Iperf perfomance

[IggyB]

Hi, this is my result on zotac ci323 nano

I'm using realtek driver i compiled in freebsd 10.3 vm then moved over to pfsense

iggy@iggy-mint ~ $ iperf -c 192.168.1.1
–----------------------------------------------------------
Client connecting to 192.168.1.1, TCP port 5001
TCP window size: 85.0 KByte (default)

[  3] local 192.168.1.170 port 37664 connected with 192.168.1.1 port 5001
[ ID] Interval      Transfer    Bandwidth
[  3]  0.0-10.0 sec  600 MBytes  503 Mbits/sec

Does this look low? Should it be higher?

[um3n]

This depends on your hardware and the networkinterfaces, but for a 1GBit interface this looks okay.

Can you give us some informations about your hardware on both machines?

If both machines are virtual you could try to give some more vCPU to the pfsense.

[Derelict]

Try iperf -c 192.168.1.1 -P4

I'm getting about 440Mbit from a mac running all kinds of crap to a 4860 running all kinds of crap . 770M with -P4. I'd like to see 900+ there but as I said these nodes/hosts are running all kinds of crap.

Get 700+ with the 4860 sending and the mac discarding. 740M with -P4

[IggyB]

@um3n:

This depends on your hardware and the networkinterfaces, but for a 1GBit interface this looks okay.

Can you give us some informations about your hardware on both machines?

If both machines are virtual you could try to give some more vCPU to the pfsense.

server is barebone pfSense 2.3.2-1

cpu n3150 4 cores
16gb 1600 ram 2x8gb
2 realtek nics
samsung 850 evo

only got pfBlockerNG running with just Easylist enabled and iperf for testing

client is skylake 6700k
intel nic I219V
16gb ram

setup is adsl2 modem only in bridge > pfSense > tp-link switch < wifi router into switch running dd-wrt for wifi only and this client is connected to wifi router ethernet port because there is a hardware issue currently with asus skylake nics triggering my switch to shut off if i connect it to switch. i'm waiting for tp-link to send me a new a/c adapter which solves this issue.

[IggyB]

@Derelict:

Try iperf -c 192.168.1.1 -P4

I'm getting about 440Mbit from a mac running all kinds of crap to a 4860 running all kinds of crap . 770M with -P4. I'd like to see 900+ there but as I said these nodes/hosts are running all kinds of crap.

Get 700+ with the 4860 sending and the mac discarding. 740M with -P4

This is what i get with that command

iggy@iggy-mint ~ $ iperf -c 192.168.1.1 -C4
iperf: invalid option – 4

Client connecting to 192.168.1.1, TCP port 5001
TCP window size: 85.0 KByte (default)

[  3] local 192.168.1.170 port 37990 connected with 192.168.1.1 port 5001
[ ID] Interval      Transfer    Bandwidth
[  3]  0.0-10.0 sec  603 MBytes  505 Mbits/sec

Invalid option –4?

[doktornotor]

Your cut'n'paste skills suck.

[IggyB]

[ ID] Interval      Transfer    Bandwidth
[  4]  0.0-10.0 sec  151 MBytes  126 Mbits/sec
[  5]  0.0-10.0 sec  161 MBytes  135 Mbits/sec
[  6]  0.0-10.0 sec  158 MBytes  132 Mbits/sec
[  3]  0.0-10.0 sec  159 MBytes  133 Mbits/sec
[SUM]  0.0-10.0 sec  629 MBytes  526 Mbits/sec

with hardware checksum offload enabled

when it's disabled i lose about 20mbit/sec

[um3n]

@doktornotor:

Your cut'n'paste skills suck.

Great - that would help him…

@IggyB:

@Derelict:

Try iperf -c 192.168.1.1 -P4

I'm getting about 440Mbit from a mac running all kinds of crap to a 4860 running all kinds of crap . 770M with -P4. I'd like to see 900+ there but as I said these nodes/hosts are running all kinds of crap.

Get 700+ with the 4860 sending and the mac discarding. 740M with -P4

This is what i get with that command

iggy@iggy-mint ~ $ iperf -c 192.168.1.1 -C4
iperf: invalid option – 4

Client connecting to 192.168.1.1, TCP port 5001
TCP window size: 85.0 KByte (default)

[  3] local 192.168.1.170 port 37990 connected with 192.168.1.1 port 5001
[ ID] Interval      Transfer    Bandwidth
[  3]  0.0-10.0 sec  603 MBytes  505 Mbits/sec

Invalid option –4?

you missplaced the parameter… you should use "P" for parallel TCP-Sessions.
But in LAN environment it should be fine with one TCPstream. You can play with TSO and LRO (if your networkhardware and co can work with that) or play with the TCP Window Size.

[Derelict]

I had originally put -C4 and edited it to -P4 so it was probably cut/pasted just fine - it's me who sucks.

Swapping that out for something not realtek would be an interesting test considering recent threads. Seems you should be getting better than that. Though if you are going to performance test I would disable pfblockerng. Or at least compare with and without. (there shouldn't be much if any difference).

[heper]

i would stop trying to measure from/to the firewall. This is pointless & incorrect.

try this & report back:
iperf-server<->routerWAN | routerLAN <–> iperf-client