openssh unix dev - Nov 2008 - OpenSSH performance with VIA padlock

Hello list,
please spend a minute considering these facts and maybe there is something to 
improve:

1) VIA Eden based board can write AES256 encrypted information on HDD at > 
60MB/s
2) iperf shows NIC speed 69MB/s
3) openssl tests have even better results
4) openssh can transfer AES256 encrypted information at < 27MB/s (and worse 
with HDD encryption)

It is better with openssh 5.x, than 4.6, but still I see no reason why results 
are so bad... Please, could someone explain? I don't believe, padlock is
used
properly...

Jan

On Tue, Nov 18, 2008 at 17:28:11 +0200, Jan Klod wrote:
> Hello list,
> please spend a minute considering these facts and maybe there is something
to
> improve:
> 
> 1) VIA Eden based board can write AES256 encrypted information on HDD at
>
> 60MB/s
> 2) iperf shows NIC speed 69MB/s
> 3) openssl tests have even better results
> 4) openssh can transfer AES256 encrypted information at < 27MB/s (and
worse
> with HDD encryption)
> 
> It is better with openssh 5.x, than 4.6, but still I see no reason why
results
> are so bad... Please, could someone explain? I don't believe, padlock
is used
> properly...
> 
> Jan
Remember, cycles are also taken up by the message digest. You don't
mention which MAC you are using nor do you give any of the OpenSSL speed
numbers for AES or any of the digests.

Also, what performance do you get for a memory-to-memory transfer over
the loopback?

-- 
Iain Morgan

On Tue, 18 Nov 2008, Jan Klod wrote:
>Hello list,
>please spend a minute considering these facts and maybe there is something
to
>improve:
>
>1) VIA Eden based board can write AES256 encrypted information on HDD at
>
>60MB/s
>2) iperf shows NIC speed 69MB/s
>3) openssl tests have even better results
>4) openssh can transfer AES256 encrypted information at < 27MB/s (and
worse
>with HDD encryption)
>
>It is better with openssh 5.x, than 4.6, but still I see no reason why
results
>are so bad... Please, could someone explain? I don't believe, padlock is
used
>properly...
	
	after the initial key exchange, the processing of SSH packets is not 
about encryption/decryption only but also about HMAC. MD5/SHA-* are much 
faster in software than AES but if AES goes to HW and HMAC stays in SW it 
can make a significant impact on the overall numbers when compared to 
OpenSSL speed, for example.

	also, the packet (1 cipher block) length is decrypted independently 
from the rest of the packet in OpenSSH. Usually, working with small blocks 
is much slower in HW than in SW due to inherent overhead of offloading 
anything to HW. In general, the size of blocks offloaded is very important - 
your benchmark numbers might be affected by different block sizes used.

	those 2 things mentioned above might be the reason why you see 1/2 
of what you have expected.

	J.

-- 
Jan Pechanec

Jan Klod wrote:
> Hello list,
> please spend a minute considering these facts and maybe there is something
to
> improve:
> 
> 1) VIA Eden based board can write AES256 encrypted information on HDD at
>
> 60MB/s
> 2) iperf shows NIC speed 69MB/s
> 3) openssl tests have even better results
> 4) openssh can transfer AES256 encrypted information at < 27MB/s (and
worse
> with HDD encryption)
> 
> It is better with openssh 5.x, than 4.6, but still I see no reason why
results
> are so bad... Please, could someone explain? I don't believe, padlock
is used
> properly...
Did you enable OpenSSL engine support when you built OpenSSH? 
(./configure  --with-ssl-engine).

Try this with 5.1p1, apparently some earlier versions didn't work 
properly (I have no such hardware to test).

Also as someone else mentioned, try a faster MAC such as umac64.

-- 
Darren Tucker (dtucker at zip.com.au)
GPG key 8FF4FA69 / D9A3 86E9 7EEE AF4B B2D4  37C9 C982 80C7 8FF4 FA69
     Good judgement comes with experience. Unfortunately, the experience
usually comes from bad judgement.