[rrd-users] RRD Database precision

[Sam Stickland]

Hi,

I'm optmising a perl RRD replay script I've written (it "replays" DSs 
from one RRD to another, preserving max and average information as best 
possible, even when the RRA definitions are notably different). At the 
moment, I've got somewhat liberal use of Math::BigFloat, and I envisage 
quite a notable speed increase if I use the native floating points where 
possible.

What precision does RRD store it's information in? Looking at the output 
of "rrdtool dump" I see 11 decimal digits and an exponent, so this 
should be storable within perl's native float, with a bit of precision 
left over (room for an extra five decimal places)? 
(http://perldoc.perl.org/perlnumber.html).

So I should be able to work with perl's native floats, and only invoke 
Math::BigFloat when I need to round it to a native integer for passing 
to RRD?

Is this right?

Sam