Cap AC at school seeing huge drop in performance when more than 25 clients

Disclaimer: I am not the most knowlegable at our company with the Mikrotik, but I am the one reaserching information right now.
PROBLEM:

  • It appears when we get 40-70 clients on one AP, the WiFi speeds drop dramatically compared to other roomms with less than 20 clients. Sometimes drop to 0.
    This has been hard to troubleshoot becuase we look at the AP remotely and the CCQ looks good and the sppeds in teh registration table looks good, but the actual experince for the user is terrible.


  • The same devices elswhere on teh campus, with the saem configuration, are performing great. Only difference seems to be the client load.


  • Clients are 85%-90% on 5Ghz, but a few are on 2.4Ghz.


  • Most often happens when the entire class is doning online testing - so everyone is activly getting content.


  • Once the client load drops, speeds are good again.

Question: Is there a “known” issue with capAC getting overloaded with 40+ clients all downloading? Does anyone think a capAX will perform any differntly/better?
Enviroment: Private K-12 school with multiple buildings. 20 cap AC, 6 capAX (new building), Mikrotik switches at IDF/MDF, 5009 core router.
Background: Originally installed capAX everywhere a year ago with CapsMAN. Ran into lots of problems, switched to capAC, still had problems, elminated CapsMAN and did direct configuration. Most problems ended. CapsMAN was the likely problem, but left capAC units in place anyway.
Over the last 7 months we get random compaints about bad internet speed on WiFi for entire classrooms (eveyone in class having an issue). When we investigate later we don’t find any issues.
Setup: I can send more details, but basically each AP has 5Ghz with main interface, plus 2 virtual interfaces, same on 2.4Ghz.

  • Traffic for each SSID is routed via vlans on bridge interface. No other “special” config. Everything built up from empty config.
    Channels are staggered. 20/40Mhz channel width on the 5Ghz and 20Mhz on the 2.4Ghz channels.
    rOS 7.18.2 / country set to United States / Skip DFS=all / protocol=802.11 / antenna gain=6
    We use Mikrotik APs and networking at 30 different clients, but the school is the only client with this level of loading on individual APs. We have experience configuring and installing Mikrotik. This issue is just more elusive for us.

I had similar problems with wireless drivers as well, but already with about 15-20 clients.

Would it be possible to switch to wifi-qcom-ac drivers?
There are some disadvantages ( https://help.mikrotik.com/docs/spaces/ROS/pages/224559120/WiFi#WiFi-Replacing’wireless’package ), but for me the advantages of many times better stability and throughput outweigh. Then also 802.11ac rates on 2.4 GHz.

The last time I visited the school during a “larger” event in the auditorium, I was very surprised… there is one cAP ac (so all clients had a direct view of the AP and 95% of them were on 5 GHz) and I’ve probably recorded my record of clients on MikroTik AP.

The throughput was… I¨ve got 30 Mbps with acceptable ping on that AP. :slight_smile:
Screenshot_com.ubnt.usurvey.jpg

Well known problem for me, so here is a few tips and questions to better understand whats going on:

  1. Tell more about deployment, physical AP deployment - are APs deployed in classrooms? To have a good modulations you need APs in every room where WiFi will be used - classrooms, administration, event halls, cafeteria, sports hall etc. Also what I call secondary priority locations (locations that can have lower quality service - less amount/more spaced APs, providing lower signal and capacity than first priority locations, but still acceptable performance) like hallways, resting spots, outside of buildings, even small, dark corridors if people gather there and stay for longer that 5seconds (hyperbole, but not far off). I hi density networks it is very important to use airtime as effectively as possible (to get the most throughput and responsiveness) and split load by deploying enough APs. Secondary priority locations might sound like a resource waste, but reality is, if a station is connected to SSID it will always try to connect to it. In school example it means every device that is in hallway (even if just passing by) will try to connect to AP in classroom, creating two problems: a) That station will have low modulations comparing to stations in classroom, that will waste airtime. b) its additional load to classroom AP. Not a fun to have problem like this while students are trying to pass online exam. Besides, its nice to have proper WiFi access all over school for everyone.

  2. TX power only 6dbm is too low. School where I used to work had 8x6m classrooms, to get and hold 256qam modulations under load I had to set TX power to 16dbm. Of course, raising power rases issues about channels - are they not overlapping? Have to admit, it was hard to set channels on 75 APs. Barely make it and secondary APs did shared channel’s (no other way). TX power and channel overlap is one of WiFis balancing acts - you have to make balanced decision or prioritize on off. To be clean, I do not use auto channel select on networks like this, its too unpredictable. Also DFS channels must be used. My goal was a clean, non-overlapping 40mhz wide 5ghz channel per first priority location whit modulations mostly (around 2/3) in 256qam.

  3. What are actual throughput requirements? One of important planning stage questions. ac lineup whit ww2 (vendor) driver can push around 140mbps under load in classroom. Here in Latvia we have up to 34 student classes + teacher and support assistant, so 36*2=72 as potential ilde device count per room.

  4. You need to use ww2 drivers in hi dense networks, legacy/in house drivers have too may issues. See my topic about it: http://forum.mikrotik.com/t/my-experience-and-issues-in-hi-density-networks-at-school/164018/1

  5. ax is a better generation that benefits mostly in hi dense networks. I have no idea how and what you configured and deployed before, RUs alone will help in classroom scenario. But I currently cant recommend ax lineup as there is issues whit wlan driver. MT should get new ax driver from Qualcomm soon as issue is confirmed.

  6. WiFi is not easy and environment in wild. So I do recommend proper planning, deployment and monitoring after. It does not help if you arrive at evening in empty school. To see actual performance you need to test it under heavy/realistic load - while students are present and using their Chromebooks or tablets. Remote monitoring is important, but there things you can only see on location in action. When deployment is done, second phase is traffic shaping (vlans,bcast/mcast limiting, queues etc) to further improve network’s performance and security.

Probably “hidden node” problem.

Wow, this is a well thought out and excellent answer from maigonis!

To answer some of your questions:
1.: originally there was old Unifi AC AP units in every 3rd classroom (drops are limited - we currently have new drops for every room scheduled to happen next month). we replaced the Unifi with Mikrotik cap AC, but had issues as the number of devices and use load grew. So we just did an install of just 5 cap AX units in a group of classrooms. We found some devices having issues with connecting to the when using the AX band and WPA3, so we have had to set them to AC and WPA2. They have some older devices. We are going the direction you recommend: AP in every class. Classes typically have 30-35 devices connected.

2.: We have manually set the APs to prevent overlapping frequencies, but we were using “skip DFS” because we are near and airport and a Navy base and I am not sure if it would be getting “suppression signals” from these locations. We also still have to support 2.4Ghz, so overlapping is going to happen there, but 85% connect on 5Ghz.

3.: Throughput requirements are not real high, 10Mbps/Mbps will work, it is just when we see it drop to 0.2Mbps in 3 classrooms, but other parts of the campus have 60/60 that we feel like idiots. We have traffic shaping queues on the core router and different vlans for staff, casting, students.

4.:When you say “You need to use ww2 drivers in hi dense networks,” are you referring to the drivers on the client or on the Mikrotik AP?. I skimmed through the thread you referenced, but will need to go back to really parse out the “useful info”

5.: Well the cap AC in place were doing terrible, so the cap AX is what we are trying…

6.: WiFi at the scale of a school where there are 500-700 devices, and suddenly 200+ need heavy bandwidth because of a test or video intensive lesson, is very hard to manage. We have deployed a lot of Mikrotik devices, but the unpredictable loads of the school are challenging. It seems our best “next steps” is to increase the AP density to and AP in every classroom.

Thank you for your time and effort.

AP in every classroom is the only option now a days. Even if you can somewhat manage passable performance today - tomorrow demand will only increase. Remember what I said about two important things in hi dense networks - use airtime as effectively as possible and split load.

As I sad, 2.4ghz will be challenging, I ended up with TX power 6dbm (typically) and radio turned on in every second AP in that school. Those results I get while monitoring network - capsman registry table and some field tests. Result is somewhat usable modulations to stations that do connect to this band (around 20-30 stations total, half is interactive boards android module that is not used). But overall, in middle of the day 2.4ghz band gets so crowded, that those medium modulations are the least problem. Bluetooth devices, student individual hot spots and God knows what else just destroys that band, it means 30-70% channel load whit no actual traffic passed from schools network.

We also are port, airport and navy presence city, but we have to use DFS channels. Otherwise there is only x4 20mhz wide channels to use (U-NII-1) - its barely better than 2.4ghz band. You just have to deal whit it, also do think about it while planning AP deployment and do configuration, so you have passable network performance even when DFS event happens. Note, ac lineup is more sensitive to DFS events, so you will see at least a few APs going DFS every day. (I did report this to MT support. At home I have ax lineup APs and did setup ac lineup (cap ac and hap ac2) as a test for comparison. ax lineup on DFS channels have been triggered 4 times in 2024 (both, total in that year), the same channels now on ac lineup - hap ac2 a few events per day, cap ac less, but still relatively hi number)

If you are not using please do use frequency usage and scan features in ROS, as its the only way to make sure channels are not overlapping. This allows you to see what AP “see”. Goal usually is a clean channel for every AP in 5ghz band.

Complete AP halt is house/legacy driver known issue, so do use ww2/vendor drivers (those are drivers on APs). Read that post I mentioned before and do check help page - https://help.mikrotik.com/docs/spaces/ROS/pages/224559120/WiFi . Setup will require time and you will need to migrate to new capsman - but its worth it. It will feel like getting new APs, networks stability and performance will be noticeable immediately (assuming configuration is good).

Looks like we have similar station amount and load in classrooms. It can be unpredictable, yes, but usually, at least for us, its web based content (like classkick), moderate video streaming from youtube (or vimeo), rarely file downloads. Usually those downloads aren’t big, but OS/app updates can put load on AP. One more reason to have AP per classroom if updates are not managed by staff after hours. Also note about load type. Typical web page traffic and downloads can handle slower network and responsiveness - real time data cant. If you want to have a good experience while Zooming, doing calls in Whatsapp, Teams etc, your channel load must be not much more than 30-40% in ac deployment (latency and jitter will noticeably increase). ax deployment should tolerate more, as RUs are used (if most of stations are ax clients).

Every part of network plays a role here. First you need good WiFi deployment (good modulations, split load), when that is done you need to do traffic shaping as I mentioned, to split better available bandwidth, prioritize traffic, eliminate unnecessary packet flow (like bcast/mcast, Apple devices for example are yapping like crazy to broadcast their service’s) etc. Its not easy and quick task, this is general advice, not specific to MT devices (mostly).


I do enjoy helping out from time to time, but as good answer on interesting topic can take so much of my time, I rarely participate in discussions.

I marked this solved because user maigonis put effort into sharing their solutions at a similar installation. maigonis appears to have gotten a configuration that works. We DID NOT.

  • We tried all of the suggestions, but the school could not sustain this ongoing “production testing”. WiFi worked under “lab testing” conditions, but failed in real world loads that could not be accurately recreated in the “Lab/office” environment.


  • We are deeply disappointed, as we have been using Mikrotik for 10 years - routers, switches, APs, P2P - but this school installation has helped us better understand why we see on the forums “…I use Mikrotik for everything except APs”.


  • The problems first started 9 months ago with a brand new roll out of capAX, we first thought the problems of “random disconnects” were related to AX, so we switched all 40 APs to capAC, but in the end diagnosed it as a capsMAN problem. Once we did a “local configuration” on all 40 capAC the random disconnects went away. We kept the “leftover” capAX for use at other locations. Then we started to see random AP overload issues where entire classrooms would have the bandwidth of all clients crawl to near 0. We tried putting 5 capAX APs back in the worst problem area thinking the new standard / new device may help. Then we did all the tweaking and changes we found recommended on the forums (becasue there was still problems). The problem kept coming back - how frustrating for us, the teachers, and the students - especially on heavily loaded “testing days” .


  • We couldn’t keep doing this to them

We reluctantly installed 5 Unifi U6 APs in the worst problem area. The next day, NO PROBLEMS. 5 days out now and NO PROBLEMS.

  • I don’t even like Unifi - the “dumbed down” GUI management, the lack of controls, lack of meaningful data, the push to get all Unifi devices - these are reasons we switched the school to Mikrotik the year after we took over IT management.


  • But, the Unifi APs clearly worked where Mikrotik APs were failing. And with very little configuration and time.


  • How is it that Unifi works so smoothly in this mildly loaded environment, with a central controller/manager, and Mikrotik fails? Mikrotik has so many amazing products and abilities, but it cannot handle 40 clients on one capAX (with local config, not capsMAN) when all the clients (students) need to take an online test?


  • First capsMAN disappointed us, now the actual device (capAX and capAC) have left us completely disillusioned. So much time (un-billable) was lost in the last 9 months on this one client because of undocumented Mikrotik issues.


  • We will still use Mikrotik for router/switches/P2P, but anywhere there may be a “load” on the WiFi, we will have to go with something else. This time it was Unifi, but we will keep an open mind.

-Keith

I can imagine that quite a few of us can relate to this post and the “I use Mikrotik for everything except APs” comment. I did one install in a very busy bar. Never again. Harmed my reputation and pride. Yes, the router and PoE switch has been rock solid but I spent days trying to get the Wi-Fi stable. So since I’ve gone back to UniFi which saddens me because I’d love Mikrotik to thrive as the software is stunning. But as many have commented, maybe too complex. Hence the reasons for a web-based controller system to simplify.

We may get some jumping down our throats with “Unifi is a pile of c**p” but I’m afraid real life experience shows otherwise. I’ve got a very complex and demanding wedding barn client where Wi-Fi is critical. No mobile signal so it has to work. They soon get complaints if Wi-Fi is down. Single UniFi 6 AP and often nearly 100 clients. Okay, less demanding throughput wise than an office but it just works day in, day out. 10 other UniFi APs on the site. Can’t remember the last ticket raised with them.

There is a very long running post on here about disconnects. We’re all hoping that it’s software and that one day it can be fixed. But sneaking suspicion that it’s hardware…