Re: newbie: Speed of loop-aes

[Markus Huehnerbein <silencer@xxxxxx>]

As I understand the man-Page of bonnie++ it says that "-s indicates the
 size of the file(s) for IO performance measures in megabytes...". So it
should be 4000 MB. Sorry, I forgot to state that in the mail.

Thanks for the hdparm suggestion I will set hdparm and run the test again.

Gisle Sælensminde wrote:
> Markus Huehnerbein wrote:
>> Hi List,
>>
>> I did some performance tests using the benchmark bonnie++ to evaluate
>> different disk encryption possibilities. The results regarding loop-aes
>> are a bit strange for my understanding. I ran the tests on a notebook
>> with Debian Etch, 1GB RAM and a Linux 2.6.22.2 Kernel.
>>
>> Bonnie was startet with the following parameters:
>> bonnie++ -b -u root -s 4000
>>
>> As a result loop-aes has a sequential input (block) of 29127 K/sec where
>> the sequential input with no encryption is 25123 K/sec. The same with
>> sequential output (block) (loop-aes: 27633 K/sec and no encryption:
>> 24267 K/sec).
>>
>> What I don't understand is how encrypted reads and writes per block
>> could be faster as plaintext reads / writes? Any comments are welcome!
>>
>>   
> Now I know very little about the bonnie++ benchmark, but it says that
> 
> 4MB is way to little to show anything. Factors like caching and process
> overhead (which benchmark did you run first?) and other similar factors
> will dominate. Try with at least 4GB. Since the bonnie program test for
> database-like access according to it's website, it will probably not
> make much sense before you test it with file sizes significantly larger
> than the main memory.
> 
> If you have not set hdparm, sometimes DMA is not used by the kernel.
> This is the safest on some buggy (but common) mainboards, and may be the
> default on your system. If this is set, non-encrypted writing will be
> much slower than it need to be. Read the manual page for hdparm.
> 
> With DMA enabled, the null cipher (that is no encryption) can be as slow
> as an encrypted filesystem, because what slows things down is that the
> block has to be copied into RAM.
> 
>> Kind regards,
>> Markus
>>
>> P.S.: The bonnie++ output:
>> LOOP-AES:
>>  ------Sequential Output------ --Sequential Input- --Random-
>>  -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
>> Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
>> 4000M 18554  59 27633   7 13696   3 18922  77 29127   4 102.3   0
>>
>> NO ENCRYPTION:
>>  ------Sequential Output------ --Sequential Input- --Random-
>>  -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
>> Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
>> 4000M 24085  77 24267   6 11978   3 24825  75 25123   3 103.9   0
>>
>>
>> -
>> Linux-crypto:  cryptography in and on the Linux system
>> Archive:       http://mail.nl.linux.org/linux-crypto/
>>
>>   
> 



-
Linux-crypto:  cryptography in and on the Linux system
Archive:       http://mail.nl.linux.org/linux-crypto/