[mrtg] Re: How much is too much?

[Henry Steinhauer]

Humm -

Well - From my experience with a slow device that responds, I can tell you that
the time appears to be when the device is polled.

The delta is computed from the current time and the prior time of when it was
polled.

Henry




From: Bradford Woodcock <bwoodcoc at bbn.com> on 03/31/2000 08:40 AM

To:   alex at ergens.op.Het.Net
cc:   SSottile at BGC.COM
      mrtg at list.ee.ethz.ch (MRTG users)  (bcc: Henry Steinhauer/Information
      Systems/National/Hewitt Associates)
Client:
Subject:  [mrtg] Re: How much is too much?



Alex, an interesting comment lets work it through a bit mainly so I
understand mrtg's use of time.

I see your point in the calculation problem however it all depends on how
(and when) mrtg determines the time difference between polls.

>Result: MRTG will calculate a difference in time which is, for example,
>five minutes.  The actual samples were taken 4.5 minutes apart in real
>life.

The question becomes from the comment above "How does mrtg calculate the
time difference?". Do you (or someone on the list) know the answer to this
as I haven't looked at the code?

- Does it use system time when the mrtg job *1st* executes? (least accurate
as you describe)
- Does it use system time for each router or interface as it goes through
the config? (more accurate)
- does it use sysUpTime at all and when?
- what testing has been done to determine the level of accuracy of the
metrics given varying system loads (probably only needed if using system
time at mrtg startup)?

And if there are inaccuracies there are a couple different views to be
taken depending on expectations. If the user wants perfect 5 minute
calculations then mrtg might not be the tool for the job (depending on
answers above). If you are mainly using it for trending (that would be me)
then having a level of inaccuracy isn't the end of the world given that the
trends over time will be fine. So long as it is at least within a couple
percent on individual 5 minute intervals I can live with it given that the
next value will simply lean the other way.

cheers,
brad...


--
Unsubscribe mailto:mrtg-request at list.ee.ethz.ch?subject=unsubscribe
Archive     http://www.ee.ethz.ch/~slist/mrtg
FAQ         http://faq.mrtg.org    Homepage     http://www.mrtg.org







--
Unsubscribe mailto:mrtg-request at list.ee.ethz.ch?subject=unsubscribe
Archive     http://www.ee.ethz.ch/~slist/mrtg
FAQ         http://faq.mrtg.org    Homepage     http://www.mrtg.org