variance - another EIGRP question

Hello All,

What is the meaning of variance ? I know that it enables unequal load
balancing in EIGRP but what it really does ? What if I enter variance
128 (max value) instead of calculating it ? It looks to me that it
works just fine.

After some labbing I noticed that I can use variance to eliminate
slower link from being load balanced when more than two links exist.
See example below. I can however achieve the same result with
"maximum-paths".

Here is an example:

    R1
/ | \
R2 R3 R4
|------|------|

R1-R2: 1,5 mb/s p2p 10.1.12.0/24
R1-R3: 512 kb/s p2p 10.1.13.0/24
R1-R4: 16 kb/s p2p 10.1.14.0/24
R2-R3-R4: Ethernet 10.1.123.0/24

Calculated variance between link R1-R2 and R1-R4 is 74. If I enter on
R1 variance 73 only two faster link will work:

I played with various variances on R1 with following results

Case #1: Variance 73:
Sh ip route:
    10.0.0.0/24 is subnetted, 4 subnets
C 10.1.14.0 is directly connected, Serial1/0.14
C 10.1.13.0 is directly connected, Serial1/0.13
C 10.1.12.0 is directly connected, Serial1/0.12
D 10.1.123.0 [90/5537536] via 10.1.13.3, 00:00:01, Serial1/0.13
                  [90/2195456] via 10.1.12.2, 00:00:01, Serial1/0.12

Case #2: variance 74:
    10.0.0.0/24 is subnetted, 4 subnets
C 10.1.14.0 is directly connected, Serial1/0.14
C 10.1.13.0 is directly connected, Serial1/0.13
C 10.1.12.0 is directly connected, Serial1/0.12
D 10.1.123.0 [90/160537600] via 10.1.14.4, 00:00:02, Serial1/0.14
                  [90/5537536] via 10.1.13.3, 00:00:02, Serial1/0.13
                  [90/2195456] via 10.1.12.2, 00:00:02, Serial1/0.12

Case #: variance 128:
    10.0.0.0/24 is subnetted, 4 subnets
C 10.1.14.0 is directly connected, Serial1/0.14
C 10.1.13.0 is directly connected, Serial1/0.13
C 10.1.12.0 is directly connected, Serial1/0.12
D 10.1.123.0 [90/160537600] via 10.1.14.4, 00:00:02, Serial1/0.14
                  [90/5537536] via 10.1.13.3, 00:00:02, Serial1/0.13
                  [90/2195456] via 10.1.12.2, 00:00:02, Serial1/0.12

Blogs and organic groups at http://www.ccie.net
Received on Thu Oct 07 2010 - 16:49:25 ART