[c-nsp] 7500 PPPoE dCEF aggregation

[Joe Maimon]

Niels Bakker wrote:

>* jmaimon at ttec.com (Joe Maimon) [Sun 16 Jan 2005, 20:54 CET]:
>  
>
>>Being that the 7500 is the largest distributed platform (that I know of) 
>>that supports the enterprise/fw feature set (jk9o3sv) does anyone else 
>>have an interest in seeing PPPoE aggregation become dCEF instead of 
>>being CEF punted to the RSP?
>>
>>Cisco claims nobody wants it and has junked previous feature requests.
>>    
>>
>
>VIPs have puny CPUs and would be real bad at it.  Recent RSPs have much
>beefier CPUs on them, it makes sense to do that processing there.
>
>I believe that's more like the rationale given out by Cisco...
>
>
>	-- Niels.
>  
>
Thanks for the response.

I wish I could share your enthusiasm for the only recent RSP, the RSP16 
which has a CPU double the RSP4 at 400Mhz IIRC. Oh and the system only 
has one of those in active use at any single time.

The VIP2-50 has the same CPU as the RSP4.

I guess I will find out soon how much "higher" the RSP16 scales.

As it stands, both GRE and L2TP are dCEF on the 7500 platform, I dont 
see how PPPoE which has only a 8 byte header compared to GRE's 24 would 
be so much harder.

Joe