[Openswan dev] Re: [Openswan Users] Hardware encryption support - VIA PadLock?

[Nate Carlson]

On Tue, 22 Jun 2004, Nate Carlson wrote:
> Actually tested things out now, AES is working great.
> 
> I'll have to do some speed tests without a kernel supporting the padlock
> stuff, and then test again with a kernel supporting the padlock code.  
> Hopefully the 1ghz processor isn't fast enough to do AES for 100mbit
> without padlock, or else it'll be difficult to test.  :)

someone replied offlist about this with a good suggestion - since they 
were offlist, not sure if they want to say who they are (speak up if you 
don't care), but here's their comment:

"You can always measure how much CPU resources you have _left_:
http://www.zip.com.au/~akpm/linux/#zc"

This is an excellent program - I just tested it out on my gateway, with 
some interesting results, copying a large file to a box behind the 
gateway at 3mbit/sec:

No IPsec:	1.5% cpu load
IPsec AES:	8.1% cpu load
IPsec 3des:	18.0% cpu load

This is on the VIA C3 Nehemiah 1.0ghz processor, without the patches in
the kernel to support the hardware encryption.

Nice big difference between AES and 3des!

------------------------------------------------------------------------
| nate carlson | natecars at natecarlson.com | http://www.natecarlson.com |
|       depriving some poor village of its idiot since 1981            |
------------------------------------------------------------------------