[c-nsp] Cat6500 modular IOS - direction?

[Phil Mayers]

On 08/28/2010 02:23 AM, Nick Hilliard wrote:
>
> On the other hand, it is disappointing that the sup2t has taken so long to
> materialise.

Right. That lengthy arrival has made some of the "higher performance" 
figures for the new device look very weak. 80gig/slot on a 9-slot 
chassis is still only 64 line-rate 10gig ports - woo, how exciting.

As for the original question:

I note that we've been told "monolithic is being phased out" a couple of 
times; it was said about SXH, more so about SXI and the publicly 
accessible sup2t info says it as well. We all know how the first two 
panned out. I suspect Cisco have learned their lesson.

I suspect modular IOS is dead, in the form we were originally promised 
it. That is to say, the model for resilience and incremental software 
upgrade going forward is dual-sup / ISSU-based. Any modularity features 
that *do* make it into IOS will be purely for the IOS developers 
benefit, and will provide little specific features that users can 
access. The dual-sup model has obvious profit benefits for Cisco.

Sadly, my experience with dual-sup for resilience thus far has been 
so-so. Since SXI and MPLS SSO, we've had a couple of crashes which 
thanks to SSO meant no loss of service. But we've had a couple more 
where either a) FIB corruption propagated with the SSO, so a reload was 
eventually necessary or b) the SSO event crashed both sups, dropping the 
box to ROMMON. So it's ok, not great. This is unfortunate given the high 
cost a 2nd sup entails; for that money I'd want slightly better 
reliability. We're aiming to complete our dual-attachment programme and 
then let HSRP/STP solve the problem, and ditch/re-use the 2nd sup.

Still, sup2T is still a little way off; who knows what it will 
eventually boot? Only Cisco.

Cheers,
Phil