Comment 8 for bug 618019

Revision history for this message
Jamu Kakar (jkakar) wrote : Re: OOPS may be underrepresenting storm/sql time

I would guess time-to-fetch can be an important measurement. A
query to fetch a 1000 objects might be very quick to execute, but
pulling the data might be slow. If you see that the query takes
100ms but the reality is that 2s are spent fetching, and you can't
easily know that that's the case, it will be hard to figure out that
you ought to figure out how to do what you want with less than 1000
objects.

I think it would be worth adding this kind of logic to Storm. If
we're unsure about whether it's truly useful we could add it
experimentally and use it in anger to determine how useful it really
is.