Prev: Nexus 7000 Question
Next: Looking for int's? 1MFT T1, VIC 2FXS, WIC 1DSU, WIC 1T link's inside!
From: russianj on 10 Apr 2008 15:42
I have two 3845 routers with IOS version 12.4(2)T1. The T1s are
connected to a private MPLS cloud used by my company for remote sites.
On the LAN side of the routers, HSRP is employed with one of the
routers, called RTR01 used as primary, with the other, RTR02 as
secondary for failover. On the WAN side, the routers broadcast EIGRP
for LAN-connected subnets.
Originally, all traffic flowed in and out of RTR01's T1 and RTR02 had
no load. EIGRP was setup so that RTR01 had a better metric for Static
(we use heavy weighted static routes to our dialup router when the
MPLS connection goes down at a remote site) as follows:
On RTR01
router eigrp 100
redistribute static 1500 10 255 1 1500
On RTR02
router eigrp 100
redistribute static 1000 10 255 1 1500
In an attempt to balance traffic across the two T1s and the two
routers, I set the metric to "1500 10 255 1 1500". Now after the
change, all incoming traffic flows through the T1 on RTR01 and all
outbound traffic flows on the T1 out RTR02. Unfortunately, I am losing
bandwidth by not having both incoming and outgoing traffic across the
two T1s and routers.
What is the next step? I am a little lost as to why the network
would work this way when two routes are being given. Should I try it
without any metrics and hope that it balances properly? Thanks.
From: Merv on 10 Apr 2008 16:03
start again ...

For outbound, take a look at GLBP - Gateway Load Balancing Protocol

For inbound, ask You provider if they can support EIBGP load
balancing (note the EIBGP part - it is important)

From: russianj on 11 Apr 2008 07:50
On Apr 10, 3:03 pm, Merv <merv.hr...(a)rogers.com> wrote:
> start again ...
>
> For outbound, take a look at GLBP - Gateway Load Balancing Protocol
>
> For inbound, ask You provider if they can support EIBGP load
> balancing  (note the EIBGP part - it is important)

The scenario that you are talking about, I believe, is to attempt to
load balance the individual packets. That is not necessarily what I am
going for. Currently, if I fill up either of the two T1s, the
additional traffic that attempts to go across does not get balanced to
the additional T1s. That is more specifically the problem. If I
understand EIGRP correctly, then if metrics are not specified, then as
connections use up bandwidth on one of the T1s, then the metric for
the other T1 will be more favorable, and traffic will go through the
second T1 accordingly. That is how EIGRP is designed to work, correct?
From: Merv on 11 Apr 2008 09:41
On Apr 11, 7:50 am, russi...(a)gmail.com wrote:
> On Apr 10, 3:03 pm, Merv <merv.hr...(a)rogers.com> wrote:
>
> > start again ...
>
> > For outbound, take a look at GLBP - Gateway Load Balancing Protocol
>
> > For inbound, ask You provider if they can support EIBGP load
> > balancing (note the EIBGP part - it is important)
>
> The scenario that you are talking about, I believe, is to attempt to
> load balance the individual packets. That is not necessarily what I am
> going for. Currently, if I fill up either of the two T1s, the
> additional traffic that attempts to go across does not get balanced to
> the additional T1s. That is more specifically the problem. If I
> understand EIGRP correctly, then if metrics are not specified, then as
> connections use up bandwidth on one of the T1s, then the metric for
> the other T1 will be more favorable, and traffic will go through the
> second T1 accordingly. That is how EIGRP is designed to work, correct?

nope

the default value for the "k" variables are K1=1, K2 =0, K3=1,K4=0,
K5=0

so link load does not come into play

the bandwidth referred to in metric calcluation is a fix value of the
interface bandwidth - it is not dynamic link utilization
From: Merv on 11 Apr 2008 18:00
On Apr 11, 9:41 am, Merv <merv.hr...(a)rogers.com> wrote:
> On Apr 11, 7:50 am, russi...(a)gmail.com wrote:
>
>
>
> > On Apr 10, 3:03 pm, Merv <merv.hr...(a)rogers.com> wrote:
>
> > > start again ...
>
> > > For outbound, take a look at GLBP - Gateway Load Balancing Protocol
>
> > > For inbound, ask You provider if they can support EIBGP load
> > > balancing (note the EIBGP part - it is important)
>
> > The scenario that you are talking about, I believe, is to attempt to
> > load balance the individual packets. That is not necessarily what I am
> > going for. Currently, if I fill up either of the two T1s, the
> > additional traffic that attempts to go across does not get balanced to
> > the additional T1s. That is more specifically the problem. If I
> > understand EIGRP correctly, then if metrics are not specified, then as
> > connections use up bandwidth on one of the T1s, then the metric for
> > the other T1 will be more favorable, and traffic will go through the
> > second T1 accordingly. That is how EIGRP is designed to work, correct?
>
> nope
>
> the default value for the "k" variables are K1=1, K2 =0, K3=1,K4=0,
> K5=0
>
> so link load does not come into play
>
> the bandwidth referred to in metric calcluation is a fix value of the
> interface bandwidth - it is not dynamic link utilization


Forgot to mention that K2, K4 and K5 were meaning for IGRP which used
periodic updates

They are not useful for EIGRP for which updates are event driven

 |  Next  |  Last
Pages: 1 2
Prev: Nexus 7000 Question
Next: Looking for int's? 1MFT T1, VIC 2FXS, WIC 1DSU, WIC 1T link's inside!