10 GbE Network Test

CHANG3D

Supreme [H]ardness
Joined
Jul 23, 2010
Messages
4,975
I am used to lan testing with QCheck and Passmark's Performance Test (Network Test). But recently I upgraded to a 10 GbE switch (Netgear M7100), and it broke QCheck; basically, QCheck gave me an infinity as the result of my throughput test. Passmark performed fine, but it's just a hassle to use (or at least it's not as straight forward as QCheck).

Is there another free tool for network lan speed tests? Or is there a really cheap tool for it?

Let's say I have less than 20 devices to test on.

I've already tried Totusoft's lan speed test and find that its free version is useless.

I'm having an itch write my own software now, but if it comes to that, I'd rather pay for a cheap software.

(Spent too much money on upgrading already - switches, servers, and computers)
 
Last edited:
I haven't used it on any 10G equipment, but iperf has always served me well for simple testing. You can use jperf if you need a gui.
 
I've never actually tried it on any of our (10Gb) equipment at work, but iperf is where I would start.
 
iperf sucks for 10g. You have to run a MOUNTAINOUS sized stack in order to even tempt 10gig into waking from it's slumber.


What you want is to use a protocol like FTP or something to test with NOT SMB period. SMB is single threaded crap coding and can never realize the full potential of 10G.

Here is using Filezilla copying a 16GB dummy test file from my desktop to my ZFS NAS piping through my Cisco 3750E 10g switch. I could probably do better numbers if I used my Cisco 4948-10G switch but it is powered down right now.
I also used a RAM disk on my PC in order to be able to even reach these speeds. My single Samsung 840 pro chokes at SataIII saturation and is in no way capable of transferring as fast as a ram disk. I could also probably get faster transfers if I just ran fiber from card to card in took my switch out of the equation.

ojr8.jpg
 
Last edited:
Code:
PS C:\Users\Dragon> iperf.exe c 10.241.100.2 -M 8900 -P 4
------------------------------------------------------------
Client connecting to 10.241.100.2, TCP port 5001
TCP window size: 64.0 KByte (default)
------------------------------------------------------------
[200] local 10.241.100.1 port 64503 connected with 10.241.100.2 port 5001
[192] local 10.241.100.1 port 64502 connected with 10.241.100.2 port 5001
[188] local 10.241.100.1 port 64501 connected with 10.241.100.2 port 5001
[176] local 10.241.100.1 port 64500 connected with 10.241.100.2 port 5001
[ ID] Interval Transfer Bandwidth
[200] 0.0-10.0 sec 2.88 GBytes 2.47 Gbits/sec
[188] 0.0-10.0 sec 2.93 GBytes 2.52 Gbits/sec
[192] 0.0-10.0 sec 2.96 GBytes 2.54 Gbits/sec
[176] 0.0-10.0 sec 2.77 GBytes 2.38 Gbits/sec
[SUM] 0.0-10.0 sec 11.5 GBytes 9.90 Gbits/sec
Seems to work fine to me, going past -P 4 doesn't really improve speeds any, and I can't completely take all traffic off that link, so iPerf gets pretty damn close to 10G without much of a sweat.
 
+1 for iperf
Its what I used and was getting around 9.8 - 9.9 Gbits
 
Code:
PS C:\Users\Dragon> iperf.exe c 10.241.100.2 -M 8900 -P 4
------------------------------------------------------------
Client connecting to 10.241.100.2, TCP port 5001
TCP window size: 64.0 KByte (default)
------------------------------------------------------------
[200] local 10.241.100.1 port 64503 connected with 10.241.100.2 port 5001
[192] local 10.241.100.1 port 64502 connected with 10.241.100.2 port 5001
[188] local 10.241.100.1 port 64501 connected with 10.241.100.2 port 5001
[176] local 10.241.100.1 port 64500 connected with 10.241.100.2 port 5001
[ ID] Interval Transfer Bandwidth
[200] 0.0-10.0 sec 2.88 GBytes 2.47 Gbits/sec
[188] 0.0-10.0 sec 2.93 GBytes 2.52 Gbits/sec
[192] 0.0-10.0 sec 2.96 GBytes 2.54 Gbits/sec
[176] 0.0-10.0 sec 2.77 GBytes 2.38 Gbits/sec
[SUM] 0.0-10.0 sec 11.5 GBytes 9.90 Gbits/sec
Seems to work fine to me, going past -P 4 doesn't really improve speeds any, and I can't completely take all traffic off that link, so iPerf gets pretty damn close to 10G without much of a sweat.


I loaded some jperf and got good results. So yeah I will take back what I said. I havent used iperf since back when it wasn't good for 10g. Now it is awesome. So yes ipef/jperf.
 
Last edited:
I also recommend iperf/jperf. I've tested 10Gb with it many times with no issues. The trick for high throughput links is to run multiple thread and also increase the TCP window size.
 
Back
Top