On Fri, 2007-08-31 at 15:32 -0400, Dan McDonald wrote:
> On Fri, Aug 31, 2007 at 12:24:44PM -0700, Garrett D'Amore wrote:
>
> <SNIP!>
>
> > We found that "in-band" processing of IPsec with SCA 4000 was a lot more
> > efficient. I'd really, really like someday to see this kind of offload
> > properly supported in the stack, without having to resort to crufty DLPI
> > mechanisms.
>
> If we get the generic hooks right, then as hardware becomes available, it can
> be exploited quicker since driver writers will know what to do. IIRC, the
> CPU speeds caught up with the SCA 4000 before we really had all of the bits
> in S9 ready for it.
Actually, I think what happened is that the world switched to AES.
I suspect the SCA 4000 3DES performance is probably *still* faster than
typical commodity CPUs. (Certainly than SPARCs.)
The SCA 4000 had either no or greatly reduced AES performance. (I don't
recall whether the final version had a 5823 on it or not... we certainly
had talked about it.)
-- Garrett
_______________________________________________
networking-discuss mailing list
[email protected]
