Poor pfSense benchmarking performance, any guesses why ?

zeluis

Hello everybody.

I've been doing some benchmarking and the results above are quite surprising and i'm lost for an explanation why ...

I've been using iperf 2.0.11 / 2.0.13

I've done an isolated setup, has the diagram illustrates, all direct connections between NICs. The first 2 tables are averages for 1/2/4/8 threads run, the bottom 2 are the individual results.

The biggest shockers, if you will, are the pfsense vs ubuntu performance diference in point to point. and the pfsense point to point (2,5 Gbps) to pfsense NAT (3,2 Gbps).

I'm not sure if i have made myself clear, but can anybody provide some tips on why i got these results.

Thanks
Jose

stephenw10

Not sure what you expect to see there. pfSense as one host in an iperf test will always show a bad result. It's not optimised as a TCP server but as a router. You should always test through it.

Steve

zeluis

@stephenw10 thanks for your input

The first 2 tables are represeting of iperf measurements between the 2 i7 and the pfSense doing the NAT.
The last 2 was to illustrate the difference in behavior for the same hardware but different operating systems.

It's such a big difference that i have a feeling that i'm missing something, but i don't know what, i tryed everything i can think of.

thanks

mer

@zeluis
What version of pfSense? I could have missed it, but I don't see it.
maybe post the rules?
Are you using any packages? If so, which ones?

zeluis

@mer pfsense 2.5.1, allow everything in all interfaces and i just installed vm-tools package in the virtualized instance.

thanks

stephenw10

Right, so in the 3rd table you are using pfSense as one side of the iperf test directly. That will always give a bad result.