🚀 go-pugleaf

Avi Freedman <freedman@freedman.net> writes:

>The problem is that the standard MRTG doesn't keep around the 5-minute
>samples for a whole month...

If samples are taken every 5 minutes, you have 8640 in 30 days.  Five
per cent of 8640 is 432.

Our objective is to discard the 432 highest samples in the last 30 days,
and find the next-lower one.  To do this, we need only save, at any
instant, the 433 highest samples seen so far.  The lowest of these, at
any instant, is the value we are looking for, at that instant.

So we need to store 433 samples, not all of them.

What we still need to decide is what each sample measures.  Should it
measure the instantaneous data transfer rate when the sample is taken,
or should it measure some smoothed value?  If it were up to me I would
measure an exponentially decaying average, similar to the way operating
systems calculate a load average.

We also need to decide whether we will cast out duplicates from our
433 samples, or not.
-- 
Rahul Dhesi <dhesi@email.rahul.net> (spam-filtered with RSS and ORBS)
   See my ORBS faq:
      http://www.rahul.net/dhesi/orbs.faq.txt
-
Send 'unsubscribe' in the body to 'list-request@inet-access.net' to leave.
Eat sushi frequently.   inet@inet-access.net is the human contact address.