I don't think this is a valid argument.
I *think* that the LSA flooding process is unaffected by this parameter. My
understanding is that the LSAs are still flooded straight away, but this
parameter determines how long before the router recalculates its own OSPF
routing table.
That means that the delay shouldn't be cumulative - the link state database on
each router gets updated regardless of the setting of the spf timers, but the
timers affect how fast the individual routers take note of the new information
in the database (so if several changes happen in quick succession, the router
only has to do the SPF calculation once). If I've misunderstood any of this,
somebody please let me know.
I agree that you wouldn't want the spf timers to be significantly different, for
convergence reasons, but I can't see why a difference of a couple of seconds
would be a problem. The hello timers etc have to match exactly, though, or
hello packets are dropped - OSPF won't function.
JMcL
---------------------- Forwarded by Jenny Mcleod/NSO/CSDA on 15/09/2000 03:26 pm
---------------------------
"Kristopher B. Climie" <[EMAIL PROTECTED]> on 15/09/2000 12:14:08 pm
Please respond to "Kristopher B. Climie" <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
cc: (bcc: JENNY MCLEOD/NSO/CSDA)
Subject: Re: SPF timers
If the routers are not calculating their topology table at the same
interval, it would not take long for their tables to become completely
out-of-whack. For instance, lets say you have 10 routers in your network,
and your spf-delay times vary by 5 seconds on each router -- not a long time
at all. But by the time you get to the 10th router, the delay is off
50-seconds, and its convergence would be worse than RIPs! The only way that
OPSF can be sure that the topology table is consistent among all routers is
that their timers match. Remember, one of the main benefits of OSPF is that
all the routers converge at the same time when a change in topology occurs.
If router A converges, and Router B doesn't converge for another 10 seconds,
Router A cannot be sure of the validity of its own table. OSPF depends on
every router having a valid view of the network, with itself as the root, to
have the most accurate information available to it to make a decision.
K
-----
Kristopher B. Climie, CCNP, CCDP
<[EMAIL PROTECTED]> wrote in message
[EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
>
>
> At
http://www.cisco.com/cpress/cc/td/cpress/design/ospf/on0407.htm#xtocid163652
3
> (which is an extracted chapter from 'OSPF Network Design Solutions', by
Tom
> Thomas), there is a bit that states...
>
> "Cisco's OSPF implementation enables you to alter certain
interface-specific
> OSPF parameters, as needed. You are not required to alter any of these
> parameters, but some interface parameters must be consistent across all
routers
> in an attached network. Those are the parameters set by the following
commands:
>
> ip ospf hello-interval
> ip ospf dead-interval
> ip ospf authentication-key
> timers spf spf-delay spf-holdtime
>
> Therefore, be sure that if you do configure any of these parameters, the
> configurations for all routers on your network have compatible values. "
>
> The first three I can understand, and I don't have a problem with these
> parameters having to match on all routers on the network. But I can't see
why
> the spf timers should have to match. And in any case, that one's not an
> interface-specific parameter.
>
> For those who haven't used this command before, 'spf-delay' is the delay
time,
> in seconds, between when OSPF receives a topology change and when it
starts a
> SPF calculation. 'spf-holdtime' is the minimum time, in seconds, between
two
> consecutive SPF calculations. The command reference on CCO doesn't
mention that
> the spf timers have to match on all routers. I can see that if they are
> mismatched by too much, it will take longer for routers to converge to a
> consistent view of the network, but would it cause any other problems?
>
> I've had a look at RFC 2328, and am no wiser, although I will happily
admit I
> did not read all 240 pages. Why would one router care how long another
router
> has waited between SPF calculations? Or is this an error in the
book/website -
> can they in fact be different everywhere (obviously it's simpler if
they're the
> same, but does it do nasty things to OSPF if they're not)?
>
> JMcL
>
>
> **NOTE: New CCNA/CCDA List has been formed. For more information go to
> http://www.groupstudy.com/list/Associates.html
> _________________________________
> UPDATED Posting Guidelines: http://www.groupstudy.com/list/guide.html
> FAQ, list archives, and subscription info: http://www.groupstudy.com
> Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
>
**NOTE: New CCNA/CCDA List has been formed. For more information go to
http://www.groupstudy.com/list/Associates.html
_________________________________
UPDATED Posting Guidelines: http://www.groupstudy.com/list/guide.html
FAQ, list archives, and subscription info: http://www.groupstudy.com
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
**NOTE: New CCNA/CCDA List has been formed. For more information go to
http://www.groupstudy.com/list/Associates.html
_________________________________
UPDATED Posting Guidelines: http://www.groupstudy.com/list/guide.html
FAQ, list archives, and subscription info: http://www.groupstudy.com
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
