Actual throughput vs reported

Aaron Toponce aaron.toponce at gmail.com
Mon Oct 8 19:21:45 BST 2012


On Monday, 01 October, 2012 22:00:56, Aaron Toponce wrote:
> It appears that the 5 keys cannot keep the enttropy pool filled faster than
> ~10 KBps, 1/2 of what is expected.

So, after playing around a bit, the performance of the keys greatly depends
on the hardware of the host. The Raspberry Pi is limited to a 700 Mhz proc.
My x86 server is an Intel Celeron Coppermine at 700 Mhz also, however, it
manages to process the keys at ~12 KBps. When on my Lenovo T61, with is a
dual-core Intel, it can process the keys at ~20 KBps.

The interesting thing is the load on each. With the Pi, the load was 1.4
while ekeyd was running, and dropped when not. On the x86 server, the load
is .06. On the laptop, .03. Knowing that internally, the keys encrypt the
packets to the kernel, I'm guessing this is where the slowdown is. The
faster the CPU can decrypt the stream, the more packets you'll get out of
the keys.

Anyway, FYI.

-- 
. o .   o . o   . . o   o . .   . o .
. . o   . o o   o . o   . o o   . . o
o o o   . o .   . o o   o o .   o o o
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 519 bytes
Desc: not available
URL: <http://lists.simtec.co.uk/pipermail/entropykey-users/attachments/20121008/a5ae295e/attachment.pgp>


More information about the EntropyKey-users mailing list