From: ying chang (ying_c@xxxxxxxxxxx)
Date: Sun May 05 2002 - 12:40:35 GMT-3
Hi Mas & Jag,
No, I don't think you'll lose redundancy with regular redistribution.
Doyle's case is more or less artificial so he can lead us to use distance.
If I think the way you are going to apply the filters without the distance,
the only problem would be suboptimal routing. In stead of say we use #1 or
#2, I think what we should do is to combine both #1 and #2, i.e. raise the
distance so the redistribute routes are not preferred over the local learned
routes.
The case in the book is for achiving optimal routing with redundancy. If
suboptimal is not an issue here, either way would be fine. But if we are
required to have the optimal routing and the redundacy, we'll have to
control the backup routes by distance to save the routes at the background
instead just get rid off them by the filters. His topology is like below.
Without any filters, r5 would take ospf E2 route r5-r2-r3-r6 route to reach
net 6.0, so at r6 filters are added to solve this problem but losing the
redundancy r5-r6 link is cut.
R2---4.0----R3
| |
| |
3.0 OSPF 5.0
| |
===== R5 ======== R6 ====== <<< Redist points
| /
2.0 RIP 6.0
| /
| /
1.0---R1------
Chang
>From: "Mas Kato" <loomis_towcar@speedracer.com>
>Reply-To: "Mas Kato" <loomis_towcar@speedracer.com>
>To: ccielab@groupstudy.com, prospectccie@yahoo.com, ying_c@hotmail.com,
> loomis_towcar@speedracer.com
>Subject: RE: Redist: Filtering vs Fail-over
>Date: Sun, 5 May 2002 01:19:38 -0700
>
>[demime could not interpret encoding binary - treating as plain text]
>I guess I've gotta re-crack the book. But until then, please help me out.
>
>If I'm injecting only native routes into an external autonomous system at
>two different places, I lose redundancy?
>
>Mas
>
> > "ying chang" <ying_c@hotmail.com> RE: Redist: Filtering vs
>Fail-overDate: Sat, 04 May 2002 23:36:04 -0400
> >
> >I have the book. In Doyle's "Case Study: Multiple Redistribution Points"
>(pp
> >787-794) shows #1 would break the redundancy. If redundancy is the reason
> >why you use multi-redistribution points, you probably should consider #2.
> >
> >Chang
> >
> >
> >>From: "Mas Kato" <loomis_towcar@speedracer.com>
> >>Reply-To: "Mas Kato" <loomis_towcar@speedracer.com>
> >>To: ccielab@groupstudy.com, prospectccie@yahoo.com
> >>Subject: RE: Redist: Filtering vs Fail-over
> >>Date: Sat, 4 May 2002 16:55:21 -0700
> >>
> >>[demime could not interpret encoding binary - treating as plain text]
> >>#1, for me, by far and away because it offers the greatest degree of
> >>control.
> >>
> >>I don't have Doyle's handy at the moment, so I'm having difficulty
>making
> >>the leap between administrative 'distance' and redundancy in a looped
> >>environment. Typically 'distance' is used to arbitrarily prefer routes
>from
> >>a given routing protocol because you know it offers better routes for
>the
> >>given topological constraints.
> >>
> >>Regards,
> >>
> >>Mas Kato
> >>https://ecardfile.com/id/mkato
> >>
> >> >Date: Sat, 4 May 2002 16:09:25 -0700 (PDT)
> >> > Jack S <prospectccie@yahoo.com> Redist: Filtering vs Fail-over
> >>ccielab@groupstudy.comReply-To: Jack S <prospectccie@yahoo.com>
> >> >
> >> >Hi,
> >> >What is the best way to tackle redistribution in a
> >> >topology involving loops? i.e., in a domain with
> >> >multiple redistribution points.
> >> >
> >> >1) Filter all routes so that only routes in that
> >> >domain are propagated
> >> >
> >> >2) Play with 'distance' command as described in
> >> >doyle's book so that redundancy is there in the
> >> >network.
> >> >
> >> >The 1st method is the easiest and the 2nd involves
> >> >careful configuration.
> >> >
> >> >Please advise what method to follow.
> >> >
> >> >Thanks,
> >> >Jack
> >> >
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:58:50 GMT-3