On Thu, 5 Feb 1998, Scott A. Keoseyan wrote:
> I noticed some examples for doing this using EIGRP were posted on this
> list... has anyone done this using more than 2 T1s (assuming all paths
> equal)?
>
> The reason I ask, is because I have the EIGRP solution implemented using
> three T1s, and noticed that the 3rd T1 does not get totally included in
> the load-balancing using this strategy... I was really wondering if anyone
> has had any better luck with this using 3 equal paths...
>
> For instance...
>
> Interface #1
> 5 minute input rate 576000 bits/sec, 125 packets/sec
> 5 minute output rate 97000 bits/sec, 144 packets/sec
>
> Inteface #2
> 5 minute input rate 577000 bits/sec, 124 packets/sec
> 5 minute output rate 93000 bits/sec, 145 packets/sec
>
> Interface #3
> 5 minute input rate 463000 bits/sec, 100 packets/sec
> 5 minute output rate 79000 bits/sec, 116 packets/sec
>
> Now, this isn't as severe as it has been... I have seen interface #3 at
> only 50% of what the other 2 are doing before... it seems to vary now and
> then, but I don't know why... all interfaces are configured identically
> on both sides... the links are essentially error free... the 'no ip
> route-cache' is implemented on all the interfaces as well.
>
> Here is how I understood it... It should be that traffic gets spread
> across the links on a per-packet basis, but maybe not??? Would there be
> another way it (eigrp) is handling the traffic across the 3 links... like
> maybe based upon connection or something to that effect???
>
>
> Scott
>
if you'll do:
sh ip ro n.n.n.n
You'll get:
Routing entry for n.n.n.n/m.m.m.m
Known via "some protocol", distance a, metric b [, candidate default path]
[Redistributing via ......]
Routing Descriptor Blocks:
a.a.a.a
Route metric is U, traffic share count is 1
a.a.a.b
Route metric is U, traffic share count is 1
* a.a.a.c
Route metric is U, traffic share count is 1
the "*" means that next packet will go trough this link.
the share count is the relative number of packets to
send on this interface, if a.a.a.a and a.a.a.b were T1's
while a.a.a.c was a 3Mb interface, you would want to see there
"traffic share count is 2" for a.a.a.c
If it's not load balanced enough in your eyes, I would suggest
that you'll check that ALL routes are the same on all the 3 links,
---------------------^^^
I.E. if one route to f.g.h.0 is only via a.a.a.a and a.a.a.b,
it will obviously not load a.a.a.c ... and depending on the traffic
you might see this unequal load you mention.
BTW, it gets even more funny with EIRGP unequal path load balancing ;-)
Thanks,
Alain.
P.S. If NT is the answer then you don't ask the right question .
Alain Golan Goldberg, Alain@goldberg.org.il, Alon@Golan.net
Golan Network consulting http://www.golan.net
Web pager: http://wwp.mirabilis.com/88888
Tel: +972-4-8234042
Fax: +972-4-8234118
Cel: +972-50-290135
UIN: 88888
nic-hdl: AGG / AG-RIPE
"I love well designed chaos" Sixtus dixit.
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:13:15 EDT