[c-nsp] 7513 T3 Card Question

[David Coulson]

Pete Templin wrote:
> Step back and analyze: do you really need NPE-G1 (a device rated around 
> 1 million pps, so we'll figure 100kpps real-world) to handle ~6-8kpps? 
> I'm showing about 4kpps on 4xCT3.  Consider NPE-225, and you'll see the 
> price drop significantly.  Plus, you won't need Gig uplinks...but it may 
> impact your scalability calculations.

Certainly don't need a NPE-G1, however I was trying to save bandwidth 
points and slots in my head. I figure that a FE or two isn't going to 
handle the ~500mbit of traffic that is potentially generated from 12 DS-3s.

> Redundant processors means your outage time shrinks from 360 seconds to 
> ~45 seconds, on the rare occasion there's a processor outage.  However, 
> you gain outage time whenever a processor "fakes" an outage, and the 
> spare RSP takes over.

Honestly, never had a 'fake' outage in my time of using RSPs. That said, 
my last NPE failure was a lot longer than 360s because it never came up 
after rebooting.

> Using 6x7206 in the space of 1x7513 means 12xCT3 are impacted, not 
> 36xCT3, and your other 64xCT3 are undisturbed.  That can be a good 
> thing.  Given the simplicity of the code in a 7200 (no need for any dCEF 
> code, no need to accommodate switching vectors for non-dCEF cards, 
> etc.), you're getting a more stable platform in the 7206 than the 7513.
> 
> Nutshell: I want 7206s (NPE-225, 2xOC3 uplinks, 4xPA-MC-2T3) for my 
> ongoing CT3 deployments.

I agree - Splitting stuff up across multiple routers is always a "good 
thing". Right now my 7500s experiences have been happy enough, plus from 
a troubleshooting perspective I've got a bias towards that rather than 
7200 since I'm more familiar with funky things they do and how to get 
around it.

David