So, this month’s T-SQL Tuesday topic is to think about a non-SQL Server technology that we want to learn.
For me, I’m going to pick machine learning.
As a DBA, I’ve always looked at machine learning as a thing for the BI guys. I’m a DBA after all why do I care about that?
Well, my attitude has changed somewhat recently. This little change all started when I listened to Alex Whittles’ keynote talk at Data Relay. He presented a demo where a computer program used Python (I’m already a huge fan of Python in SQL Server as you may know) and SciPy (a machine learning, data sciencey type module) to play and learn a game. Alex demonstrated how, over time his robot was able to increase it’s score through machine learning algorithms.
WOW, myself and Adrian looked at each other as a little light bulb come on over our heads. For the rest of the conference I attended a number of sessions that I wouldn’t normally attend, stuff for the BI guys. A great session from Terry McCann and an interesting one from Simon Whiteley really got the creative juices flowing. Could the DBA use this technology to model things like performance trends, predict capacity and answer that question that we’re always asked, “have we got room on the SQL Server for just one more DB?”.
So where do I go from here? My first port of call is going to get my head around Python, I’ve got a background in C programming to that shouldn’t be too difficult. Once I’m happy with that, it’ll be a case of hitting the blogs, courses, books and anything else that I can get my hands on to help understand the strange mysteries that are Machine Learning.
Where can I go with this? As DBAs, we’ve got a ton of data available to us in DMVs, Query Store, etc. Wouldn’t it be great if we could hook a little robot into all that and start building up models of how our servers behave. Keep an eye out for the inevitable blog post that are going to come out of it.