Re: Cache traffic - odd?

From: Daniel Schroder <daniel@dont-contact.us>
Date: Thu, 4 Sep 1997 16:55:59 +0200 (SAT)

  /sbin/ifconfig on the proxie gave me
  a marginal difference between RX and TX.
   
  Wich kinda makes sense considering that one
  has parents or siblings. An ICBM ping is also
  a packet ?

  So with each request a packet is sent out for
  a query ? I'd be surprised if that were still
  the case with no siblings or parents.

  Does MRTG report packets or data ? ( I must implement
  it when I get a week for my birthday :)

--Daniel Schroder
  Networld
  http://livewire.new.co.za (F) 419 3212
  mailto:daniel@new.co.za (T) 419 4430

On Thu, 4 Sep 1997, Simon Rainey wrote:

> We run ten heavily used squid caches. These are located at three locations.
> Each site has a 100Mbit/s backbone and the servers hang off Cisco Catalyst
> 10/100 switches.
>
> We use MRTG to monitor the traffic levels for each server, taking the data
> from the Catalyst switches.
>
> An analysis of the squid log files shows about 22% less data fetched than
> served. However the MRTG traffic analysis shows about equal traffic flow,
> and sometimes the inbound traffic greatly exceeds the outbound. Some of
> this can be explained by client HTTP requests, sibling UDP queries and DNS
> traffic, but I would have thought that these would be small compared to the
> actual data served.
>
> I don't doubt for a minute the integrity of the squid logs. I'm more
> inclined to mistrust the data I get back from the switches. But then one
> might expect any errors affecting the analysis of traffic in one direction
> to affect the analysis of traffic in the other direction to an equal
> degree. The delta would then be unaffected.
>
> Anyone got any thoughts on this?
>
> Cheers,
> Simon.
>
> At 14:49 04/09/97 +0100, Leigh Porter wrote:
> >Hi folks,
> >
> >We run two pretty well used cache servers and I have noticed
> >some odd traffic stats.
> >
> >The idea of a cahce server is to serve objects locally, an offset of
> >this is that it saves bandwidth.
> >
> >We use MRTG to monitor the ethernet line going to the cache server
> >from a Cisco switch and even though the server gets almost a 30%
> >HIT rate there is often more traffic going OUT TO THE SERVER than
> >there is come FROM THE SERVER - which is odd!
> >
> >Occasionally it will go the other way around.
> >
> >Anybody know why this is or have similar monitors of bandwidth
> >usage they can share!?
> >
>
> -------------------------------------------------------------------------
> Simon Rainey Direct Line: 01235 823238
> Principal Internet Development Engineer Fax: 01235 823424
> RM Internet for Learning E-mail: srainey@rmplc.net
> New Mill House, 183 Milton Park, Abingdon, Oxfordshire, OX14 4SE, England
>
Received on Thu Sep 04 1997 - 07:55:44 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:36:56 MST