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

Re: [Openvpn-users] Bandwith usage

  • Subject: Re: [Openvpn-users] Bandwith usage
  • From: James Yonan <jim@xxxxxxxxx>
  • Date: Wed, 30 Mar 2005 02:58:52 -0700 (MST)

On Wed, 30 Mar 2005, Patrick Marquetecken wrote:

> > On Wednesday 30 March 2005 10:00, Patrick Marquetecken wrote:
> >> Hi,
> >>
> >> I have setup a test lan to see if openvpn is a good solution for us.
> >> I dit tranfers and stress test with and without openvpn.
> >> Without openvpn we can use the compleet 100mbit bandwith, but with
> >> openvpn it goes to a max of 60%, i supose its because of the
> >> compression. Is there a way to use the other 40% of our 100bmit
> >> connection, i find it a bit waste not to use it, or will heavy usage
> >> on the tunnel enlarge the bandwith usage ?
> >
> > Without knowing exactly how you performed the test, it's tough to
> > understand why you are experiencing this discrepancy.
> >
> > A pretty simple (albeit not very realistic for production environments)
> > test would be to transfer a large (say 500MB) file via FTP or HTTP,
> > once outside the tunnel and once through the OpenVPN tunnel.
> >
> > In case you do see significant differences in transfer rates, you should
> > then monitor the machines that act as VPN gateways - perhaps they are
> > not powerful enough to handle the load. Although I must say that I
> > haven't experienced any slowdowns through tunnels implemented with
> > <300MHz x86 machines.
> >
> > -A
> I have used a program called LanTrafic witch creates trafic in all kind
> block sizes and file sizes.
> We have done a test with 16 simultane (the max of the program) connections
> and moved about 100GB of data.
> So we saw the bandwith usage is max 50% - 60% of a 100mbit connection
> between two machines with a direct cable.
> The machine is a pentium4 3GB ram, dual proc and dual GBit ethernet
> connecto tors the same machine.

That's about what I would expect.  With this kind of test, you will max 
out the processor capacity before you hit network saturation.

In most cases this is non-representative of real-world usage where the 
bottleneck will be in network throughput long before the CPU utilization 
hits 100%.

Also, given that you are running on a dual processor machine, you could 
set up two OpenVPN daemons on different ports and have clients randomly 
connect to either one using two remote directives on the client.  This 
would give you some load balancing across your dual CPUs and probably get 
you close to saturating your 100mbit connection.


Openvpn-users mailing list