I thought i would share my real world experience with enabling RTS/CTS and tweaking hardware retries.
We have a mix of about 500 Mikrotik CPE and 50 or so base stations all running between 2.9.51 and 3.10 +
Some of our sites have 30 + customers coming off one radio, and we would see average latency (idle) of about 20/30ms. We offer plans that start at 3mbit/sec so the need for throughput is quite high.
I recently read on the forums that RTS/CTS would improve things quite drastically, but only if you install it on the AP + client - so, i set about installing it on both. Below are some smoke ping graphs from 3 seperate customer sites to give you an idea of the difference - its impressive!
Throughput is definately improved as well - site 2 can acheive 9.7mbit/sec download and 9mbit/sec upload on the same sector as 31 other customers - and this is in a 10mhz channel.
I have noticed that if you install 3.30 + Wireless beta on an AP with a mix of 3x and 2.9x clients you get an improvement, but problems quickly ensue - the full potential only appears to be reached when you have both clients & the AP upgraded to have 3.30 + Wireless beta.
Final note - my experience is that setting hw-retries to 8 appears to be the optimal value - atleast in not crazily noisey enviroments.
Edit: Forgot to add that nstream is enabled - framer policy = best fit, 3200
This update is totally worth the time and effort involved if you can.
Thanks for that - but i do want to know, what (in your educated opinion) has attributed to the increase in throughput and significant decrease in latency?
When you say “RTS on the AP should be set to the highest level possible aka not at all” - what do you mean?
My understanding was that RTS / CTS was ‘ready to send, clear to send’ - i.e, the client requests to send and the AP broadcasts ‘clear to send’ to all stations, indicating that other stations cannot send until the requester has completed sending.
Furthermore - 3.30 + Wireless beta is the only current version (other than 4.x) that has the RTS/CTS implimented - of course, correct me if i’m wrong.
Finally, our entire network (as per my post) is running nstream with polling - the smokeping graphs were after implimenting RTS/CTS AS WELL AS nstream - you can see the difference yourself
Wait a moment. You might be seriously confused here. Did you try the wireless beta package WITHOUT RTS/CTS?
Based on what I have seen, the wireless beta package improves nstreme and fixes the high latency problem when using nstreme.
In other words, RTS/CTS might have nothing to do with these improvments.
Yes - prior to enabling RTS/CTS i had tried the wireless beta package (i’d much prefer to not have to upgrade all the clients as well)
I was aware that the wireless beta package had significant improvements for nstream, and while some drop in latency (especially while the station was idle) was apparent, i dont think the improvements were as drastic as what i am seeing now.
Perhaps will try disabling RTS/CTS on the AP / Clients and post my results back here after a couple of days, for interests sake.
RTS/CTS has always been supported in the AP. Here is a quote from Normis:
"OK, it’s like this:
Previously RouterOS AP always replied with CTS to those who asked RTS, but RouterOS client never asked for RTS. Ie. the support was not in the client. Then we introduced NStreme polling to solve the hidden node problem.
Right now, in the Wireless Test package, we have also included RTS support for the RouterOS client, but you have to configure it: http://wiki.mikrotik.com/wiki/Wireless_ … S.2FCTS.29"
Honestly, it sounds like anyone’s guess why you have improvements. Everything you did is more or less backwards from the way it should be done so trying to figure out why it works better is darn near impossible.
JWCN, could you please indicate to me where it says that the wireless beta package needs to be rolled out to the clients, not AP for the improvements to be felt?
When i search the forums it appears people are saying they load it onto the AP and huge improvements are seen immediately. There are also comments about enabling RTS/CTS with Nstream showing favourable results (though there is no proof of this).
I’m sorry, it would just be helpful if there was very clear (and easy to find) documentation surrounding this stuff.
I will do some testing myself and work out what has cause such a big improvement, i still have plenty of other AP’s / Clients to upgrade, so i can do it the other way around.
