One thing I should ask now that there's a count in question here as well and note of its efficiency. I believe with myisam that the count was a part of the metadata, but I believe is not the case in the use of innodb. Is there ever a point in time where the use of count becomes inefficient to the degree where keeping a separate counter would be warranted (using INNODB)?
I ask since I'm considering something I'm working on. Whilst optimistically I'd be looking at needing to do aggregate count, min/max, avg, etc on grouped by data in the counts of hundreds of millions, realistically I'd be looking at hundreds of thousands or maybe tens of million. Since I'll always build based on optimistic potential over more realistic potential, would it be better to count separate data or use simple addition over that of using counts?
All and all I would presume that the addition will always be faster, but its not that I prefer that approach (since I make use of more than just counting and whatnots). I'm more curious if I'd be looking at extreme degradation of using properly indexed records in the tens of millions versus a simple select query of an existing count.
BTW, I would need to keep the records regardless of if I run a separate counter or not. Just curious if you think there would ever be a need for concern.
header('HTTP/1.1 420 Enhance Your Calm');