WARP Project Forums - Wireless Open-Access Research Platform

You are not logged in.

#1 2017-Nov-22 22:45:16

starlord
Member
Registered: 2017-Nov-07
Posts: 12

Throughput bottleneck

Seems like there's a bottleneck on WARP's throughput.
I'm using the 802.11 Ref Design v1.7.2 and I have setup one node as an AP and the other as a STA.

Following the example: throughtput_two_nodes.py

I check the throughput values for all sorts of data rates by setting the 'mcs'.
Its odd that the throughput never goes beyond ~30Mbps even when I setup to HTMF 64 QAM 5/6.

P.S. I am targeting a throughput around 50Mpbs.

Offline

 

#2 2017-Nov-24 13:51:42

murphpo
Administrator
From: Mango Communications
Registered: 2006-Jul-03
Posts: 5159

Re: Throughput bottleneck

A peak throughput around 30Mbps is the expected result - see our (old but still valid) throughput characterizations for details.

MAC and PHY overhead is the primary cause of the difference between the raw PHY rate (54Mbps for NONHT MCS7 or 65Mbps for HTMF MCS7) and the achieved throughput. Later versions of 802.11 have addresses this bottleneck with longer transmissions, thereby reducing the relative overhead of PHY preambles, MAC headers and MAC idle periods.

Offline

 

#3 2017-Nov-26 20:46:22

starlord
Member
Registered: 2017-Nov-07
Posts: 12

Re: Throughput bottleneck

Thanks for the response.

By the way, when I check the throughput between two WARP nodes using IPERF3 software, the throughput does not go beyond 8Mbps. Following is the experiment setup:

1 node set as AP (connected to a PC via eth A)
1 node se as STA (connected to a PC via eth A)

I am not sure why is this happening. Do you have an idea?

Can you recommend a thrid-party/open-source throughput tester software?

Offline

 

#4 2017-Nov-27 09:10:15

chunter
Administrator
From: Mango Communications
Registered: 2006-Aug-24
Posts: 1212

Re: Throughput bottleneck

Is the iperf experiment UDP or TCP? UDP will be the most similar traffic load to the LTGs used in the throughput characterization that murphpo linked to. Also, how big are the frames being created by iperf? Are they near 1500 byte maximum supported by the design? Is your computer doing something weird like IP fragmentation to create multiple smaller packets out of a larger one? You can answer these questions by using Wireshark to monitor the Ethernet interface on your computer.

Offline

 

Board footer