[c-nsp] IS-IS Topology database

[Mark Tinka]

On Monday 29 September 2008 14:11:28 Oliver Boehmer 
(oboehmer) wrote:

> a few comments after taking a quick look:
>
> SPF and PRC-interval are quite aggressive. 1 msec initial
> wait is appropriate for single link failures, but if you
> have multiple failures within a short time frame (for
> example SRLG- or node-failures), you might need to run
> two SPFs, so many deployments use 50ms initial wait. 20
> msec interval is quite low, some folks' SPF takes much
> longer than this. So I would consider increasing this.
> Same reasoning applies to lsp-gen-interval, for SRLG
> failures you might need to generate two LSPs.

Our initial implementation of these were based on Cisco best 
practices documents. Our tests (both link and node failure 
across 6 routers [NPE-G1/G2]) worked great.

However, thanks for the feedback. I'll modify the values and 
lab another (more) aggressive test as soon as I can.

> Not sure if I would consider "ignore-attached-bit" a Best
> Practice.. It is useful in some environments, but many
> others would rely on it.

Our deployment is based on using IS-IS to carry Loopback and 
infrastructure addresses only. iBGP deals with the rest of 
the prefixes.

As such, we need end-to-end reachability of the (provider) 
edge routers across the entire network. This means longest 
match for /32's (v4) and /128's (v6).

Since we practice route leaking for IS-IS, and all 
non-infrastructure routes are carried in iBGP, we ignore 
the ATT bit in the network. Besides, we wouldn't want to 
run the risk of a default route in the network trying to 
handle traffic in a way we have not predicted.

But, as you say, some deployments may find it useful.

> "log-adjacency-changes all" generates some more log files
> (for example adjacency down when you shut an interface)..

> AS you tune for fast convergence, "process-max-time 50"
> and "ip routing protocol purge interface" (or "ip
> slow-converge" in non-12.0S/non-12.2S trains) would also
> be recommended.

Thanks.

Will get some time to test these out too.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20080929/5982a53c/attachment-0001.bin>