[mrtg] Re: Server Specs.
tony bourke
tony at vegan.net
Sun Jun 10 16:47:36 MEST 2001
Hi Jon,
I think the problem is the IDE bus. If you can get one, a SCSI system
would be able to handle quite a bit more interfaces, and if it were a dual
processor then even better.
IDE uses the actuall processor to control the drives. The result is when
there is alot of disk usage, not only is the CPU working on your task but
working on writing blocks to your hard drive. Nothing like a hung IDE
bus during a tar -xvf.
SCSI has a dedicated processor on the controller for the block writes, so
there is quite a bit less processor load when disk access is done.
Normally you wouldn't see much of any difference for a desktop machine,
but the difference is huge when dealing with servers and high disk I/O.
I've run into this very same issue with MRTG before, and SCSI made a huge
difference in performance and scalability.
Oh, and staggering them out amoung 5 minutes will help quite a bit too.
Hope that helps,
Tony
On
Thu, 7 Jun 2001, Jon Viventi wrote:
>
> RedHat 7.1
>
> Yeah, the box works flawlessly when it first boots, but each sucessive time
> it runs it loads up more and more ram until it slows. I upgraded to 384 to
> see if that would help (from 256meg), but it just used up that ram too....
>
> Maybe i'm running too many processes.. I have 10 running.... Maybe I should
> combine some of the smaller ones down to 5 processes and start one each
> minute?
>
> It also seems to be wasting time waiting for down sites. Plus, I'm
> monitoring over a few 32k PVCs....
>
> I dont know. What does everyone think? Thanks
>
> -----Original Message-----
> From: mrtg-bounce at list.ee.ethz.ch [mailto:mrtg-bounce at list.ee.ethz.ch]On
> Behalf Of Tim Kennedy
> Sent: Thursday, June 07, 2001 8:56 PM
> To: Jon Viventi
> Cc: John Oliver; mrtg
> Subject: [mrtg] Re: Server Specs.
>
>
>
>
> I used to use a dual xeon 400, with 512mb of ram, and ultra-wide scsi
> disks, and i used to poll about 115,000 interfaces every 5 minutes.
> we ran 5 processes in parallel, or it took about 11-12 minutes to complete
> a cycle.
>
> the OS was freebsd.
>
> that was a couple of years ago, though.
>
> Recently I had another server that was running on two 10gig ultra-66 IDE
> disks, running linux, that was performing horrendously, and crashing a lot.
> I upgraded the disks to 2 18g ultra wide SCSI disks, and the box has been
> performing flawlessly since.
> This box was mosltly running a webserver, though it did some other stuff.
> Nothing that should have put enough load on it to matter, though.
>
> just reaffirms my long standing prejudice against IDE drives.
> (no flames, please)
>
> It does sound like the box should be performing better, but I can't
> really say what might be the problem.
>
> What OS are you using?
>
> -Tim
>
> On Thu, 07 Jun 2001, Jon Viventi spake thus:
>
> >
> > One of the mrtg processes sometimes takes the full 5 minutes to run, so I
> > get an error message when the next one attempts to run. I've fixed that
> for
> > now by spreading them out over time, and I can probably squeeze more onto
> > the box by splitting up that one customer, but I dont think much more. Its
> > pretty well at 60-80% cpu all the time and the drive is going constantly.
> >
> > -----Original Message-----
> > From: joliver at ee.ethz.ch [mailto:joliver at ee.ethz.ch]On Behalf Of John
> >
> > > Right now I have a PIII 600 desktop w/ 384meg RAM and 20gig ultra 66 IDE
> > > drive. It seems to have reached its limit at 2300 interfaces.
> >
> > Have you monitored that machine to see what's giving out at ~2300
> > interfaces?
> >
> --
> Timothy Kennedy
> Network Operations Manager
> YellowBrix, Inc.
> tkennedy at yellowbrix.com
> Cell: 703-216-5957
> PGP Fingerprint: 07 F1 31 99 89 BC 91 AD 08 E5 B5 67 51 1B 53 49
>
> --
> 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
> WebAdmin http://www.ee.ethz.ch/~slist/lsg2.cgi
>
>
> --
> 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
> WebAdmin http://www.ee.ethz.ch/~slist/lsg2.cgi
>
--
-------------- -- ---- ---- --- - - - - - -- - - - - - -
Tony Bourke tony at vegan.net
--
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
WebAdmin http://www.ee.ethz.ch/~slist/lsg2.cgi
More information about the mrtg
mailing list