[OpenVPN home] [Date Prev] [Date Index] [Date Next]
[OpenVPN mailing lists] [Thread Prev] [Thread Index] [Thread Next]
Web openvpn.net

Re: [Openvpn-users] Hardware acceleration

  • Subject: Re: [Openvpn-users] Hardware acceleration
  • From: Mike Tancsa <mike@xxxxxxxxxx>
  • Date: Wed, 15 Aug 2007 09:37:37 -0400

At 08:57 AM 8/15/2007, Prasanna Krishnamoorthy wrote:
>On 8/15/07, nobledark@xxxxxxxxxxxx <nobledark@xxxxxxxxxxxx> wrote:
> > Hi again,
> >
> > I've seen a couple of threads on this topic but could use some more
> > info - how does one utilize hardware acceleration with OpenVPN?
> > Does it make a significant difference vs a modern cpu? Ideas on
> > equipment, etc are appreciated.
> >
>It does make a significant difference. For instance a Via C7 with hw
>encryption is supposed to provide a significant difference, compared
>to any modern processor. However, I was unable to get it to work. :-(.

The place where it can make a difference, even on modern hardware is 
where you have multiple streams.  For example, on our backup server, 
we have a dozen ssh sessions coming in at once.  Offloading the 
crpyto to the hifn card helps keep the CPU down a lot.  On embedded 
devices, it doesnt make a difference, and can slow things down a 
little if you have just one operation happening.

e.g. on a straight ftp fetch on FreeBSD, across an open vpn tunnel, 
with the aes-128 being offloaded to the hifn card, the fetch is 
actually a little slower

With Crypto offload
fetch -o /dev/null
/dev/null                                     100% of   26 MB  352 kBps 00m00s

4801, no crpto card

# fetch -o /dev/null
/dev/null                                     100% of   26 MB  396 kBps 00m00s

The variation is about +-10kB/s. AES-256 might be different. I will 
see if I can try and do some benchmarks later as well as track CPU 
utilization to give a better idea.


Openvpn-users mailing list