Community discussions

MikroTik App
 
maigonis
Member Candidate
Member Candidate
Topic Author
Posts: 181
Joined: Sat Jul 20, 2019 8:16 pm

Abstract throughput calculation formula in hi-density networks

Mon May 01, 2023 3:55 am

Interesting and useful formula how to calculate throughput per AP. Maybe someone will find it useful, base is Cisco documentation (don't quote me on this), compared results whit my finding in work network + some of my variables in it based on my knowledge and observations. I'm still testing this for myself, have to mention this, here is the example:

400/2x0.7/2x0.8=56mbps

400mbps represents maximum phy speed/modulation, that you can get from 2x2 special streams connection using 40mhz wide channel, ac standard. "/2x0.7" part represents wifi headroom in general (half of phy and additional 30% off), when network is used by many devices. "/2x0.8" part starts to calculate worst case scenario in this example, if you want to know the best throughput (very optimistic) number you can stop here. But be sure all devices have and do connect 2x2, also signal coverage is good and modulations are holding very well (holding 256-QAM even under load). In real life I doubt it will happen, you, most likely, face mixed spacial stream environments and modulation can drop to 64-QAM levels under load. Airtime does not care if station have 1x1 or 2x2 special streams, they both get the same amount of airtime, 1x1 just gets less data thru - half of the amount as 2x2. So "/2" represents worst case scenario, when stations are, or are using, 1x1 special streams only (stations can and do drop streams as environment gets worse, also AP and station compatibility issues and misconfiguration can cause this). Adjust this as necessary, you can monitor stations in wireless registration or capsman registration sections. If you see mostly 1x1 leave this as 2, but if you see more 2x2 connection replace it whit *0.7 for example, so device ratio is more 30% 1x1 and 70% 2x2 (I hope I do math correct). "x0.8" represents modulation drop. A lot of things can impact modulation, like not so great signal level, interference from other APs, stations and environment. If you see in registration section that modulation is good you can skip this, but realistically it will wave, even for a bit under load as mentioned before, so adjust as you observe (again, please use ROS provided tools, they are useful). In my work network (its a school whit no so great AP placement at the moment) I see drops to 16-QAM levels or even lower, as signal coverage is not that great at places and situations, but its only for a few devices and at moments, so I'm abstractly averaging it out (that number still feels optimistic). Take into account TX and RX modulations, both at bad levels will impact airtime as channel is used by both sides (WiFi basics 101, at least up to WiFi5 - only one device can use channel, other wait their turn. WiFi is a half duplex connection, so its RX of TX, not both at the same time).

Result number is quite low, but realistic. I often see very low numbers on interfaces and i didn't understand why. But when you start to dig into how WiFi works, observe and test you start to understand why. I usually calculate the worst case scenario, so I can plan and adjust traffic shaping accordingly. Remember that this is very abstract (wifi in general is) calculation, a lot of variables can change this number, also different AP models and vendors will give different results, make adjustments (we use cap acs and hap ac2s). Result should be something you can at least take into account, if it works better - that's great! In this case I know that in classroom every student should have around 2mbps. We have 34 students per classroom plus teacher, so 35 stations. 56/35=1.6mbps, not full 2mpbs, but remember that this is potentially worst case scenario overall, also not all stations will use internet at the same time every day. Usually I see throughput jumping from 50 to over 100mbps, so that means this is quite close calculation. As more spatial stream device ratio increase, I see more throughput per AP. (1x1 ~50mbps, more 2x2 ~100mbps, see the pattern here?) Also as modulation is holding good it will allow more data to pass.

If you have access to Ekahau try to simulate this. Make 8x6m classroom, put 35 stations whit 1x1 using 2mbps SLA, also 35 1x1 phones whit background sync. Background sync will add that additional ~-20% of airtime decrease I mentioned because of APs placement (etc) Also we have tablets (ipads) for use in classrooms, so 70 devices connected in total per classroom can be possible anyway. As result you will see airtime usage a bit over 70% - that is the limit when things are getting nasty. Also Ekahau does not know real life environment and its interference, specific stations used etc. So again - a lot of variables, so that's why its quite hard/impossible to precisely calculate this. Look at this as abstract calculation to get understanding of what to expect.

Its way past midnight,sun is almost rising and I'm almost sleeping. I hope this will not be just a crazy midnight tough that I have tough out loud. If you have some improvement in mind or I did (major) mistake somewhere let me know down bellow.

I'm going to sleep - good night.

PS: Also remember that, as more stations connect to AP the slower the network, in terms or latency, will become. So general recommendation is around 30 stations per AP (band).

Who is online

Users browsing this forum: arebelo, GoogleOther [Bot], neki and 34 guests