I've done it a number of different ways, depending on the circumstance. But
to address the two solutions that you propose...I already saw Chuck's post
about the need to actually advertise the loopback addresses, so I'll must
make the assumption that you DO need to - If not, then assign them in your
OSPF broadcast networks in the order that you want DR/BDR election and only
advertise them locally (intra-area), and design the hierarchy to your
convenience.
Solution 1. Advantage - easier to summarize across areas. You can still do
the area range just for the loopbacks, so that 10 loopbacks in an OSPF area
are summarized back to area 0.0.0.0, which in turn sends summaries from the
other areas.
Solution 2. Advantage - probably easier to administer all the loopback
addresses out of one central 'pool'. It's the way that I've done it in the
past - take a /22 or /23 block of address space and 'reserve' it for device
loopback interfaces.
Drawback - lots of host routes everywhere. (When I'm really on the game, I
can 'regionalize' those loopbacks and create summary routes using EIGRP - a
little harder to do in OSPF unless they're across areas, but you can plan
the loopback addresses regionally a little more carefully)
One of the members of this list wrote a book about IP Addressing design - he
probably has an entire section on how and why's of advertising loopbacks (I
don't know, I haven't bought the book yet), and I remember him posting about
being NOT fond of /32's for loopback addresses (I believe his recommded
maximum was /30)
the best question I could ask is, Which way would you rather manage the
scheme, and which would make the most sense for your operation?
----- Original Message -----
From: "COULOMBE, TROY"
To:
Sent: Friday, September 14, 2001 12:30 PM
Subject: Loopback Address scheme (long) [7:20002]
> Alright gang,
>
> Usually I just lurk. However, for those of you doing support for
networks,
> I would like to know (sample) your loopback address scheme. We have a Hub
&
> spoke network, with our core network consisting of core services (server
> farm, etc) and our two WAN routers. Off these WAN routers are our regions
> (and off them branches). Currently there is little redundancy (as far as
it
> being meshed--we have redundant links/isdn backup, etc). We are a purely
> OSPF shop.
>
> So, Core/WAN is Area 0.0.0.0 & Regions/Branches are areas x.x.x.x (Each
> (region and their branches) are separate areas...) However, due to
> transitions, summarization is not practical (things have moved within the
> corp.)
>
> Thoughts on BEST PRACTICE of loopback addressing??? Should I take a /32
> from within the major network of each region/branch & use that, or should
I
> use one class C (and /32 that) and just spread it amongst the sites?
Also,
> take into consideration that we are moving towards a more partial meshed
> network.
>
> Confused? See examples below
>
> Core network:
> 10.1.0.0 /16 (summarized)
> 10.1.1.0 /24 (server farm)
> Region 1
> 10.32.0.0 /16 (not summarized)
> 10.32.1.0 /24 (local net 1)
> Branch 1
> 10.32.1.0 /24 (local net)
> Branch 2
> 10.32.2.0 /24 (local net)
> Region 2
> 10.44.0.0 /16 (not summarized--but in general this is the major
> network we are going to)
> 10.44.1.0 /24 (local net 1)
> 10.44.2.0 /24 (local net 2)
> Branch 1
> 10.32.7.0 /24 (local net) **** see can't summarize above due
> to discontiguous nets ;- Branch 2
> 10.44.217.0 /24 (local net)
> ==========================================================
> so solution #1 would be
>
> core network devices use addresses (like)
> 10.1.2.2 /32 (using a new network here!!!!)
> 10.1.2.3 /32 (using a new network here!!!!)
> etc etc.
> Region 1
> 10.32.2.2 /32 (using a new network here!!!!)
> 10.32.2.3 /32 (using a new network here!!!!)
> Branch 1
> 10.32.3.2 /32 (using a new network here!!!!)
> Branch 2
> 10.32.4.2 /32 (using a new network here!!!!)
> Region 2
> 10.44.3.2 /32 (using a new network here!!!!)
> 10.44.3.3 /32 (using a new network here!!!!)
> Branch 1
> 10.32.8.2 /32 (using a new network here!!!!)
> Branch 2
> 10.44.218.2 /32 (using a new network here!!!!)
> ==========================================================
> or solution #2 would be using ONE loopback network...
>
> core network devices use addresses (like)
> 10.99.99.2 /32
> 10.99.99.3 /32
> etc etc.
> Region 1
> 10.99.99.4 /32
> 10.99.99.5 /32
> Branch 1
> 10.99.99.6 /32
> Branch 2
> 10.99.99.7 /32
> Region 2
> 10.99.99.8 /32
> 10.99.99.9 /32
> Branch 1
> 10.99.99.10 /32
> Branch 2
> 10.99.99.11 /32
> ==============================================================
>
>
> and maybe I am just sweating the small stuff & it really doesn't
> matter...just pondering it, because quite truthfully, I seem to only do
> Hub&Spoke OSPF networks...all my mesh/part. mesh networks were EIGRP...
>
> and I figured I'd share this as it might also be a prob/solution that get
> peoples minds on off the tragedy.
>
> TroyC
> -=-=-=-
> With sufficient thrust, pigs fly just fine.
> ---RFC 1925
_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=20024&t=20002
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
