SQL Server 2012 licensing, changes-a-coming?

Reading a post by Jeremiah Peschka over at brentozar.com/ and as I commented it raises some interesting questions above the standard “How much more is SQL2012 going to cost us” ones.

Current employer has been generally moving towards a strategy of larger ‘real’ SQL boxes and using smaller virtual app servers. This is partly due to this place’s Oracle heritage, where Oracle’s mid tier structure really pushed the idea of everything running from packages in the DB.

So all the heavy lifting and processing was pushed back onto the DB iron as there was plenty of power in the newer multi-core processors which weren’t under quite so much pressure as more of the DB could be moved into memory and the SAN got faster. This was great as the app servers got correspondingly smaller so you could get more VMs onto a given infrastructure, which was great as OS licenses were cheaper than DB licensese.

Now that SQL2012 is moving to per core this model might not make as much sense as it did before. Suddenly all those extra cores in the DB boxes are an expensive liability. Based on Microsoft’s released Licensing Datasheet we’re good for the upgrade (Software Assurance to the rescue) as we’ll get all our current SQL2008 CPU licenses turned into the appropriate number of cores of SQL2012 licensing. But our standard DB box is based on 8-core procs at the minute, and according to the datasheet the a SQL2012 core license is going to be 1/4 of the cost of a SQL2008 CPU license. Which makes each processor in our boxes twice as expensive to license under 2012. Suddenly having all that horsepower starts to look like a liability that business will want to reduce.

Reducing the grunt at the backend will mean either:

  • Developers having to become a bit cleverer with their code. spending more time on it.
  • Bigger App servers

Neither of which come cheap.

It’s going to be interesting (for me at least) to see how management pass the costs around. Especially as SQL2012 is on managements wish list (after some pushing) thanks to the HA (group failover) and related reporting features (read ability on mirrors). Something to look forward to chasing up after the Christmas break