From: chiragshroff (chiragshroff@myrealbox.com)
Date: Fri Jun 13 2003 - 07:37:37 GMT-3
good answer......
regards
chirag
----- Original Message -----
From: Mike Williams <ccie2be@swbell.net>
To: 'John Matijevic' <matijevi@bellsouth.net>
Cc: CCIELab@Groupstudy.com <ccielab@groupstudy.com>
Sent: Friday, June 13, 2003 02:05 AM
Subject: RE: eigrp and variance command
> When you use the variance command, EIGRP will take the minimum metric
> (best) metric for a given destination, then it will consider other
> routers with a higher metric, as long as the higher metric is < (best
> metric) * variance.
>
> Let's say you have three possible routes to the 10.1.1.0/24 network and
> they have EIGRP metrics of 20, 35, and 50. Normally (without the
> variance) the route with the metric of 20 would be installed and that's
> it. If you go and configure a variance of 2, now EIGRP will install the
> routes with metric of 20 and 35 (since 35 is less than or equal to 20 x
> 2 = 40). It will then proceed to send 2 packets over the metric 20 link
> and 1 over the metric 35 link and back and forth.
>
> Here is the output from a site we have. This site uses variance to load
> balance across a 384kbps ADSL circuit and a 768Kbps Frame relay circuit.
>
> Routing entry for 10.1.0.0/16
> Known via "eigrp25", distance 90, metric 3459328, type internal
> Redistributing via eigrp 25
> Last update from 10.63.0.81 on ATM0/0.35, 00:00:05 ago
> Routing Descriptor Blocks:
> 10.63.0.197, from 10.63.0.197, 00:00:05 ago, via Serial0/0.1
> Route metric is 3459328, traffic share count is 2
> Total delay is 4930 microseconds, minimum bandwidth is 768 Kbit
> Reliability 255/255, minimum MTU 1500 bytes
> Loading 10/255, Hops 1
> * 10.63.0.81, from 10.63.0.81, 00:00:05 ago, via ATM0/0.35
> Route metric is 6792704, traffic share count is 1
> Total delay is 4930 microseconds, minimum bandwidth is 384 Kbit
> Reliability 255/255, minimum MTU 1500 bytes
> Loading 31/255, Hops 1
>
> The thing to note here is that it shows the "traffic share count". As
> you can see the 768Kbps link has a share count of 2 and the 384 has a
> share count of one. This means it will send 2 packets down the 768Kbps
> pipe then one down the 384Kbps pipe, just as planned. Just be careful.
> Many times, depending on the interface, just because the bandwidth of
> one link is 2x or 3x the bandwidth of the smaller link, that doesn't
> mean a variance of 2 or 3 will load balance properly. This is because
> the EIGRP metric depends on the interface types. So if you have a
> 768Kbps ATM PVC and a 768Kbps Serial connection, the delay on the ATM is
> 80 and the delay on the serial is 20000(!!) This make a noticable
> difference in the EIGRP metric. So what you need to do is to do a 'sho
> ip eigrp topology <network> <subnet mask>' and analyze the metrics over
> the difference links. Then you must adjust the delay so that the
> metrics are the multiplier you want. So, if you have 2 links and one is
> 2 times the bandwidth of the other but the delays are way off, you may
> want to change the delay on one interface so they are the same on both.
> This will usually make the metrics within 2x of each other, and a
> variance of 2 will work great then.
>
> Good Luck!
>
> Mike W.
>
> -----Original Message-----
> From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
> John Matijevic
> Sent: Thursday, June 12, 2003 12:18 PM
> To: ccielab@groupstudy.com
> Subject: eigrp and variance command
>
>
> Hello,
> I want to better understand the variance command, I know that it is used
> to provide unequal load balancing among mutliple paths, and you specify
> by variance 1-128 metric varinace multiplier. I have read the doc cd
> regarding this and could not get more specific info that I need.
> Basically I would like to know how to verify that the command took
> place, do I need to look at the metrics on the routing tables? Also how
> is the varinace applied to the metric what is the formula? I am
> basically just entering the command and assuming thats its load
> balancing but im not sure how to verify. Any help regarding the matter
> is greatly appreciated. Sincerely, Matijevic
This archive was generated by hypermail 2.1.4 : Fri Jul 04 2003 - 11:10:57 GMT-3