[rrd-users] Re: multiple RRAs in a single RRD

[Alex van den Bogaerdt]

On Wed, Nov 20, 2002 at 09:12:57PM +0000, Jim Breton wrote:

> What is the advantage of having multiple archives (with different sample
> intervals) if you are only graphing one of these?

You're not, as Tobi explained.

> For example, if I am graphing network i/o on a system and taking samples
> every 5 minutes, why would I also want to create another RRA which takes
> samples once per hour?

An RRA doesn't take samples.  An RRA stores normalized and consolidated
data.  It is ready to graph data.

>                         I realize that would make a useful graph as well,
> but I don't see how to "get it out" of the RRA when you graph it, since
> it's under the same Data Source; so as far as I can tell the only way to
> determine hourly results based on the above RRD would be to extrapolate
> your 5-minute samples.  But I'm sure I must be missing something!

If you're going to generate an image where each pixel shows one hour
of data, you don't want to to the math on 12 5-minute samples for
each of your 400 pixels.  It is much faster when you already have
the data in a 1-hour resolution at hand.  You don't *need* it, it
is just nice to have.

a) retrieve 12 samples, add them, divide by 12, display
or
b) retrieve 1 sample, display

Another good reason is that you may want to choose not to store 2 years
worth of data in a 5-minute resolution.  Unless of course you want to
actually display the data from way back in high resolution.

a) 365+366 days * 288 samples per day = 210528 samples
or
b) 2.5 days * 288 samples  and  365+366 days * 1 sample = 1451 samples

HTH
Alex

--
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:rrd-users-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/rrd-users
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi