From: Howard C. Berkowitz (hcb@gettcomm.com)
Date: Sun Jan 11 2004 - 13:34:56 GMT-3
At 8:32 PM -0800 1/10/04, Richard Davidson wrote:
>I have not done this but could you run another ospf
>process on one the area 0 on r5 and redistribute it
>into area 2?
>Rich
I'm not sure I'm saying the same thing, but in the real world, with
this configuration in which there is no nonzero area connected to
both partitions, I'd not try to fix the partition, but run two OSPF
processes, each with its own area 0.0.0.0.
When I've done this for production, I've usually found that I get
very nice control not by redistributing directly, but by implementing
a backbone-of-backbones with static addresses (if the addressing is
clean) or BGP. This scales much better when you need more than two
OSPF domains. Offhand, the only two-domain production system I
designed had separate domains for Eastern and Western Canada. More
often, I did a domain-per-continent or multiple-domains-per-continent
for multinational companies.
The US Postal Service (not my design, but that of a respected
colleague) has many OSPF domains linked by a BGP backbone. Region and
area assignments fall very naturally out of the 3-2-4 ZIP code
hierarchy.
>--- Paul Chen <cpjchen@starhub.net.sg> wrote:
>> Hi,
>>
>>
>>
>> Say, I have the following set up :
>>
>>
>>
>> Area 10 ----R1----- area 0 ---- R2----- area 2 - ---
>> R5--- area 0
>>
>>
>>
>> The regular method to mend a partitioned area 0 is
>> to use a virtual link.
>> The next method would probably be the use of tunnels
>> .
>>
>>
>>
>> Just a hypothetical question :
>>
>>
>>
>> Q. What happens if both methods are NOT allowed ?
>> What else could we
> > consider doing to mend the partitioned area 0 ?
This archive was generated by hypermail 2.1.4 : Mon Feb 02 2004 - 09:07:39 GMT-3