This month's T-SQL Tuesday is hosted by Jen McCown of @midnightdba fame. She wants us to write about strategies for managing the enterprise, and as a database consultant, I find myself giving advice on this kind of thing to customers remarkably often.
No, I'm not going to do stories about LobsterPot customers. We don't do that. What happens between us and the customer stays between us and the customer. However, I can talk about the kinds of things that we look for when we talk to a customer.
The thing that I want look at in this post is about that twitch that you get when something doesn't feel right. The 'spider-sense' feeling that Peter Parker gets when there's something that's not quite right.
Experience is what helps people know what 'normal' looks like. I've heard stories that people who are trained to spot counterfeit bank notes don't spend time with fake notes, they spend time with the real ones (on loan, presumably). They learn what 'normal' looks like, and get really familiar with it. That way, when something isn't quite right, they can spot it and raise an alarm.
For DBAs taking care of an environment, whether small or large, they will have learned what 'normal' looks like. They should have benchmarks that tell them what how various metrics perform, and for those things that they don't have formal metrics on, they should still be familiar enough to recognise when something isn't right, even if they can't tell exactly what. Their 'spider-sense' should tingle.
If you don't have benchmarks, then get them. Every time you find something that doesn't seem right, you will wish you had a benchmark on that thing to be able to quantify your suspicion. Feeling like something isn't right is great, but it won't be long until someone asks "How do you know this thing needs attention?" and "I just know" probably won't cut it. If you're a consultant, you can probably get away with "In my experience…", because that's what they're paying you for, but having supporting evidence – actual numbers – can help, particularly if you're the main person responsible and are needing to persuade someone to find the money for a CapEx.
Having the numbers handy is useful for a lot of situations, but there are a whole bunch of tools also available to look at too. A while back I wrote about how DBAs could use the same kinds of tools that other industries hire data professionals to provide, in a post called "DBAs and the Internet of Things". If you take this kind of approach and start analysing the readings from all kinds of things that affect your database, then you can get ahead of the game. Feed this stuff into something like Azure ML for predictive analytics, and you might be able to have an even-better spider-sense, where the prediction isn't just based on your own opinions, but on what has caused failures in the past.
Too often, the significant thing is some small detail that most people wouldn't notice, but before of your experience and expertise, you can spot it and work out whether it's significant or not. Then if you don't have that particular thing benchmarked, or analysed by other tools, you can include it to see what's going on.
…and develop superhero powers for managing your enterprise. It's something we regularly recommend to DBAs.