We tell people that Database In-Memory is all about improving analytic processing. The definition that I use is "using aggregation to find patterns or trends in data". I believe I saw this used in one of Juan Loaiza's OpenWorld presentations and it made sense to me. After all Database In-Memory is really good at scanning and filtering a lot of data, and most of the articles on this blog echo that theme.
The problem is, the definition that I use doesn't necessarily make sense for everyone. When we do presentations or give webinars many people don't seem to really get what we mean by analytics.
While watching the Golden State Warriors two weeks ago from my hotel room the TV commentators began discussing how the Warriors use analytics pretty extensively and they mentioned that this is in no way mainstream in basketball.
This reminded me of where baseball was back when Michael Lewis published Moneyball: The Art of Winning an Unfair Game, the story about how the Oakland Athletics (another SF Bay Area team) used an analytic, evidence-based, approach to competing in Major League Baseball. At the time they were severely chastised by the baseball establishment and the story made for a pretty good book and movie.
So I did a little surfing on what the Warriors define as analytics and found the following quote in an article on SF Gate:
I'll re-phrase Kirk Lacob's definition of analytics to "using information or data to make informed decisions". And that's what Database In-Memory is all about. The ability to run adhoc, analytic queries on source data without impacting the existing workload, except for the CPU cycles required to scan the data in-memory. There is also the additional benefit, or side-effect, in a mixed workload environment of allowing the removal of analytic indexes that probably only exist to support analytic reporting. You can even read about a real world use case where Bosch was able to remove 76 analytic reporting indexes to improve the performance of their SAP CRM system.
Now that we have a better idea of what analytics are, next up we will try characterizing an analytic query.