[Swan] Using AES-NI doesn't improve performance in most cases
Andreas Herz
andi at geekosphere.org
Mon Feb 3 13:15:58 EET 2014
On 03/02/14 at 02:42, Alex Elsayed wrote:
> Well, your throughput may not be increasing in the tests you ran, but what's
> the load like? It may be that the crypto isn't the _bottleneck_ (and thus
> making the crypto faster won't result in better benchmarks), but saving
> resources that might otherwise be spent on crypto for other things may still
> be valuable.
I agree, but why is this one special case (2.6.32 KLIPS AES-NI) that is
increasing the throughput on the same machine, same setup as in the
other 3 centos scenarios?
That is what confuses me :)
--
Andreas Herz
More information about the Swan
mailing list