[mrtg-developers] Re: How many Routers we can poll using MRTG

[Henry Steinhauer]

Bjon -

I am monitoring more than 7,000 interfaces with running NT.  This is done with a
dual compaq machine with 512 Meg of memory.
500Mhz processors.  It was left over from another project and I happened to be
in the right place at the right time.  The old machine was dual 300 Mhz with
256Meg and it needed some help bad.  It just could not keep up and more people
were counting on the results being current.   A problem when it is running slow.

Tobias is right in that it is very I/O intense.  I use SCSI drives.  Keeping
sure that the free space is at least 50% on the main drive.  This seems to help.
Disk is cheap compared to mis-information.

I also fork the tasks on NT by running many programs using the start command to
launch each of the CFG files.

This helps since it is waiting on the I/O from the device (SNMP) and I/O from
the Disk for the updates.  (I'm still on the old Rateup.  When I get back in the
office, I hope to finally convert to the RRD version.)  :-)

Henry




From: Tobias Oetiker <oetiker at ee.ethz.ch> on 09/29/2000 03:30 AM

To:   Bjorn Nordbo <bn at nextra.com>
cc:   mrtg-developers at list.ee.ethz.ch
Client:
Subject:  [mrtg-developers] Re: How many Routers we can poll using MRTG




Today you sent me mail regarding [mrtg-developers] Re: How many Routers we...:

BN>
BN> Jaji Hegde wrote:
BN> > 1. How many routers we can poll with polling frequency of 5 minutes.
BN> > Roughly, if you have done any experiment that would be great. Lets
BN> > assume on average 3 interface for each device.
BN>
BN> I estimate we could do at least 3000 targets with MRTG 2.9pre26 and
BN> rrdtool on an Ultra 2. That's 1000 routers /w three interfaces each.
BN>
BN> > 2. If the polling doesn't finish by 5 minutes, when cron tries to start
BN> > the second mrtg what happens.
BN>
BN> The new job will die and the old one will finish. With rateup, this
BN> wasn't a big problem, but with rrdtool the graphs gets really ugly.

*  can you elaborate on the 'ugly' bit ?
*  you may want to use the daemon mode then the mrtg will run
   continuously if it does not finish withing the interval

tobi

BN>
BN> > 3. How CPU intensive is mrtg.
BN>
BN> Not very; lots of time is spendt on I/O. But you would want to fork
BN> a lot.
BN>
BN> > 4. How can we scale it to poll more routers in the given polling
BN> > frequency.
BN>
BN> Add more iron?
BN>
BN>

--
 ______    __   _
/_  __/_  / /  (_) Oetiker, Timelord & SysMgr @ EE-Dept ETH-Zurich
 / // _ \/ _ \/ / TEL: +41(0)1-6325286  FAX:...1517  ICQ: 10419518
/_/ \.__/_.__/_/ oetiker at ee.ethz.ch http://ee-staff.ethz.ch/~oetiker


--
Unsubscribe mailto:mrtg-developers-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:mrtg-developers-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/mrtg-developers








--
Unsubscribe mailto:mrtg-developers-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:mrtg-developers-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/mrtg-developers