Monday, November 19, 2018

You CAN change MCS rates

I've heard before that you should NEVER play with MCS rates on 11n and 11ac as they usually lead to instability on the network, but usually no one backs them up with any evidence (guilty of that myself). So I've decided to give it that theory a try and write a short blog about it.

Recently I was working on an upgrade to a very old warehouse system that is built in a very 3D way. There a rows of 20m metal aisles stacked up about 20m high, full with products ranging from cereal to metal cans. There are shuttles picking those products on both sides of the aisles and they rely on a timely delivery of RS-485 data via WLAN. For this a MOXA W2150A model is used, which supports 1x1 11n (a fuller feature set can be found here).

In such environments a designer can very easily come across a multipath problem, which causes many retries due to the signal getting bounced around so much that the higher data rates become almost useless and in fact a hindrance to timely packet delivery. In order to alleviate that problem designers often disable the high data rates, so to save the client devices the agony of shifting down through the "fast" rates (ballpark above 24Mbps) to the "working" ones (ballpark up to 24Mbps).

But as mentioned above, MOXA does support 11n, and 11n has some really good features that help with efficiency and packet delivery. I'm mainly talking about LDPC, STBC, block ACKs, etc. Plus the higher speed of lower rates compared to 11a/g would be good. (For reference on those click here and here).

The idea was to imitate the current working 11a deployment, where rates 6, 9, 12 and 18Mbps are used, which means using MCS rates 0, 1 and 2.

On Aruba, you set these parameters under the HT profile

wlan ht-ssid-profile <PROFILE_NAME>
   supported-mcs-set 0-2
You verify the rates by capturing a beacon frame from an AP that is broadcasting the SSID

The 0000 0111 means that the lower most MCS rates (the ones we've set) are being advertised and will be used for all transmissions in both directions.

Finally check that the device is connecting and transmitting packets and data, with a trivial connection test, and most importantly if shuttles are still moving.

So I'd suggest the next time someone tells you you should never do (or always do) something, you'd be smart as to ask why. Then you can either believe them and NEVER/ALWAYS do something, or ask them to provide some evidence as to why that's so.