This editorial was orignally published on Sept 26,. 2014. It is being republished as Steve is on holiday.

When I asked recently what compelling reasons there are, or aren't, for upgrading SQL Server, it seems cost was the issue for most people. I can understand that, especially as licensing for VMs has made it an expensive endeavor to upgrade their systems. Many, many companies have started to revisit their policies regarding upgrades and seriously evaluate the decision to undertake a version change. While I'm not sure if many companies are starting to look at keeping all instances of SQL Server for ten years, I know many are looking to keep their noncritical systems on the existing versions as long as possible.

As I read through the posts on my piece, I found this great quote from one of the commenter: "I'm a great believer in "If it ain't broke, don't fix it" and so if a system is running acceptably for it's purpose, I'll leave it alone." To a large extent, that's how I've viewed systems. If things are working, it becomes hard to justify an upgrade. In fact, I had a client that was running a SQL Server v6.5 instance in 2005, a full ten years after that version was released. Why? Mostly because the database backed a card key system that worked. While the database wasn't supported, neither was the card key system, and since an upgrade would have been close to $100,000, it wasn't worth performing. We did upgrade to a virtual server to remove hardware dependencies, but otherwise left the system alone.

That brings me to this week's question, which is one that I've asked myself for years: Is SQL Server mature?

If a version of SQL Server can handle the load placed on it, is there any good reason to upgrade. I suspect that for many applications, the base features of SQL Server are good enough. If these features have performed well for a number of versions, then why change? While I like the changes in T-SQL, especially error handling, I'm not sure those changes are worth paying tens of thousands of dollars for. I'm not sure the improvements are really justified for many applications, and I'm not sure many developers can even find the time to implement them.

More and more I think that key to continued adoption and growth of SQL Server is licensing by scale, not edition, and not by feature. Just like Azure, let me pay for the cores and RAM I need, and let me easily grow that as needed.