From: Alan Basinger (abasinge@xxxxxxxxxx)
Date: Thu Jan 18 2001 - 13:27:14 GMT-3
I think if I am not mistaken I read that bandwidth and delay are the 2
primary values used in calculating path selection but in the event of a tie
in those values the remaining values are considered?
Alan Basinger
Systems Engineer
SBC DataComm
Houston Texas
abasinge@swbell.net
| |
||| |||
.|||||. .|||||.
.:|||||||||:.:|||||||||:.
C i s c o S y s t e m s
Certified Gold Partner
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
Chuck Larrieu
Sent: Thursday, January 18, 2001 10:14 AM
To: Schmitt, Greg; 'CCIE newsgroup (E-mail)'
Subject: RE: Metrics when redistributing.
I believe the experiment I detail below indicates that in effect it does
not matter. (E)IGRP metric is bandwidth and delay, unless you physically go
in and change them. My experiment indicates that even though the
redistribution command requires numbers, the default calculation ignores
those values.
Chuck
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Schmitt, Greg
Sent: Thursday, January 18, 2001 5:30 AM
To: 'Chuck Larrieu'; 'Alan Basinger'; 'Virnoche, Phil'; 'CCIE newsgroup
(E-mail)'
Subject: RE: Metrics when redistributing.
Greetings,
Just watching this thread, and Alan is correct, you have the load and
reliability switched here. It's probably a typo, but I
wouldn't want any newbies to get confused here.
I can't imagine you would ever want to have your load setting maxed out
(255) and your reliability set to the min. Here is
the explaination of the igrp metric settings from the ios:
r5(config-router)#router ig 10
r5(config-router)#redis os 10 met ?
<1-4294967295> Bandwidth metric in Kbits per second
r5(config-router)#redis os 10 met 10000 ?
<0-4294967295> IGRP delay metric, in 10 microsecond units
r5(config-router)#redis os 10 met 10000 10 ?
<0-255> IGRP reliability metric where 255 is 100% reliable
r5(config-router)#redis os 10 met 10000 10 255 ?
<1-255> IGRP Effective bandwidth metric (Loading) where 255 is 100%
loaded
r5(config-router)#redis os 10 met 10000 10 255 1 ?
<1-4294967295> IGRP MTU of the path
r5(config-router)#redis os 10 met 10000 10 255 1 1500 ?
match Redistribution of OSPF routes
metric Metric for redistributed routes
route-map Route map reference
<cr>
So, the normal setting for an ethernet port would be someting like:
redis os 10 met 10000 10 255 1 1500
Cheers,
Greg Schmitt
Internetwork Solutions Engineer
ThruPoint, Inc. (formerly Total Network Solutions)
Voice: 410-349-9772
Cell: 443-822-5183
Pager: 888-773-0423 or pager.gschmitt@thrupoint.net
e-mail: GSchmitt@thrupoint.net
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
Chuck Larrieu
Sent: Wednesday, January 17, 2001 10:17 PM
To: Alan Basinger; Virnoche, Phil; CCIE newsgroup (E-mail)
Subject: RE: Metrics when redistributing.
Ya got me. Some things I think it better to just be a monkey. The command
calls for numbers. Use standard easy to remember numbers.
Slattery seems to prefer the load = 255 and the reliability = 1
Just before Thanksgiving I posted the following results of an experiment I
did ( posted to the regular groupstudy list ) I dug it out and here is what
I found at that time:
---------------------
begin the previous post
Bandwidth Delay Load Reliability MTU resulting route metric
1000 100 255 255 1500 12,100
1000 50,000 255 255 1500 62,000
1000 100 255 1 1500 12,100
1000 100 1 255 1500 12,100
1000 100 1 255 1 12,100
2000 100 1 255 1 8,576
it would appear, then, that ( as we have been told by Cisco ) that only
bandwidth and delay are the driving force in determining the route metric.
Slattery's book contains a formula for calculating the metric, but I can't
get the numbers to match up. There seems to be a constant of 11,000 that I
can't account for. And while it is intuitively clear that higher bandwidth,
as configured in the last line of the table, should lead to a lower metric,
I would expect that twice the bandwidth would result in half the metric, not
2/3's of it.
EIGRP works in a similar manner, so I presume that the calculations are
roughly the same, and that the load, reliability, and MTU factors are of no
import, unless one has manually changed the K values using the metric
weights command.
Hmmmm..... now that's interesting. I thought these values had to match on al
routers in the AS. Just changed the defaults on one router, and all routes
are still appearing, even after multiple clear ip route * commands.
--------------------
end of previous post
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of Alan
Basinger
Sent: Wednesday, January 17, 2001 6:28 PM
To: Chuck Larrieu; Virnoche, Phil; CCIE newsgroup (E-mail)
Subject: RE: Metrics when redistributing.
Chuck isn't it 1 load (min) and 255 reliability (max). That usually is the
case in most scenarios and in most of the information I have read including
EIGRP network design. I would assume you would need to know the metrics of
the protocol your distributing from and to? Also it depends on if you want
to manipulate which routes or paths are preferred in that particular routing
protocol. For example. In rip you inject a route that is 10 hops away but
adjust the metrics to make it appear 2 hops away to RIP so RIP will use the
T3 instead of the 56k link. (This is probably not something one would want
to do without understanding the consequences of an adjustment like this but
you get the idea). I know in BGP you use these to manipulate path selection
and can do the same in IGP's. But then I could be completely wrong :) Hope
this helps.
I have pulled my hair out many a night trying to memorize and understand the
metrics and how they effects path selection for all the routing protocols
and decided screw it I'd rather just look them up haha. Now if I could only
find them all in one place on the doc CD life would be grand. Can any of you
out there help on that one??
Alan Basinger
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
Chuck Larrieu
Sent: Wednesday, January 17, 2001 4:51 PM
To: Virnoche, Phil; CCIE newsgroup (E-mail)
Subject: RE: Metrics when redistributing.
The Slattery / Burton book Advanced IP Routing in Cisco Networks is a bit
better in explaining things. While both Slattery and Caslow, in many
instances, assume far more background than does Doyle, in some cases
Slattery offers some very good examples and discussion.
A couple of comments below:
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Virnoche, Phil
Sent: Wednesday, January 17, 2001 11:19 AM
To: CCIE newsgroup (E-mail)
Subject: Metrics when redistributing.
Question: When redistributing routes into OSPF and EIGRP, how does one
determine the metrics to use? Is there any good documention that exists that
eplains this?
CL: in terms of good practice, one should probably not be too fancy, but
should stick with accurately representing the network topology However,
experienced engineers do use metrics to manipulate or otherwise influence
routing policy.
Caslow's book only says that it is necessary to provide metrics but is very
lacking on how these are determined.
FOR EXAMPLE:
router eigrp 1
redistribute static metric 10000 100 255 1 1500
CL: this places static routes into the EIGRP routing process, and attributes
to them an EIGRP metric based on 10 megabits bandwidth, 100 microsecond
delay, a load of 255 ( max ) a reliability of 1 ( minimum ), and an MTU of
1500 bytes If you are in to self inflicted pain, you can use these numbers
to calculate what shows up in the routing table in the metric column.
router ospf 1
redistribute eigrp 96 metric 20 metric-type 1 subnets
CL: this places EIGRP originated routes into the OSPF process as external
type 1 routes ( as the route passes through the OSPF domain, OSPF metrics
are added to the redistribution assigned metric of 20 ) and all subnets are
passed into the OSPF domain as well. ( as opposed to a summary, or not at
all if you omit the "subnets" keyword. ) In OSPF the metric is cost, which
in turn is 10^8 / bandwidth. The metric of 20 means the administrator is
telling OSPF to consider the EIGRP originated routes as the equivalent of a
5 megabit line, or a bit more than three T1's.
Danka !
Philip G. Virnoche CCNA
Network Engineer - AT&T Wireless
phone: 425.580.5239
cell: 206.601.3134
"HAM AND EGGS - A day's work for a chicken; A lifetime commitment for a
pig."
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:27:34 GMT-3