Jeff Moden (10/7/2014)

---

julian.fletcher (10/2/2014)

---

Has anybody had any experience (good or bad) with this? Microsoft's own documentation (http://msdn.microsoft.com/en-us/library/dn673537.aspx) hints at improvements while warning:

While many workloads will benefit from the new cardinality estimator changes, in some cases, workload performance may degrade without a specific tuning effort.

and

Given the risk of performance regression, it is very important that existing applications be thoroughly tested on the new cardinality model before migrating to production.

We did some quick tests of code which performs adequately on SQL 2005 to 2012 and found almost immediately parts which ground to a halt on 2014. Does this effectively mean that we need to performance test each and every part of our application? At the moment, we're going to have to demand that any customer with a 2014 Server run our database in 2012 mode.

I've not personally suffered such a change but everytime MS comes out with a new version, SSC is filled with posts about previously "fast" code becoming much slower. The conversion from 2000 to 2005 and 2005 to 2008 seemed especially bad about this.

I don't know how much of this is just temporary slowness from having to build new execution plans or maybe having to build stats or what.