How many clients can a MR33 handle

route_map
Building a reputation

How many clients can a MR33 handle

Hi Guys

 

How many clients can a MR33 handle? the below snippet is from the link below

 

Note: Each access point has a technical limitation of 128 clients maximum per radio. Please note that this is only a theoretical maximum; in practice, interference caused by multiple clients communicating simultaneously will cause this limit to be far lower.

 

https://documentation.meraki.com/MR/WiFi_Basics_and_Best_Practices/Approximating_Maximum_Clients_per...

 

That being said i have 3 MR33's and 2 mx64  and i need to support +- 500 users for an event, its a once of event.

Other events will have +-1000 users? can meraki handle these numbers?

72 Replies 72
BrechtSchamp
Kind of a big deal

This topic will help you:

https://community.meraki.com/t5/Wireless-LAN/AP-with-support-to-200-clients/td-p/3325

 

In short:

-The MR33 is not meant for high density. It would be better to move to at least MR42 but ideally even higher MR52 or MR53 (combined with multigigabit switching)

-While your mileage may vary a good rule of thumb is to aim for no more than 30 users per AP, maybe a bit more for high density (around 50) but not too much. This also depends on the usage per user of course.

-Have a look at the high density docs:

https://documentation.meraki.com/Architectures_and_Best_Practices/Cisco_Meraki_Best_Practice_Design/...

route_map
Building a reputation

for arguments sake, if i have 500 clients connecting and i had 10 APs this should be fine.
but i have 1 switch and 1 firewall (MX64), would these be a problem?
Kamome
Building a reputation

MX64's firewall throughput is 250Mbps. If all 500 users makes traffic more than 200Mbps(80% of maximum capacity of MX64), you can get some slow down. Plus, recommended client number of MX64 isa 50. So 500 users can be quite heavy for MX64 alone.

BrechtSchamp
Kind of a big deal

That depends on the usage by your users but yes, you need to consider the whole path to the internet for bottlenecks.

 

Have a look at the throughput numbers in the MX datasheet. And size your WAN link...

 

Also give a little thought to redundancy. If you're depending on one switch and one MX and one of them guess down you've got a problem on your hands... So apart from the bottleneck of a single uplink on your switch for redundancy alone you might want to deploy multiple switches.

PhilipDAth
Kind of a big deal
Kind of a big deal

>That being said i have 3 MR33's and 2 mx64  and i need to support +- 500 users for an event, its a once of event.

Other events will have +-1000 users? can meraki handle these numbers?

 

The MX64 can support around 50 users.  You need an MX250, which can support up to 3,000 users.  This is the MX sizing guide.

https://meraki.cisco.com/lib/pdf/meraki_whitepaper_mx_sizing_guide.pdf

 

The MR33 is the wrong AP to choose.  You want an MR52/MR53/MR55.  The MR42/MR53 are 4x4 access points (an MR55 is 8x8).  You will want to limit the channel width to 20Mhz.  At 20Mhz the MR52/MR53 can use all four radios servicing clients concurrently - while an MR33 can only use two radios - so an MR52/MR53/MR55 can handle twice as many clients straight away.

Also if this is one big open plan space you'll need to use external focused antennas.  If you mount to the roof, and depending on the roof height you could consider using a downtilt antenna.

https://meraki.cisco.com/products/wireless/antennas-power#

 

I would be using at least 20 of the MR5x series access points.

route_map
Building a reputation

understood, i already have the MR33s and MX64

I also have 4 MX74 and MA-ANT-20 antenna's

 

quick summary of the event i am going to cover. it is an open park

 

length = 101 metres

width =  58 metres

 

would the the MX74s cover 500 users

NolanHerring
Kind of a big deal

@PhilipDAth I personally think the MX64 should/could work fine for the client count of 500 (not sure I would go past this though). Unless they are ALL actively pushing traffic 100% at the same time, I don't think it will have a noticeable impact. I would prefer an MX67 of course, but I use the sizing guide as just that, a guide. I have an MX64 with about 500 clients a day (internet pipe only being 150Mbps at this specific site though) for guest, and it chugs along just fine. Again though the circuit isn't maxed out all the time, spikes here and there etc. I am planning on upgrading it to an MX84 in the near future once I pull it out of my DC which is being upgraded to an MX250.  Overall the type of event would help both of us decide what would be best because if its a party and only a few people are using their phones at the same time, or if its some sort of training and everyone is on laptops at the same time, could change this.

 

At 20Mhz the MR52/MR53 can use all four radios servicing clients concurrently - while an MR33 can only use two radios - so an MR52/MR53/MR55 can handle twice as many clients straight away.

 

^However I'm confused by this line here. When you say radio's do you mean antenna? And when you say concurrently are you specifying MU-MIMO? Off the cuff I don't agree with this, unless I'm misunderstanding what you mean.

 

 

@route_map 

Ironically there was a recent blog post (yesterday) about this from an associate that should help with exactly this problem.

 

https://wifiburns.wordpress.com/2019/03/28/how-many-clients-can-we-have-per-radio/

 

RF being a finite resource and all, client capabilities, all come into account.  Are all the clients transferring/streaming at the same time?

 

For a high density situation like this, if you CAN use 40MHz, then that will increase efficiency and throughput overall. If you can't because now you have access points stepping on each other, then default to 20MHz which is usually the safest scenario. Key here is you don't want any access points using the same channels (assuming this is in some sort of isolation and clean RF environment). This also applies for 2.4GHz, and you've only got 3 channels there to play with so if you have 10 access points in your final design, you'll be needing to turn off some of the 2.4GHz radios. 

 

In the space you mentioned its probably not going to be possible to use a 4th 2.4 radio without it having to contend with one of the other 3 you have enabled. So for this, I would separate the SSID you are going to use into a 2.4 only and 5-only so that clients have a smooth roaming process.  I would treat this as you want full coverage and some contention, versus no contention and coverage issues.

 

 

Technically the MR52/53 is a more capable access point as it probably has more powerful hardware, and they have 4x4:4 stream capabilities, however you have to keep in mind the clients being super important (hint - there are no 4SS clients).  Download MU-MIMO will only work with clients that support 802.11ac Wave-2, and that whole idea basically became a joke in the real world. So at the end of the day, the AP will still be sending to one client at a time, just like an MR33.

 

If you compare a MR33 (2x2:2) vs MR52 (4x4:4) for example. If the client is a smart phone, it most likely only a 1SS or 2SS. I know the Macbook Pro has a 3x3:3 radio so you could see more throughput for that specific client. However if you have a client that is 1SS or 2SS, you 'technically' will still see some gains, just not via the spatial streams. Those would come in the benefit of having more antennas on the MR52 providing better diversity, STBC, better signal/SNR etc. So in a lab test setup you would see some gains, maybe 10 or 20% 'better' (or so I've read).

 

This blog here also explains everything clearly as well: http://divdyn.com/wi-fi-throughput/

 

In this space, what is more important in my opinion is going to be the actual RF design.  If you just through 10 access points (any model) into a space like a basketball stadium because 'math wise' it works out, it won't necessarily guarantee that the clients will evenly distribute themselves between all 10 access points.  You could have one AP with 80 clients, and another with 10.  You need to use directional antenna, static channel/power, probably RX-SOP etc. 

 

Its a whole process that needs to be done correctly to ensure everything runs as optimal as it can, which is why people like me have a job 😃

 

At the end of the day, you can do a perfectly suitable design with MR33 and everything will work. If you use MR52, sure you get some gains, at a higher cost, and everything will also work.  Since it looks like you already have the MR33, you need to figure out what clients you will have, and figure out the math from the links above to determine how many AP's you need.

 


@route_map wrote:

understood, i already have the MR33s and MX64

I also have 4 MX74 and MA-ANT-20 antenna's

 

quick summary of the event i am going to cover. it is an open park

 

length = 101 metres

width =  58 metres

 

would the the MX74s cover 500 users


So that is about 350 feet by 200 feet or so, so assuming your mounting these antenna on the ceiling (again too many unknowns for me to give perfect advice), and you have your TX power set correctly etc., coverage shouldn't be an issue at all. Actually coverage would be achievable with that setup with just ONE if you had max power lol (don't do this though).  However, with only four access points you will probably be running into capacity issues at this point, and probably maxing out the hardware limit of client counts on the AP potentially if for whatever reason you have 'too many' clients getting stuck on one of the APs.

 

Off the cuff, from a mixture of clients of like (150 1SS, 150 2SS, 150 3SS) your probably going to need like 10 access points. This is just off my brain cells, use the calculator above with accurate data input to get realistic values. I'll reiterate though that this 60k or 70k square foot space, I think you'll have a problem wrangling the RF design with only omnidirectional antennas that your going to use. I think RX-SOP might be the saving grace here but do not use this if you don't know what it is. 

 

 

 

Nolan Herring | nolanwifi.com
TwitterLinkedIn
NolanHerring
Kind of a big deal


@route_map wrote:

it is an open park

I literally somehow missed this tidbit of information lol. 

How are you mounting any of this gear? This could change everything drastically lol.

 

Nolan Herring | nolanwifi.com
TwitterLinkedIn
PhilipDAth
Kind of a big deal
Kind of a big deal

>I also have 4 MX74 and MA-ANT-20 antenna's

 

@route_map  you'll find these work fine when there is no one at your venue.  However as soon as you have 500+ people connecting and transferring data it will fail.

You need to be using directional antennas.  I guess Omni's might work if you can situate them within the actual crowd itself and cable them back to a switch somewhere (don't try and use mesh).  You would want to put them up on poles so they are above the people (human bodies will soak up the signal otherwise).  Maximum height 3m.

 

I also disagree with @NolanHerring .  I think the MX64 will not handle having 500 concurrent users.  It will also fail to work properly.

 

 

On the whole - I think you are going to crash and burn.

NolanHerring
Kind of a big deal

@PhilipDAth  Are you referencing experience or simply data sheet limits?  You might be right, maybe 500 is the breaking point I honestly don't know.  The MX64 I'm using hits about 300 clients a day, and I have not seen any signs of buckling or anything though.

 

Below being an average day. ISP pipe is 150Mbps down/20Mbps up (boy I wish it was symmetrical).

mx.jpg

 

I agree that using 4 access point isn't enough though for sure. And actually, now that I know its an open field crowd, using the people bodies can come to your benefit. Its how we design inside stadiums, we put AP's under seats and the human body attenuates around 3db so it helps to regain channel re-use.  Of course I would now be worried about someone knocking them over, messing with them etc. lol.

 

I'd be more concerned about the wireless being overloaded vs the MX though in this situation based on mine running 300 clients on average pretty smoothly (so far).

 

I'm curious to hear how it all works out though @route_map  , so if you could provide an update after whatever solution you implement, that would be appreciated 😃

Nolan Herring | nolanwifi.com
TwitterLinkedIn
PhilipDAth
Kind of a big deal
Kind of a big deal

>You might be right, maybe 500 is the breaking point I honestly don't know

 

The difference is your 500 clients are spread out.  This is an open air venue where it is likely to have 500 clients all turn up at the same time for a venue, and all leave at the same time - so it is 500 concurrent users.

DonAnnett
Getting noticed

It's only 500 concurrent if all 500 are staring at their phones at the same time, the entire time. In our experience we grossly overestimate usage requirements based on 'people count'. When we go back and look at real usage during events it's always shockingly less than predicted.

 

Erlang anyone? ....

NolanHerring
Kind of a big deal

Good point Phil, that makes sense. But let me take it one step further down this rabbit hole haha

'concurrent'

Do you mean simply clients grabbing an IP off the scope and not really passing much traffic (in peoples pockets) or do you mean they are all trying to pass at the same time?

Not sure what the venue is, maybe its a music festival or some concert etc. I would be surprised though if there is more than 50% of anyone using/passing traffic at the same time. Again I don't know, only the original poster could tell us. 😃
Nolan Herring | nolanwifi.com
TwitterLinkedIn
PhilipDAth
Kind of a big deal
Kind of a big deal

>Do you mean simply clients grabbing an IP off the scope and not really passing much traffic (in peoples pockets) or do you mean they are all trying to pass at the same time?

 

I don't tend to see "quiet" devices at events.  Their using Instragram (under 25's), Facebook, DNS queries, messaging notifications, chat, etc.  Even old school email.  Even when the phone is in the pocket it will tend to generate some packets.

 

And the hardest of packets to process are short packets.

PhilipDAth
Kind of a big deal
Kind of a big deal

>we put AP's under seats and the human body attenuates around 3db so it helps to regain channel re-use

 

I do like this idea.  I tend to use focused antennas, but this is a smart idea to deliberately kill and limit the range of the signal.

NolanHerring
Kind of a big deal


@PhilipDAth wrote:

>we put AP's under seats and the human body attenuates around 3db so it helps to regain channel re-use

 

I do like this idea.  I tend to use focused antennas, but this is a smart idea to deliberately kill and limit the range of the signal.


Its a win-win because now the AP's don't hear each other that well (or the clients) via FSPL. And the clients will basically be 'forced' to connect to the nearest AP they can actually hear, increasing MCS rates, lowering retries etc.  Never done this in an open field so I'm curious to see how he implements it.  😃

Nolan Herring | nolanwifi.com
TwitterLinkedIn
BrechtSchamp
Kind of a big deal

@NolanHerring and @PhilipDAth if you haven't seen it yet I thought this presentation from Cisco Live US 18 about those topics was very interesting:

High Density WiFi for Large Public Venues and Stadiums - BRKEWN-2013

Matt Swartz, Principal Engineer, Cisco

Joshua Suhr, Solutions Architect, Cisco

 

You can watch it on demand on ciscolive.com.

route_map
Building a reputation

You guys are legends, so much valued information
Now i am kinda scared to do this event as i do not want to crash and burn,
This is my first event and want to showcase Meraki and Purple wifi for splash page and analytics.
please ask the questions and i will answer as i am available right now. you guys are very help full.
I want to do more events with larger crowds.
quick question 4x MX74 wont do the job?

 

egg_found.png

Egg found!!

NolanHerring
Kind of a big deal

4x MR74 will give you wireless coverage, but not capacity support. 500 'potential' clients is too many for only 4 access points for something like this.

Also, if you want the event to be as smooth as possible, and end-user experience to be friction-less, I would do yourself a favor and remove any splash page. Just have an SSID that is wide-open so people connect, and boom that is it.

Splash pages tend to have issues, users have problems with them, they get frustrated, they'll think the wireless sucks even though its technically something else like a DNS issue or splash page slow to respond or something else that isn't RF related.

Meraki gives you analytics built in, just make sure you enable Detailed Analytics under the General settings. Not sure what Purple Wi-Fi will give you, other than a splash-page of headaches. 😃
Nolan Herring | nolanwifi.com
TwitterLinkedIn
route_map
Building a reputation

Hi Team

 

Thank you for the great advise, some of it went straight past me.

I ended up designating 2 hotspot areas. Client had 2 activation areas for their customers to come and signup and download their app.

One last point or question. I hope the below illustration makes sense

 

internet -->Huawei B618(LTE-A Router)--->>MX64---MS120-24P----MR42 x62

 

Not to many people connected at the same time so i didnt have too many issues, but the next event there will be many more users and more activation areas.

 

Speeds here for LTE-A range from 10-50Mb. just that the LTE router i am using has a capacity of 64 clients so is this another bottle neck i might have? I am thinking of clever and efficient ways to set this up, any ideas welcome

NolanHerring
Kind of a big deal

The LTE router most likely won't be an issue with max client of 64 since you have it connecting to the MX64, which will be NATing, so it should only see it as 'one' client.

However, if your backhaul is going to be 10-50Mbps, it simply won't matter how well you design this system. That backhaul is going to get saturated almost instantly.
Nolan Herring | nolanwifi.com
TwitterLinkedIn
route_map
Building a reputation

thanks Nolan

 

If you dont mind, what would you suggest

 

I just read your blog, very interesting

https://nolanwifi.com/2018/10/25/you-down-with-l-t-e-yeah-you-know-me-raki/#more-654

 

MMoss
Building a reputation

If I'm following this correctly the client had a single hot spot and you deployed a pair of MR33's for the users to connect, register, and download the app? If it's a similar setup, but on a larger scale then I would consider splitting the network up into multiple networks running of individual hot spots as their internet pipe. This way you would be able to split your traffic up so it's not destroying your bandwidth. Like Nolan said, your true bottleneck is when the users start blowing up the connection. At least this way you can minimize the strain somewhat.

Other options would be to set bandwidth limits, content filtering out items like streaming, or outright blocking all traffic with the exception of what you need to allow for registration and app downloads. This can be noted on the splash page that traffic is limited to these items so people down bombard you with your wifi is down comments.

I think the MX 64 is fine, and the single switch can even be used to split the traffic to each MX with VLANS. depending on how many concurrent users you expect to have per AP the MR 33 could work, especially if placed properly and the traffic limited to the clients needs. The more access you allow though will force you into getting larger units to handle more concurrent clients and traffic. It's a balancing act at that point, limit the Wi-Fi so people get on, register, download and disconnect or allow access to the open internet and in which case that LTE router will likely make it moot anyhow.

If you have relatively good cell service in the area you can factor that in and assume some if not most will use cellular data over the free Wi-Fi, I don't recall what kind of even this was so sorry is this is a moot point out of the gate. When I travel I personally use guest Wi-Fi only when I need to use my laptop and cellular if I'm going to be there for awhile like a hotel.

Just something to consider, I've been doing a lot of pre-planning using Ekahau the last few months, I can tell you what the specs are of my network, and I can get a general idea of whats in use on a daily basis for the existing Meraki network. You may have 500 people but concurrent users will be far fewer, and being connected and using data are quite different as well. 20 users connected and maybe 3 or 4 are actually going something to register use. The rest are exchanging information only when background processes are running or passing packets back and forth to keep it associated to the AP.

Hope this is useful, sounds like you have a lot of guess work and eyeballing with an event like this.
route_map
Building a reputation

Thanks for the feedback. It is alot of gues work as you say.
And i have decided to have hot spot areas which will its own internet.
since i am using 1 Dashboard the SSID will remain the same, but the AP will cover its area only for the users to be able to download the Customers app
MMoss
Building a reputation

Stay flexible, the tried and true of building up a long term deployment is not going to work in this case at all, but make sure you use the data collected here so you can apply it to future events be they good or bad. Itll help with the guess work if you can get some sort of baseline that will likely improve the more data you can collect.

 

This is a outside of the box scenario and you may find going forward that running multiple smaller networks with multiple LTE internet connections is the way to go, especially in an area where you man need coverage in multiple area's but trying to cable off of 1 switch, 1 firewall, and a single internet connection is more troublesome and time consuming than it's worth.

 

Good luck

BlakeRichardson
Kind of a big deal
Kind of a big deal

@route_map  Maybe consider an MX67C for your next event, it has built in LTE.

MMoss
Building a reputation


@BlakeRichardson wrote:

@route_map  Maybe consider an MX67C for your next event, it has built in LTE.


We have several of these and I'm very happy with them. I have 20 facilities and have asked that we put one in each of the ones without a 2nd WAN connection. We get 50Mbps on Verizon, but only 5Mbps on AT&T, but that's the service avaliable on AT&T in this state unfortunately. 

VincentPanico
Getting noticed

“EGG FOUND!”
DerekPeterson
Conversationalist

Dangit. 22 seconds too late.

 

EGG FOUND!

MistahReeves
Conversationalist

Damn, you beat me by 35 seconds!

Mike333
Conversationalist

EGG FOUND!
AlexSilva
New here

"EGG FOUND".

RKacz1
Comes here often

EGG FOUND!

RKacz1
Comes here often

Darn password reset...

KimStromberg
Here to help

EGG FOUND!

SactoTam
Here to help

EGG FOUND!

Ricardo
Comes here often

Egg Found

raschank1
Conversationalist

EGG FOUND!

Zman712
Just browsing

Found the egg!

MacConcierge
Conversationalist

Egg Found
KaylaEik
Here to help

I found the egg!

 

EGG FOUND!

George_Silva
Conversationalist

EGG FOUND.

DanMirsky
Conversationalist

Egg Found

Josh9400
Just browsing

Found It.

jhaberman
Conversationalist

EGG FOUND!
Markz
Comes here often

Egg Found!

fcoelho
Conversationalist

egg found
cbradley489
Comes here often

egg found

Lee_C
Comes here often

Egg Found!!

dalmiroy2k
Getting noticed

“EGG FOUND!”

Dudleydogg
A model citizen

I see a MREGG33

Eglet  EGG FOUND

DBruns
Conversationalist

EGG FOUND

panyez
Conversationalist

EGG FOUND!

 

SCOTT_FORREST
Conversationalist

EGG FOUND

 

MMoss
Building a reputation

Egg Found
NolanHerring
Kind of a big deal

Unless I read the rules wrong, only the first person to reply gets the win guys.

No more spam please haha 😃
Nolan Herring | nolanwifi.com
TwitterLinkedIn
Connectech
Conversationalist

Egg found but a bit slow. 4 people beat me to it. 

Fun contest.

mmsamuel
Comes here often

egg found

 

ChrisTorres
Conversationalist

EGG FOUND

Ashieanna
Just browsing

Egg found!!

David_law
Conversationalist

 “EGG FOUND!”

Marty
Conversationalist

EGG FOUND!

MarcP
Kind of a big deal

All answering to get to the next tier 😄 😉

TravisFleming
New here

Egg Found!

luckado
Comes here often

Found the egg!  Late apparently 🙂

CharlesIsWorkin
Building a reputation


@luckado wrote:

Found the egg!  Late apparently 🙂


me 2. Egg Found

KRobert
Head in the Cloud

EGG FOUND!
CMNO, CCNA R+S
Cesar_Hernández
Just browsing