Re: Can area routing reduce the routing traffic in OSPF?

["Liu B." <[e-mail]>]

Thanks, Vishwas. So, can I conclude as this: a low routing cost in OSPF
mainly depends on manually configured prefix lists (to limit redistribution
of AS external LSAs) and long refresh period. In other words, in a dynamic
network environment (not OPSF domain itself, i.e., AS external LSAs are
highly frequently generated), OSPF routing protocol is not suitable in term
of its not negligible routing cost in that case.

Any experience on how to control routing cost in real network administration
is more than appreciated, as well as literatures.

Thanks again for everybody's time and help.

yours
Bin

-----Original Message-----
From: Manral, Vishwas [mailto:[e-mail]]
Sent: 10 August 2002 10:45
To: [e-mail]
Subject: Re: Can area routing reduce the routing traffic in OSPF?


Hi Bin,

I will try to answer to ur questions.

In order for heirarchical routing to scale the amount of information between
domains is kept at the minimum. Not all information from EGP is injected to
IGP and vice versa.

Here is something from one of the mails I had exchanged: -

" It seems to be an error to encourage redistribution from other protocols
into IGP's. It has been observed every now and again that due to some policy
errors that pumping in routes from EGP to IGP causes the entire EGP routing
table to be redistributed, which causes instability in the IGP domain.

Also unconstrained redistribution from IGP to EGP could cause unwanted
exposure of internal domains topology.

It is therefore suggested that the primary mechanism to redistribute routes
from and to IGP should be by manually configured prefix lists."

Yes, area boundaries do not prevent the excess AS-External LSA's, which are
flooded as vectors thru out the OSPF domain, however areas do modularize the
OSPF domain into smaller units, and prevents flooding of other types(besides
type-5 and type-11) of LSA's across these units. There are also some areas

The refresh period for OSPF is 30 minutes, we can randomize refresh and can
disperse the routing traffic caused by refreshes.

Thanks,
Vishwas

-----Original Message-----
From: Bin Liu [mailto:[e-mail]]
Sent: Friday, August 09, 2002 10:06 PM
To: [e-mail]
Subject: Can area routing reduce the routing traffic in OSPF?


Hello there,

Even though area routing is proposed to reduce the routing traffic, I quite
doubt its effect. OSPF is mainly applied for for transit AS in term of its
ability to accommodate a large number of external routes. However, as we
know, each AS external LSA floods throughout the whole network
transparently. So when the majority of LSAs in the database of OSPF router
are AS external LSAs, which dominate the amount of link bandwidth consumed
by OSPF traffic [OSPF protocol analysis], the benefit of reducing routing
traffic by dividing network into areas is seriously undermined.

If my inference is right, then
1. What is the benefit of area routing, which makes routing not flexible.
2. Does that the longer update period in OSPF (i.e., 30 minutes between the
origination of LSAs) accomplish the work, i.e., reduce routing traffic.

Many thanks and looking forward to seeing reply.

yours
Bin Liu

ps. Relevent information
In 1991, the number of external LSAs in 15 router NASA Science Internet
(NSI) is 496, in another 14 router BARRNet, the number is 1816 [experience
with OSPF, RFC 1246].