Subject: Re: CryptoGraphic Disk.
To: Andrew Gillham <gillham@vaultron.com>
From: Roland Dowdeswell <elric@imrryr.org>
List: current-users
Date: 10/04/2002 18:35:14
On 1033765235 seconds since the Beginning of the UNIX epoch
Andrew Gillham wrote:
>

>Very cool, thanks Roland.  Do you have any benchmark results available?

No problem.

Here's some random bonnie numbers.  I used 256 bit keys for both
AES and blowfish.  Please note that I just did these quickly now
and did not utilise good experimental techniques.  The i386 is my
workstation running X and so has some load on it (such as typing
this email.)  For the sparc64 test, I accidentally used different
sizes with bonnie, so the seek times are completely invalid.

All that said, it does give some basic idea of the performance.
I'll come up with some better number in single-user mode and all
that over the weekend.  :-)

Also, please note that 256 bit AES is substantially slower than
128 bit or 192 bit.

i386 (P4/1.7GHz, 256MB RAM)

              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
plain     280 22198 29.5 23103  9.8  6247  4.7 17158 36.3 16320  7.8  91.3  1.9
aes-cbc   280 12122 79.5 14426 82.0  5572 33.5  8607 20.7  9629  4.3 113.7  2.5
blowfish  280 16826 77.4 21105 78.8  5857 23.7 12012 28.4 14577  6.8 127.3  2.6

note: the i386 has all manner of interactive programs running on it, so
      the number are much more suspect.  (I did turn off xmms, but that
      was it.)

sparc64 (Ultra5, 128MB RAM)

              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
plain     256  4594 34.1  4626  6.6  1814  3.3  4606 46.2  4703  9.0  42.6
aes-cbc   100  1387 96.8  1481 100.5  698 47.1  1449  9.0  1195  2.5  70.7
blowfish  100  3875 96.1  4600 89.8  1612 24.1  3276 35.0  2882  5.5 123.7

note: the sparc64 reported quite odd things for %CPU on random seeks, so I
      eliminated them.  The file sizes are also different, skewing the results.

--
    Roland Dowdeswell                      http://www.Imrryr.ORG/~elric/