of hardware was required to support those large scale implementations, but I would be very surprised if any competing platform wouldn’t also require similar grunt.

    The PDW / Exadata / General MPP thing is – as you point out – to deal with extremes of scale. But up to the tens of terabyte scale, vanilla SQL Server is perfectly capable. Most BI / DW implementations don’t operate at that scale (yet).

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>