Idea suggested by Biju KRISHNAN on Sizing guidelines for a drill only cluster
"I just finished watching a presentation by Neeraja and Aman offering a lot of insight into drill deployment in production. This was a recorded session and someone asked the same question to them. Their answer was that this really depends on how much data is read during a query. Totally agree.
What would be nice is to know from your professional services team, as to how many drill bits and what sort of config do some of your reference customers have.
Here in the info graphic
What a Difference a Year Makes: Happy Anniversary, Apache Drill | MapR
You have examples of cardlytics and fishbowl, you have mentioned the number of records and query response but not the specs of resources used. Thatz what would be super useful if someone is willing to share. "