AirVPN performance better than expected.

[Ernst89 11]

When deciding which pfSense router hardware to use with AirVPN I like to test CPU/System performance to know what the limits of the hardware are. I normally do this on my LAN so I can be sure to eliminate network problems.

 

The basic method I use is to set a simple OpenVPN tunnel between two machines using the shell command line openvpn.

 

I then use iperf3 to measure the throughput of the tunnel.

 

This works for my router, pfSense, Intel Celeron N3150. I get a test result of 127Mb/s, slighy higher if I used the rdrand engine ~134Mb/s.

 

All well and good. This is the router hardware I now run my pfsense connection to airvpn with.

 

The problem is, and it is a silly problem, is that the performance to AirVPN is much better than my theoretical LAN maximum. I get 155Mb/s, which could be limited by my ISP rather than router CPU.

 

I have checked all the OpenVPN parameters between the tunnel to AirVPN and the tunnel on my LAN and they look similar enough apart from AirVPN is using dyname/TLS keys.

 

Obviously my test is wrong, but I just can't see what is wrong with it. The only way I can get my LAN openvpn tunnel to work as fast as the WAN one to AirVPN is to raise tun-mtu to about 1900. But I'm pretty sure Airvpn use 1500.

 

I don't expect anyone will be able to help but there is always hoping ;o)

[Ernst89 11]

Apologies I intended to post this to  the problems page.

 

The fact that AirVPN is not behaving as expected is a worry to me. My primary concern is that encryption implemented correctly.

 

It may seem good that Airvpn is faster than expected, but it would be easy to achieve that result by not encrypting at all.

 

The way to build trust in AirVPN is to understand what it is doing. At the present I do not understand. 

[zhang888 1066]

There is no way to know without knowing the server specs of your OpenVPN server. If it's the same Celeron, obviously your LAN-LAN performance

would be lower than the speed you can get on WAN while connected to high performance Xeon servers used by Air.

Also, make sure you adjust the buffer sizes in your test connection to be as close to the Air config you are using.

[Ernst89 11]

I'm using a Haswell and an AMD 5350 both of which are faster than the N3150.

 

I cross check them against each other to prove this is the case.

 

What I need to be able to do is understand the AirVPN OpenVPN tunnel, is there a spec somewhere?.

 

I have run with the logs at verb 4 to get 200+ parameters and whilst some are different the significant ones seem to be the same.

 

Does anyone else have alternative methods of testing?