inline :)
.siva
>
>
> > > Granted, but if the packet is VIP to VIP with dCEF and it is Auton/SSE
> > > switched, the packet never hits the RSP, at least thats what I remember.
> >
> > Not quite. Packets still touch the RSP, but the RSP doesn't have to
> > do a cef lookup on the packet. But as far as the path across the
> > cybusses goes, the packet goes from VIPa->RSP->VIPb (or sometimes
> > VIPa->RSP->VIPa, see my other email). It's just Much Faster to get
> > the packet in and out of the RSP, since we don't have to pass the
> > packet up from MEMD to the RSP CPU.
>
> This is all covered very nicely in _Inside Cisco IOS Software
> Architecture_ (Bollapragada et al, Cisco Press 2000, ISBN 1578701813).
> A must-have for readers of this list, despite the fact that they don't
> delve deeply enough into certain dead-end corners of the history of
> router evolution for my tastes (for instance, difference between SP
> and SSP on the 7000 platform) :-)
>
really need to finish reading the rest of the book, and this may already
be covered. the essential difference between the SP and the
SSP is an additional switching feature -
ip route-cache sse
ipx route-cache sse
.
.
etc
the forwarding is cache based (i don't remember if we download the
cache onto the SSE or if the SSE does a lookup in the RP memory,
need to check the code to verify that)
so the SSP contains a superset of the SP switching techniques
(you can also do " <protocol> route-cache cbus " on the SSP.
because of improved forwarding asics on the SSP. the SSP also comes in
2 flavors - with 512 kb of packet memory and 2 megs of packet memory.
the carving of packet memory into interface buffers is very similar to
the way the RSP does it.
in plain vanilla lab forwarding, the SSE can do about 270 kpps.
cheers
.siva
> ---Rob
>
>
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:39 EDT