Before you get too excited
This isn’t a dive into any of the new automated tuning features of SQL Server 2017.
I’m interested in them because Microsoft thinks they can put us out of business with them, but…
See, a lot of the automated performance tuning features assume you EVER HAD GOOD PERFORMANCE IN THE FIRST PLACE.
I mean, sure, it may have seemed okay when your database was 5, 10, or 20 GB, but now it’s 200 GB and things have just been getting worse for months. Or years.
These features are cool, but they don’t go in and fix your crappy code, your wrong settings, or get you off that VM with 8 GB of RAM that your VM Admin swears by the grace of his body pillow girlfriend is all a SQL Server will ever need.
After all, he read a book once.
Best of the worst
Our typical customer is worried about…
Keeping the lights on:
- If they’re on the right hardware
- If they can or should move to the cloud
- If they’re meeting RPO/RTO
- If their HA/DR strategy is correct
- How do I validate server settings?
- How do I figure out what my problems are?
- How do I find my worst queries?
- How do I tune my indexes?
- Bonus points: How do I interpret output from your free scripts?
Some customers are more savvy, have pains around parameter sniffing that they’ve identified, and want help with those. They’re tired of restarting the server, freeing the proc cache, or recompiling queries. But the thing is, there’s usually so much other stuff that gets uncovered, parameter sniffing getting fixed is almost a side effect of other changes.
- An index adjustment took the bad plan choice away
- It wasn’t parameter sniffing, it was a scalar function
- There was a blocking problem
- Non-SARGable predicates buried four nested views deep
So, while I’m totally keen on automated performance tuning, and SQL Server 2017 being able to compare and correct bad plan choices, I’m still not worried about my job.
We still talk to people on 2008-2012 who are locked in because of vendor requirements, or other reasons, who just aren’t going to see those new features in the near future.
And plus, they still assume performance was good in the first place.
Thanks for reading!