EPMP Throughput

Hi Guys,

I just wanted to do a sanity check here to see if I'm missing anything.

Setup:

  • ePTP link of EPMP1000-GPS (ePTP Master) with a standard EPMP1000 (ePTP Slave).  I've also done the same test with 2 standard EPMP100s
  • Capacity at 100%
  • Tx Power of +15dBm on each
  • Rx Power of -53dBm on each end
  • SNR: ~43dB on each end
  • Gear: Macbook Pro <-> Dell Power Connect 2816 <-> POE Injector <-> EPMP <-> EPMP <-> POE Injector <-> Dell Power Connect 2816 <-> Macbook Pro
  • Verified that there is no noise on (or close to) the channel.

Throughput testing:

  • Running 'iperf -s' on one side
  • Running 'iperf -c <ip> -d -i 1 -t 30' on the other side
  • Seeing aggregate of about 70mbps to 80mbps, depending on the run
  • Running with multipe streams ('iperf -c <ip> -d -P 5 -i 1 -t 30', for example) results in about the same.

According to this document, I should be getting up to 108mbps aggregate.  

  1. Am I missing anything?  
  2. Or should ~70-80 mbps aggregate be about right for TCP?
  3. Perhaps the 108mbps aggregate is more for a mix of UDP and TCP?

Any/all input is appreciated.

For clarification, I'm assuming you're using a 20MHz channel?

What firmware?

What mode - TDD, PTP, ePTP?

UP/Down ratio or Flexible?

What MCS up and down is it most often using?

Did you try iperf with '-u' UPD flag and/or '-r' to test up/down independantly?  Not saying that's a better test, just saying the comparison might be informative.

j

I didn't have 'E-mail me when someone replies' checked, so I didn't see that you had responded.  Thanks for the reply and sorry for my delayed response.

  1. Yes, 20MHz, thanks.
  2. Firmware: 2.4
  3. ePTP mode
  4. Up/down ratio of 50/50
  5. Attached are the MCS readings... I think this is pretty strong, right?

I did do with -u flag, but I did it at a limited bandwidth (~30mbps bi-directional) to see what non-max-load latency looked like.  (I'm interested in maximum TCP performance, so the UDP performance is not of interest for max load.)

Yes, I did do '-r' also, and was able to achieve ~58mbps (IIRC) in each direction like this.  The limiting factor comes when I try to do both simultaneously.  (The test machines are able to get 1,883 mbps aggregate across the switches being used, so it shouldn't be the test setup). 

I think that I may be reaching the limit of the radio... I just want to make sure I'm not missing anything.

Thanks again for your help.