[c-nsp] FWSM and IPv6

[Justin M. Streiner]

On Thu, 24 Jun 2010, Benjamin Lovell wrote:

> Did a quick search and found that IPv6 packet are still processed by the CPU 
> not the ASICs on the FWSM. Also only works in routed, not transparent mode. I 
> don't know any hard numbers for forwarding performance for the CPU but I 
> would guess it's unacceptably low.
>
> As for software versus architecture limitation, it's tough to say. I would 
> guess architectural limitation but either way I doubt it will change as FWSM 
> is near the end of it's life cycle for new feature development.

I haven't done any real-world tests because the vast majority of our FWSMs 
are running in transparent mode, but an SE I spoke with seemed to feel 
pretty strongly that the performance would be unacceptably low.

I do hope that the new product that will succeed the FWSM isn't bound by 
the many of the same arcane limitations that hobbled the FWSM (ACL memory 
limitations, >5.5 Gb/s connection to the 6500 backplane, etc).  I haven't 
seen anything on it in awhile, so I'm assuming that new product is still 
under NDA.

jms