Wireless connectivity

[artfulmatt]

Hi all,

 

We have used Xirrus for years and recently our off site 6FC closed and the students were moved back with the lower school kids.

 

I wanted a survey completed as I weren't sure how it would perform with the extra load from the 6th formers and the BYOD.

 

I then found that every array was broadcasting at 100%, so each device was being screamed at and getting confused. The radios were also bonded.

 

BUT.....since the arrays have had the power dropped im getting complaints everywhere that WIFI is **** lol

 

I have quite a few rooms with DB reading -75

 

Im tempted to turn the arrays back to full power, but annoyed knowing its not setup correctly.

 

Anyone had similar issues?

[njc235]

Throwing more power at a poor wifi deployment is rarely the answer. I would recommend a proper site survey to see where your dead/poor coverage areas are and see if the APs can be redeployed more effectively. Any decent reseller will be able to help for relatively low outlay compared the cost of a new wifi and the whining of poorly supported customers.

[AlanD]

If you are serious about high density users and high data rates you can only achieve the high QAM rates with an AP in the same room as client...and around 3m away at most..and that is with full power. The problem arises is when you have a large block of classrooms and want to reuse those channels. This is almost impossible with 2.4ghz, but might be possible with 5ghz, especially if you stay away from the wide channels introduced in “ac” wireless. Sometimes you can leave a room without an AP if it’s surrouned by other APs .. and those APs are not going to heavily loaded..

 

I’m never that enthusiastic about WiFi surveys...which were pretty much essential when you had a thin sprinkling of 2.4ghz APs. ...but today are largely going to tell you that if you want high data rates you won’t get them through a wall with 5ghz.

 

And I’m particularly not keen on surveys done by those selling you their kit...who will make anything and promise everything to fit your budget while selling as much as possible.

 

I don’t see any problem with power at 100% ...unless it’s 2.4ghz and you are attempting to reuse those channels....in which case..it’s either get coverage ...and put up with massive cochannel interference and lower data rates...or sacrifice coverage to get better data rates. Remember one weak connection in some distant room at 5mb/s is taking All the air time to deliver that. And while airtime fairness may give a local class of laptops a look in, that distant connection..possibly the heads office ...is going to suffer enormously. So it depends what your priorities are...but getting both range and client density are mutusll6 exclusive in a classroom block.

[njc235]

If you are serious about high density users and high data rates you can only achieve the high QAM rates with an AP in the same room as client...and around 3m away at most..and that is with full power. The problem arises is when you have a large block of classrooms and want to reuse those channels. This is almost impossible with 2.4ghz, but might be possible with 5ghz, especially if you stay away from the wide channels introduced in “ac” wireless. Sometimes you can leave a room without an AP if it’s surrouned by other APs .. and those APs are not going to heavily loaded..

 

You seem to forget the single channel model.

I’m never that enthusiastic about WiFi surveys...which were pretty much essential when you had a thin sprinkling of 2.4ghz APs. ...but today are largely going to tell you that if you want high data rates you won’t get them through a wall with 5ghz.

And I’m particularly not keen on surveys done by those selling you their kit...who will make anything and promise everything to fit your budget while selling as much as possible.

 

 

Most real VALUE added resellers will consider your requirement, business case, budget, and environment before recommending a vendor to suit. This will often result in a lower AP count if done properly.

put up with massive cochannel interference and lower data rates...or sacrifice coverage to get better data rates.

 

The single channel should be considered.

 

 

getting both range and client density are mutually exclusive in a classroom block.

 

Not true if done correctly.

[AlanD]

OK. Once upon a time...in the distant past - when all we had was 3 channels of 2.4GHz (or may be only two channels if you used a wider 40MHz channel)...single channel architecture was ...fantastic actually - and really clever. And if any reader making a choice about wireless vendors doesn't understand it then its worth reading up on because its clever - and gets around some of the limitations imposed by co-channel interference and large numbers of clients in a single collision domain. Basically the APs are on the same channel - and at full power. And the chipsets virtualise their MAC addresses so that clients just see a single access point regardless of where they are - and if the management system decides to move clients to load balance - clients never know about it.

 

And when you had a single laptop card wheeled between classrooms it was great.

 

The trouble now is that all classrooms have the equivalent of laptop carts - either because there are mobile phones , or BYOD tablets, or indeed laptops and that gives you a bandwidth problem that can only be solved by breaking up that domain (collision domain that is...) by layering (to use single channel terminology) or different channels to use the conventional term. And when you do this - you start to lose many if not most of the advantages of the single channel solution. So you might decide you will run a single channel solution with a really high speed 160MHz or 80MHz channel width....but actually this will not work well in a really dense busy environment...and you get much better overall throughput using access points with narrower channels and smaller collision domains - particularly when some of the clients are older and can't take advantage of the wider channels so that significant bandwidth is wasted.

 

So yes - for a single user you would get a really high speed - and even for single class of laptops the single channel solution works well....but start expanding that on a bigger scale and it starts to fall apart...and you would start to need lowering the transmit power to reduce the size of the collision domains to try to reuse some of the bandwidth.

 

Add to that the fact that some of the chips sets don't cater for virtualised MAC addresses and the result is that single channel - although once a deal break - no longer offers schools what they are after. And you don't have the same set of problems with 5GHz because it doesn't go through walls that well...and there are lots more channels.

 

And rather than using new wide channel "ac" access points - you might discover that putting a second "n" access point into rooms on narrower 20MHz channels might actually get you better results for the whole class than using wide channels. Some access points have configurable 5GHz or 2,4GHz to help you achieve this.

 

Putting access points out side a classroom to serve multiple areas is unlikely to get you the best results..although I agree you can omit an access point from inner classrooms with some success if they can be covered by several other access points ...especially if that means you don't need to reuse a channel somewhere else and therefore run APs at higher power settings.

 

...and I think there are compromises...the solution which gives you the highest speed to a single client - is not generally the solution that will give you the highest overall bandwidth to a high density of clients in a building. And when you start to reuse channels - which can happen even with 40MHz channels especially as DFS channels often gives false detections...turning down the power will reduce co-channel interference and kepp your bandwidth up - but yes, at the expense of more remote clients getting a poor signal,

 

As always don't take my word for this. I always advise everyone to get some test kit on site - several access points.. spread over several rooms...and hammer it with 30 or more laptops, We tried Meru (now Fortinet), Meraki, Aerohive and Unify...and within the range of errors produced by repeating tests found little real difference in actual throughput...but that was pre Wave 2 "ac" so perhaps things are different now. As I say - test it for your self - don't take my word for it...and don't be taken in by smooth talking vendors who convince you they have your best interest at heart...because they are after you money...lots of it...sometimes with annual charges as well....and if you can't tell the difference when testing...I'm pretty certain you won't see much difference when classes use it....

[psydii]

But back to the OP's question As for what you need to do right now:

How many AP/Radios can your clients out at -75 see?

Can any other APs see the radio that the -75db clients are on?

 

If the answers are approximate to 1 or 2, and other APs can only seen the -75 client associated AP radio at -85 or worse you can probably get away with lifting the power a bit on that one radio without causing problems.

 

I'm not a xirrus user so I'm just assuming you can tweak the power on single radios rather than having to power up the whole array.

 

Indeed, I've always been slightly suspicious of Xirrus arrays in general and in schools in particular. The primary driver for performance is how far away is the busiest client that can be heard by the other clients/ap on the same channel. Clients that are far away have poor SNR and poor throughput and therefore consume a disproportionate amount of air time. Xirrus is about throwing the channel as far as it can with its microwave horns.

 

A slow talker on the same channel, even if it is on another AP that can be heard by a 2nd AP/Client will slow the 2nd AP/Client effective data rate. The SCA does some clever maths and tweaks to the frames to try counter this. MCA instead use independant radios on different frequencies, usually at lower power. Both do some sort of airtime fairness management. When considering SCA the question is: does the trade off from Signal Strength balance or beat the collision domain congestion when compared to MCA? I don't know.

 

However, I've recently done a bake-off, but only with MCA vendors' kit. Outcome: With a -65db minimum at the client, if you have a cell with 15+ (we tested up to 60) busy clients in a room there isn't much difference whether you have an ap in the room or sharing the one next door. When you are saturating the network with data transfers either configuration quickly bottoms out to the same rate (less than 10Mbit/sec per client). This was stress testing with real world data transfers.

 

MU-MIMO does seem to make a difference in scenarios where you have lots of 1x1 and 2x2 clients and a 4x4 AP that are all MU-MIMO capable. On a 3x3 MU-MIMO Radio I saw poorer consistency in data throughput, and it took slightly longer to complete the tests.

 

In practice most use cases will not be this demanding. 60 machines with just a few at a time doing a big transfer sharing an AP across two rooms (-65db min) gave perfectly acceptable real world (Photoshop) performance, when compared to similar workloads over a 100MBit/Sec wired ethernet network.

 

Obviously this setup does depend on what your walls are made out of to achieve -65db. Interestingly 2017 5GHz radios seem to be orders of magnitude better at working through the same walls than 2006 era radios.

 

To plan for a new network that you intend to depend upon (i.e. your continued employment depends on it working) you absolutely need a proper multi-AP on a many-Sticks survey to see what will work for you building and projected use case. 256-QAM is an expensive goal but even with it you still end up with sub 10Mbit/sec performance per client in heavy traffic scenarios (but yes, blisteringly fast if most devices aren't doing a transfer).

Edited December 1, 2017 by psydii