From: Jason Sinclair (sinclairj@xxxxxxxxxxxxxxx)
Date: Wed May 29 2002 - 20:12:42 GMT-3
Carlos,
Sorry - typo! You are 100% right. The load is calculated every 5 secs NOT
mins.
Thanks for picking that up.
Cheers,
Jason Sinclair CCIE #9100
Manager, Network Control Centre
POWERTEL
Ground Level, 55 Clarence Street,
SYDNEY NSW 2000
AUSTRALIA
office: + 61 2 8264 3820
mobile: + 61 416 105 858
* sinclairj@powertel.com.au
-----Original Message-----
From: Carlos G Mendioroz [mailto:tron@huapi.ba.ar]
Sent: Wednesday, 29 May 2002 19:45
To: Jason Sinclair
Cc: 'kris.keen@aon.com.au'; Nick Shah;
ccielab@groupstudy.com; John Mistichelli; Joe Morabito;
nobody@groupstudy.com
Subject: Re: using LOAD for IGRP
Jason,
the interface load is computed every 5 secs, not 5 minutes.
The load average is computed using the following formula:
average = ((average - interval) * exp (-t/C)) + interval
where t = 5 seconds and C = 5 minutes, exp (-5/60*5)) =
.983
and interval is the 5 seconds measured load.
EIGRP load using is a bit tricky regarding updates. AFAIK,
the load
will be used by the router (provided k2 != 0) to compute the
best (or acceptable metric when variance > 1) path always,
but
flash updates will be sent to peers only when the routes
actually change
because of the load changing.
(i.e. the router will not be communicating the load changes
to the
neighbours
all the time, but only when it affects its own routing)
Remember that the routers compute the metrics based on
minimums/maximums/totals
of some parameters on the whole path (the update does not
carry the
metric).
BTW, courseware says (and instructors duly repeat :-) that
the metric
also
uses MTU, but I don't understand how...
FTR, the metric is calculated by:
Metric = K1 * BandW + (K2 * BandW)/(256-load) +
K3*Delay
and modified by
Metric = Metric * [K5/(reliability + K4)]
if K5 != 0.
Jason Sinclair wrote:
>
> All,
>
> Some more info regarding this - load statistics on an
interface are gathered
> every 5 minutes and depending on the load-interval
configured on the
> interface will determine how accurate your load
calculations are.
>
> Cheers,
>
> Jason Sinclair CCIE #9100
> Manager, Network Control Centre
> POWERTEL
> Ground Level, 55 Clarence Street,
> SYDNEY NSW 2000
> AUSTRALIA
> office: + 61 2 8264 3820
> mobile: + 61 416 105 858
> * sinclairj@powertel.com.au
>
> -----Original Message-----
> From: kris.keen@aon.com.au
[mailto:kris.keen@aon.com.au]
> Sent: Wednesday, 29 May 2002 10:24
> To: Nick Shah
> Cc: ccielab@groupstudy.com; John
Mistichelli; Joe
> Morabito; nobody@groupstudy.com
> Subject: Re: using LOAD for IGRP
>
> correct.
>
>
>
-----------------------------------------------------------------
> Kris Keen - CCDA, CCNP, CNE
> Network Support Specialist - Network
Systems
> Aon Risk Services Australia Limited
> (612) 9253 7272
> 0404862970
> E: Kris.Keen@aon.com.au
>
>
>
> "Nick Shah"
>
> <nshah@connect
To: "John
> Mistichelli" <jmistichelli@yahoo.com>, "Joe Morabito"
> .com.au>
> <joe.morabito@cox.net>, <ccielab@groupstudy.com>
> Sent by:
cc:
>
> nobody@groupst
bcc:
>
> udy.com
Subject: Re:
> using LOAD for IGRP
>
>
>
>
> 29/05/2002
>
> 09:46 AM
>
> Please respond
>
> to "Nick Shah"
>
>
>
>
>
> If I am not mistaken, Joe's Q was that is
bandwidth required
> to be
> specified
> on the interface if load was to be taken
into
> consideration(by specifying
> metric weights 0 1 1 1 0 0), to which I
had replied that if
> actual b/w is
> specified, load is calculated according to
that, otherwise
> load is
> calculated based on default b/w (eg. 1544
for serial)..
>
> Correct ?
>
> Thanks
> Nick
> ----- Original Message -----
> From: "John Mistichelli"
<jmistichelli@yahoo.com>
> To: "Nick Shah" <nshah@connect.com.au>;
"Joe Morabito"
> <joe.morabito@cox.net>;
<ccielab@groupstudy.com>
> Sent: Wednesday, May 29, 2002 5:17 AM
> Subject: Re: using LOAD for IGRP
>
> > Hey Guys,
> >
> > The equation is as follows:
> > Metric = K1 * BandW + (K2 *
BandW)/(256-load) +
> > K3*Delay
> >
> > The values default to:
> >
> > K1=1
> > K2=0
> > K3=1
> > K4=0
> > K5=0
> >
> > If you set K2=1 load will be taken into
account.
> > Try the command:
> > metric weights 0 1 1 1 0 0
> >
> > Reference:
http://www.cisco.com/warp/public/103/3.html
> >
> > Hope that helps
> > John
> > 7536
> >
> >
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:59:11 GMT-3