[LAU] encryption and performance penalty

[David Christensen]

On 11/14/2014 07:58 AM, William Light wrote:
> I'm using luks/LVM on my Thinkpad T430s, using a -ck kernel and I've got
> 16Gb of RAM. I don't notice or perceive a slowdown at all, even though I
> bet e.g. startup would be a little snappier without encryption on. I
> haven't A/B tested it, encryption is just a prerequisite for my systems
> these days, and the performance is still excellent.


Nice.  That machine has a Core i7-2530M processor, which also has Secure 
Key (e.g. hardware pseudo-random number generator, PRNG):

 
http://www.amazon.com/Lenovo-ThinkPad-T430s-23539WU-Notebook/dp/B009UESY2I

 
http://ark.intel.com/products/64893/Intel-Core-i7-3520M-Processor-4M-Cache-up-to-3_60-GHz?q=Intel%C2%AE%20Core%E2%84%A2%20i7-3520M%20Processor%20%284M%20Cache,%20up%20to%203.60%20GHz%29


So, you should be able to generate pseudo-random numbers very quickly. 
This will expedite erasing encrypted partitions and devices, generating 
white noise, and other uses.


Here's a crude PRN benchmark for my Pentium 4 3.4E GHz HT machine:

     2014-11-14 16:58:52 dpchrist at p43400e ~
     $ dd if=/dev/urandom of=/dev/null bs=1M count=100
     100+0 records in
     100+0 records out
     104857600 bytes (105 MB) copied, 13.0431 s, 8.0 MB/s


And, for my Core i7-2600S machine:

     2014-11-14 17:00:09 dpchrist at i72600s ~
     $ dd if=/dev/urandom of=/dev/null bs=1M count=100
     100+0 records in
     100+0 records out
     104857600 bytes (105 MB) copied, 6.90355 s, 15.2 MB/s


(It appears that /dev/urandom uses a single thread, as the Pentium as 
one core and the i7 has four, but the i7 isn't four times faster.)


How well does your laptop perform on the above /dev/urandom benchmark?


/dev/random blocks waiting for entropy on both of the above machines. 
How about your laptop?


David