The value of tuning retry parameters in real-world applications

[-=thrasher=-]

One thing that shits me is there is a real lack of tuning info out there for radio comms. I learnt mine through experience but not everyone can do it for a living, so here are some results.

This test illustrates the value of retry counter tuning depending on condition.
The bottom line is, if you are in a clean RF environment (for your channel), have good LOS, close range and good antennas, there are benefits in reducing the retry limits for *LONG* packets.
Longer, dirtier, weaker hauls can often benefit from increasing the retry limits for *SHORT* packets to reduce packet loss (and speed TCP throughput). Please observe the different results for different packet sizes (lrl favours downloads, srl favours VoIP and small, frequent packets.)

Using firmware 1.9.2.7-3c, and my tuning script.
The details of the setup are many and varied, so I won't go into it here. Suffice to say I think it's a good test you can trust Basically I have cranked up power, dropped attenuation and locked the speed at 2Mbit/s. Also locked external antennas.
Good conditions means AP's have dipoles connected, 500mm apart on vertical polarisation.
Poor conditions means no antenna, 250mm apart.
Please note that fragmentation threshold and RTS threshold are left default.
Those will be a test for another time.


I am adjusting the short retry limit, and the long retry limit.
Basically it means "how many times should I try to resend a packet before giving up?"

Set this using "wl srl xx" and "wl lrl xx" where xx is a number between 1 and 255 (really 1-16)
I am setting each endpoint (WL-500G) to the same.

Note: lrl/srl = 17 results in 100% loss, this is max parameter.
Defaults:
wl srl 7
wl lrl 4

**Testing criteria, result format**
Round Trip Time measurement, Packet loss: ICMP Ping, 32 bytes, max time 500ms
SPEED; mix/avg/max/loss
Measured over 60 seconds
As a bitrate test, iperf tcp is conducted one-way alongside the ICMP.

RESULTS
-------

I have a lot more raw data than this that took hours to collect but I was happiest with the testing used for this set of data.



c6.GOOD CONDITIONS wl s/l 5/1 - 1.42Mbit; 84/126/197/0%
c6.GOOD CONDITIONS wl s/l 5/5 - 1.41Mbit; 76/147/232/0%
c6.GOOD CONDITIONS wl s/l 5/10 - 1.32Mbit; 65/125/226/0%
c6.GOOD CONDITIONS wl s/l 5/15 - 1.08Mbit; 64/149/420/0%


c6.POOR CONDITIONS wl s/l 1/5 - 0.02Mbit; 2/2/4;38%
c6.POOR CONDITIONS wl s/l 1/15 - DNF
c6.POOR CONDITIONS wl s/l 5/1 - 0.96Mbit; 48/174/337/8%
c6.POOR CONDITIONS wl s/l 5/5 - 0.95Mbit; 63/208/364/5%
c6.POOR CONDITIONS wl s/l 5/10 - 0.96Mbit; 52/192/327/6%
c6.POOR CONDITIONS wl s/l 5/15 - 0.92Mbit; 76/186/381/15%
c6.POOR CONDITIONS wl s/l 7/5 - 0.93Mbit; 88/246/473/0%
c6.POOR CONDITIONS wl s/l 8/5 - 0.93Mbit; 57/154/365/0%
c6.POOR CONDITIONS wl s/l 9/5 - 0.85Mbit; 58/187/376/1%
c6.POOR CONDITIONS wl s/l 10/5 - 0.77Mbit; 20/175/576/1%
c6.POOR CONDITIONS wl s/l 12/5 - 0.79Mbit; 38/176/387/0%
c6.POOR CONDITIONS wl s/l 15/5 - 0.78Mbit; 82/196/481;0%


c6.EXTREME CONDITIONS wl s/l 5/5 - 0.08Mbit; 2/11/108/29%
c6.EXTREME CONDITIONS wl s/l 15/5 - 0.29Mbit; 2/286/777/47%

[tommurphy73]

Hi,

Have you done any further work on these or any other settings as I found this post to be very interesting and useful.

Regards
Tom