From: Howard C. Berkowitz (hcb@xxxxxxxxxxxx)
Date: Fri May 17 2002 - 14:41:22 GMT-3
At 12:47 PM -0400 5/17/02, Lupi, Guy wrote:
>I was wondering if there was ever a study done to figure out the practical
>limit of LSAs and routes that can exist in an OSPF domain and not have a
>crippling effect on routing and convergence. This would be assuming a core
>with high end routers (GSR, 7513's), and lower end stub routers in NSSA
>area's. If anyone has any opinions or links/documents they would like to
>share it would be much appreciated. I have read a few books and I don't
>recall seeing any hard numbers listed.
There's not a simple number, but there are ways to approach it
quantitatively. The network topology, stability, and route mixture
has a lot to do with it.
First, remember that the Dijkstra only applies to intra-area routes.
The processor workload for the Dijkstra proper is proportional, in a
good implementation, to (number of subnets * log(number of routers)).
In less good implementations, the load may go up with the number of
subnets, with the worst case being ((subnets**2)*log(routers)).
Intra-area and external LSAs increase the load linearily rather than
exponentially, so the degree of stubbiness and the like will affect
the processing load. These are rarely the cause of peak loads, which
are more likely due to an excessive number of intra-area LSAs.
But this is just the computation. A critical element is how often
the routing computation has to be rerun, due to route flapping.
That's one of the major reasons for limiting the propagation of
routes outside an area, to avoid frequent global recomputation
through the OSPF domain. The absolute number of routes is much less
the issue.
The preceding is another reason to keep area 0.0.0.0 small, because
it MUST be aware of all intra-area and external routes (yeah, there
can be some exceptions), so you want to minimize the workload of area
0.0.0.0 keeping track of its own resources.
There is some formal work on convergence benchmarking just starting
at http://www.ietf.org/html.charters/bmwg-charter.html
HTH.
-- "What Problem are you trying to solve?" ***send Cisco questions to the list, so all can benefit -- not directly to me*** ******************************************************************************* * Howard C. Berkowitz hcb@gettcomm.com Chief Technology Officer, GettLab/Gett Communications http://www.gettlabs.com Technical Director, CertificationZone.com http://www.certificationzone.com "retired" Certified Cisco Systems Instructor (CID) #93005
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:59:00 GMT-3