In Ye Olde Olden Dayyes, companies would outline hardware budgets and refresh cycles.
Every 3-5 years, they’d lick a thumb and hold it up to the CPU fans. If they felt a soft breeze, they’d buy the same number of CPUs. If they felt a hard breeze, they’d buy a few more. This made the hardware buy more expensive, and of course drove up licensing costs.
For physical servers, performance tuners generally keep costs level. This started to change with VMs, and server consolidation projects. But even here, costs rarely went down more after the mortar dried.
Frankly Mr. Shankly
This is way different with the cloud. In Azure, the ever-infuriating DTU reigns supreme, but machines are still priced by size. You can change that size, and how much a machine costs you by pushing a few buttons.
That means that you, as a performance tuner, can translate your skills into the kind of bottom line stuff that makes the people who decide salaries and bonuses very happy.
It does, however, require you to fight an extremely contrary urge: Hardware can only ever get bigger.
Shrinking Your Server
There’s obviously a lot to consider before decreasing hardware. You’ll need some long-term trending of performance data to prove your efforts aren’t a fluke, and that dialing CPUs back from 16 to 8 (just as an example) won’t backfire the next time there’s a surge of users. You also need to keep some hardware slack around if you’re expecting future growth.
These are the kinds of things that companies are going to look for more and more from their DBAs and Developers, as things shift Cloudwards. People with the right set of skills go from being cost centers to savings centers.
Gaining the skills and confidence to do that can be a hard-won task. It doesn’t happen overnight, and it doesn’t happen unless you devote the time to it. Obviously my angle on it is that it makes you getting a budget for paid training (hopefully from us!) really easy.
Training is really important for you generally, to make sure your skills don’t end up full of moth holes, and that you learn the kind of changes you can make that will improve performance. Not the kind of flailing around that ends up with people chasing the wrong things and searching for a magical feature that will fix anything — Partitioning is a great example of this.
Cutting cloud costs has been coming up more and more as the reason why people have ended up in our classes. I’m sure other companies out there who do SQL Server training will tell you the same thing.
If there’s a click-baity scare line in here, it’s that if you don’t do it, someone else will.
I’d much rather you hear that from me now than after someone else has stolen your muffins.
Thanks for reading!
Brent says: one of my favorite SQL ConstantCare® moments so far is when I explained to a customer that their SQL Server is bored, and showed the metrics with proof. They dropped the server from 24 cores down to FOUR, and were fine afterwards. In the cloud, it’s so much easier for DBAs to provide value that pays off right away.