NIC performance on netgate RCC-VE 2440

Cybertoy

Hi,

I have a brand new Netgate RCC-VE 2440 that I ordered with CentOS on it. The network it's connected to is cat7 cabling with GB switch. Under CentOS I was measuring the full performance of roughly 115MB/s throughput. Once I installed pfsense on it though that number went down to about 40MB/s. In order to measure I'm using a combination of dd and nc commands.

Are there any special settings within pfsense that I need to configure in order to get the full GB performance of the NIC?

regards,
Cybertoy

cmb

What specifically are you doing to test?

You'll want to enable powerd and set it to hiadaptive or maximum if you haven't already.

Cybertoy

Thanks for that quick reply!

The exact way I measure is:

On another linux box on the network I issue this command:

nc -v -l 2222 > /dev/null

Then on the pfSense box:

dd if=/dev/zero bs=1024K count=512 | nc -v <theotherbox> 2222</theotherbox>

Thanks for the powerd tip. That has now increased the throughput to about 60MB/s … still not the maximum though. I have also tried the "maximum" setting on the powerd but that didn't improve it anymore.

regards,
Cybertoy

heper

you are not measuring throughput. you are measuring speed towards/from the pfsense itself.

Cybertoy

@heper:

you are not measuring throughput. you are measuring speed towards/from the pfsense itself.

Sorry … you're right. That got lost in translation. I am concerned that on the local LAN I don't get 1gpbs. Once I have that I will start to work on the throughput.

Doesn't change the problem though that I'm not getting >100MB/s.

regards,
Cybertoy

Guest

I have a brand new Netgate RCC-VE 2440 that I ordered with CentOS on it.

Did you replace the CentOS with pfSense image from the ADI account after registering the RCC-VE 2440?
Or did you take the ordinary community pfSense image?

The network it's connected to is cat7 cabling with GB switch.

Ok, but in normal CAT.7 is coming together with other jacks or plugs at the ends of the cable
so what is this in real? CAT.6 or CAT.6a will be more then sufficient for that in realy life a CAT.5e
wire will do that job with ease.

Under CentOS I was measuring the full performance of roughly 115MB/s throughput.

Hence, this is a Linux OS and pfSense is based on FreeBSD and this will not be the same, please accept this.
The second thing is that CentOS is not doing any kind of NAT and also no firewall rules must be worked through.

Once I installed pfsense on it though that number went down to about 40MB/s.

On what pfSense was installed directly? (eMMC storage, mSATA, SSD, HDD, SATA-DOM,…..)
Again, what pfSense image you were taking for that? (community image or ADI image)
What is the Internet connecting method? (PPPoE, static public IP,....)

In order to measure I'm using a combination of dd and nc commands.

iPerf or NetIO from one PC or server to another one as the client would be showing you up real
numbers that are protocol independent. So you are measuring the speed of the hardware it self
and more from the eMMC storage then the throughput from the NICs.

Are there any special settings within pfsense that I need to configure in order to get the full GB performance of the NIC?

At first what is full GB performance for you? pfSense is a software firewall and need much more power
to realize this likes a normal software router such DD-WRT or OpenWRT for sure. It has to work out firewall
rules (pf) where NAT is a part from internally of pf. And owed to some other things this number can really
easy be different from other number shown by the same hardware in such cases. So more input please!

Is this a PPPoE connection?
pfSense is at this time only using one CPU core at PPPoE
Inside of the pfSense ADI image customizing and tunings where already done!
Please use the ADI image from the Netgate or pfSense store after the hardware registration through the account!
PowerD (hi adaptive) should be set on
mbuf size can be high up but is not a must be!
NIC tuning can be done also by your own according to the pfSense Doc´s.

At last, comparing Linux without NAT and firewall rules against FreeBSD based pfSense with turned this
on is not really the best show down as I see it right.

Once I have that I will start to work on the throughput.

What exactly? LAN or WAN 1 GBit/s is here talked about?
On LAN or WAN you will be better sorted with netIO or iPerf from client to server method as you
where doing it. So what is now the problem? Please do a fresh install, configure the WAN and LAN part
and then connect two PCs or Laptops and measure through the pfSense box, thats it.

Doesn't change the problem though that I'm not getting >100MB/s.

Where now exactly please. Is this the LAN or WAN throughput we are talking about?
With a C2758 SoC (SG-8860) you can read here what they got as a real throughput!
WAN throughput
LAN throughput
iPerf test C2758

And if we are both talking here about the WAN throughput of 1 GBit/s you should perhaps also willing in
to read this statement shown under the link at first that explains what kind of GHz you will need to archive
a real 1 GBit/s at the WAN port. CPU Selection

You have a RCC-VE 2440 and this might be more comparable to the SG-2440 with an Intel "Rangeley"
Atom C2358 1.7 GHz. So at the WAN port you will see definitely something nearly 1 GBit/s for sure and
with ease, but at the WAN port may not really.

Cybertoy

Thanks for the detailed reply. I will work through these items and come back around Monday as I'm out of town. Just quickly though:

I'm using the ADI community version as documented here: https://netgate.com/docs/rcc-ve-2440/pfsense.html
I am doing this on the LAN interface only

What I will do is run the test with pf disabled so that it's not working any rules. It shouldn't as I don't have any on the LAN.

regards,
Cybertoy