iPerf Testing between laptops connected through cnWave


I have cnWave setup like below diagram using V5K and V1K:

When doing iPerf testing from laptop A to laptop B, I only managed to get 400++Mbps:

But when test from laptop B to laptop A, I can achieved 900++Mbps:

There is no issue for both direction when testing using built-in iperf inside the radio. Both can get 900++Mbps. Is there any reason for this? Or am I missing something on configuration part.

Please advise.

Hello @rahmat-roots ,

Can you please swap laptops. We would like to check if laptop B’s performance as receiver is the bottleneck.

Hi @Kiran_Avva,

Already swap the laptops. Result still the same. Also swap the poe injector just in case.
FW is 1.2.1

Hi Rahmat,

I suspect there is not enough bytes in flight to fully utilise the wireless capacity. Could you please repeat the test with 4 parallel streams (-P4) or try increasing the window size in iperf.


Not trying to stir anything up here but why do we care about iperf testing? The bottomline is we are being judged by customers on what they get from various bandwidth/speedtest servers. We can show great speeds using iperf, or bad speeds with iperf, but the reality is that does not matter, I get that it is a tool used for testing and I am not very versed on the ins and outs of this process, but I was hoping someone can explain to me why we would even bother testing with iperf? Thank you

Hi Jonathan,

I agree with you that the end user customer throughput is what matters most. Iperf still has it’s place for testing in a lab environment, a network without an Internet connection, a network with an Internet connection of less capacity than the cnWave mesh or isolating a single link or portion of the network for profiling.

cnWave also supports running iperf onboard the node itself. This can be useful for testing the capacity of a single individual link in a predictable way. In a complex mesh network, it may not be as obvious what path your traffic is taking to the customer.

Hope this helps,

1 Like