>To get the DUAL algorithm to operate more accurately you can enable
>the other metrics. The only metrics enabled by default are bandwidth
>and delay. Reliability, Load, MTU, and Hop count, can be added into
>the Composite Metric for path determination. I have heard you should
>really know what you are doing when you try this. I would think that
>doing something like this would need to be done on all the routers
>in this EIGRP AS. Jeff Doyle's book is pretty good for EIGRP.
>
>>>>Brian
Well, if I can assume the role of someone who really knows what he's
doing, several comments.
First, while (E)IGRP is aware of MTU and hop count, they are never
part of the metric calculation, although they are used for other
purposes.
Second, I have NEVER found a good reason to touch load. The
consensus among routing protocol architects is that considering load
on the next hop medium is a bad optimization that often leads to
route oscillation. Current thinking in traffic-aware routing is that
load can reasonably be considered only in terms of (loosely speaking)
end to end path.
I have used reliability, but it was an old application with IGRP and
there are better ways today to solve the specific problem with which
I was working.
Even there, traffic engineering is often a matter of preallocating
resources for the favored traffic, not trying to respond dynamically
to load. There's been work on OSPF load-sensitive routing in the IETF
OSPF working group archives and probably some experimental RFCs.
Look for QOSPF (don't remember the authors, although I seem to recall
they were at AT&T and some universities), and the OSPF (and ISIS)
Optimized Multipath work by Curtis Villamizor.
The more networks I design or review, the less I tend to think in
terms of tweaking metrics and the more I think of getting the correct
topology (including restricting the scope of route updates with
OSPF/ISIS areas, BGP route reflectors and confederations, etc.).
I'm not suggesting that metrics don't have an important role. But
many introductions to routing tend to overemphasize them, because the
introductions are focusing in on how the routing protocols select
routes among the potential routes offered to them. In other words,
the routing protocols make the best of what they know, but they don't
necessarily have all information.
In my Designing Routing and Switching Architectures book, you'll find
that I give far more emphasis to topology than metrics.
>n
>
>>From: "Ejay Hire" <[EMAIL PROTECTED]>
>>Reply-To: "Ejay Hire" <[EMAIL PROTECTED]>
>>To: [EMAIL PROTECTED]
>>CC: [EMAIL PROTECTED]
>>Subject: Re: bandwidth statement on Frame Relay interface
>>Date: Thu, 26 Oct 2000 07:11:49 CDT
>>
>>The idea is to have EIGRP accurately select the best path based on the
>>actual traffic that the interface can pass. If you have multiple PVC's that
>>have different CIR's, then a single setting will cause it not to reflect
>>accurate path metrics. It becomes even more complicated if the sum of the
>>CIR's exceedsthe actual bandwidth of the frame-relay delivery mechanism.
>>(t-1...)
>>
>>
>>----Original Message Follows----
>>From: "info" <[EMAIL PROTECTED]>
>>Reply-To: "info" <[EMAIL PROTECTED]>
>>To: [EMAIL PROTECTED]
>>Subject: bandwidth statement on Frame Relay interface
>>Date: Wed, 25 Oct 2000 16:59:39 -0400
>>
>>I've read a couple books that recommend setting the
>>bandwidth equal to the CIR on a Frame subinterface.
>>....as opposed to the port size....in an EIGRP environment.
>>
>>Anyone out there do it differently? Any recommendations
>>or rules of thumb to apply to this issue?
>>
>>
>>
>>
>>_________________________________
>>FAQ, list archives, and subscription info:
>>http://www.groupstudy.com/list/cisco.html
>>Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
>>
>>_________________________________________________________________________
>>Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
>>
>>Share information about yourself, create your own public profile at
>>http://profiles.msn.com.
>>
>>_________________________________
>>FAQ, list archives, and subscription info:
>>http://www.groupstudy.com/list/cisco.html
>>Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
>
>_________________________________________________________________________
>Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
>
>Share information about yourself, create your own public profile at
>http://profiles.msn.com.
>
>_________________________________
>FAQ, list archives, and subscription info:
>http://www.groupstudy.com/list/cisco.html
>Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
