Will Access Points ever need two data points?

[kennysarmy]

In designing the data points in a new 4-classroom block I was going to spec. single data points for the AP's.

 

But given that this building may be around 10-20 years time, would you put in double data adjacent to each AP?

[mbedford]

IMO. Always run a double, even if not for WiFi, at some point you are going to be glad you ran that second line.

[Primus]

IMO. Always run a double, even if not for WiFi, at some point you are going to be glad you ran that second line.

 

Agreed, we never run singles.

[Scotbynature]

Yeah got to agree, if costs permit always go for more. Cover all bases as trying to run another line later will be considerably harder. The 2nd point doesn't even need to be patched in at the minute so all you are spending is extra copper cable and patch panels.

[Norphy]

It won't hurt. Possibly NBaseT might be more more widespread as time goes on, negating the need for an extra cable but considering it's only four classrooms, the extra cost of running those cables will be next to nothing and might well come in useful in future.

[Michael]

In designing the data points in a new 4-classroom block I was going to spec. single data points for the AP's.

 

But given that this building may be around 10-20 years time, would you put in double data adjacent to each AP?

 

Yes - one step closer before they're all natively fibre I suspect!

[Davit2005]

Biggest cost is nearly always going to be labour so budget for double needed at least where possible. We always do now, all refurbs and new builds are doubled up for APs Edited March 15, 2021 by Davit2005

[kennysarmy]

Cheers guys.

Cost is a factor and project already over-budget, however I think we'll go for doubles.

[mavhc]

What cat cable are you putting in?

[EddyAllen]

Yes - one step closer before they're all natively fibre I suspect!

 

Can't wait until just running fibre trunk cables everywhere is the done thing and we can do away with copper! Just need to get workstations to adopt LC connectors instead of RJ45 and we will be laughing!

 

Imagine rooms with these instead of RJ45's

LC 25x50 Angled Fibre Module, White - Tuk | CPC UK (farnell.com)

 

 

/dreamon

[Davit2005]

Can't wait until just running fibre trunk cables everywhere is the done thing and we can do away with copper! Just need to get workstations to adopt LC connectors instead of RJ45 and we will be laughing!

 

Imagine rooms with these instead of RJ45's

LC 25x50 Angled Fibre Module, White - Tuk | CPC UK (farnell.com)

 

 

/dreamon

 

Sceptical about the longevity of having a fibre patch cable to a desktop in a school environment unless we can get fibre cables up to the task.

 

You also need to factor in costs of SFPs and the switches that will need to be mostly SFP/SFP+. I think we are a bit away from that at the moment although switches with SFP ports are becoming more widespread and cheaper.

Edited March 15, 2021 by Davit2005

[EddyAllen]

Sceptical about the longevity of having a fibre patch cable to a desktop in a school environment unless we can get fibre cables up to the task.

 

You also need to factor in costs of SFPs and the switches that will need to be mostly SFP/SFP+. I think we are a bit away from that at the moment although switches with SFP ports are becoming more widespread and cheaper.

 

I thought so too until I found fs! They are about £10 for SFP+ modules and the patch leads are pretty cheap too. Agreed though, having seen what kids can do to headphone leads I can online imagine the rate of attrition on LC's!

[TronXP]

Not to mention kids pulling them out to shine "lasers" at each other.

[kennysarmy]

What cat cable are you putting in?

 

Cat6 is on the quote.

[TechMonkey]

Not suggesting it is a great idea but if you really need to save money you could ask them to run double cables but only patch and terminate one. That way you could later go round yourself or get someone to add them, just make sure they are well labelled. Running the cable won't cost much difference as they are already pulling one cable and cable costs pennies. I doubt there is much of a saving though.

[psydii]

Only the biggest and most powerful and *earliest* APs have ever needed two cables so far. And after a generation, you get the same capabilities through single port.

 

No wifi comes close to sustained 100Mb/s+ performance when under load from multiple devices (e.g. a class set of devices). No sign of any practical standard that will change that. We throw photos and videos around our ac wifi, and its just fine.

 

I think we have reached a point that a properly designed wif put in anytime since AC wave 2 is good enough. Just as gigabit lans are good enough. At the agrigation point (servers/internet) multi-gigabit may be useful, but at the endpoints unless you are trying to throw 4K+ around then 1 cat6 per ap fine. If you are expecting to be throwing 4k/8k around, the shared medium of wifi is never going to be the right choice.

[mavhc]

Once we have APs that work on all channels at the same time then it'll be better.

 

You should be better much higher speeds than 100Mbps though. Question is how often do your APs exceed 1Gbps in either direction?

[psydii]

You should be better much higher speeds than 100Mbps though.

 

Really? We had Aruba, Aerohive and Meraki come in and run a bake-off with 3 aps (3 from each vendor) across six classrooms with 180 devices and what we found was after you got to about 10 devices, under real world conditions (loading user roaming profiles) performance was consistantly sub 10Mb/s. Once past the peak, performance improved to the point where users did not notice a difference between wires and wireless devices, until 30 devices tried to upload 1Gb of photos each, at which point performance suffered again.

 

This was all ac wave 2 kit. The performance measured was broadly in line with the details published by Aruba and Cisco in their Very High Density design documents.

 

We engineered our solution to work around this by removing roaming profiles (they are depreciated and incompatible with OneDrive), and working with departments to ensure understood to spread the load of students uploading photos across a lesson, and to not wait until the last five minutes.

 

To be clear, we regularly see 190+Mb/s per AP (sampled over 2 minutes, so the peak will be much higher), but never anywhere near 1Gb/s unless it is a large transfer from one (or maybe two) devices, and there is very little else on that AP.

[psydii]

To return from my digression...

In designing the data points in a new 4-classroom block I was going to spec. single data points for the AP's.

 

But given that this building may be around 10-20 years time, would you put in double data adjacent to each AP?

 

If you are not putting one ap in each room (do what your RF design says, not what random people on forums advise)... but you are looking to the future (those cables are going to be ther ein 20 years)... make sure you have cabling in place to allow you to put an AP in ANY room, even if it isn't used in this year's itteration of your wireless network.

[jmak]

I can see the temptation of only having one of the 2 cables terminated (haven't seen anyone recommending only running one) if budget is really tight, but if it's not terminated, presumably you don't get a test certificate for it and wouldn't find out it's faulty until years down the line.

[mavhc]

Roaming profiles are 1000s of tiny files, so not surprising. Also depends on the drivers, firmware and hardware of the clients of course

[RobD]

Really? We had Aruba, Aerohive and Meraki come in and run a bake-off with 3 aps (3 from each vendor) across six classrooms with 180 devices and what we found was after you got to about 10 devices, under real world conditions (loading user roaming profiles) performance was consistantly sub 10Mb/s. Once past the peak, performance improved to the point where users did not notice a difference between wires and wireless devices, until 30 devices tried to upload 1Gb of photos each, at which point performance suffered again.

 

This was all ac wave 2 kit. The performance measured was broadly in line with the details published by Aruba and Cisco in their Very High Density design documents.

 

I'd be interested to know what channel width you were using?

 

If you are not putting one ap in each room (do what your RF design says, not what random people on forums advise)... but you are looking to the future (those cables are going to be ther ein 20 years)... make sure you have cabling in place to allow you to put an AP in ANY room, even if it isn't used in this year's itteration of your wireless network.

 

I agree with this (and this), wireless designs are so important in high density environments.

[psydii]

I'd be interested to know what channel width you were using?

 

80Mhz enabled, with no overlap between the wide channels between the test APs, and the base channels uncontended. We also made sure MU_MIMO and Beamforming was enabled. All devices were at -65dB or better, spaced as they would be in a 1:1 deployment.

 

Two of the vendors sent 3 digit CWNE's. They seemed happy enough with the process and results.

 

Our key finding (because the exponential fall-off in throughput was already known to us and is widely documented) was that MU_MIMO and Beamforming did make a small but appreciable difference in a busy environment such as ours. Those vendors that did better with that tech had better overall throughput.

 

Subsequent anecdotal evidence suggests that running at 20 or 40Mhz wide decreases the preformance of our wlan in a way our users notice.

 

The lack of high-density reference designs for 802.11ax suggests to me that there is minimal improvement in enterprise/public-space scale WLANs to be had.

[RobD]

80Mhz enabled, with no overlap between the wide channels between the test APs, and the base channels uncontended. We also made sure MU_MIMO and Beamforming was enabled. All devices were at -65dB or better, spaced as they would be in a 1:1 deployment.

 

Two of the vendors sent 3 digit CWNE's. They seemed happy enough with the process and results.

 

Our key finding (because the exponential fall-off in throughput was already known to us and is widely documented) was that MU_MIMO and Beamforming did make a small but appreciable difference in a busy environment such as ours. Those vendors that did better with that tech had better overall throughput.

 

Subsequent anecdotal evidence suggests that running at 20 or 40Mhz wide decreases the preformance of our wlan in a way our users notice.

 

The lack of high-density reference designs for 802.11ax suggests to me that there is minimal improvement in enterprise/public-space scale WLANs to be had.

 

No wonder you had poor performance with 80Mhz wide channels with 180 clients, unless you have a specific reason or a lot of direction antenna?

[mavhc]

I can't see anyone doing proper reviews of enterprise wifi.

 

Beamforming requires spacing between clients, not great for 30 pupils in a classroom.

 

OFDMA is supposed to mean that a slow client doesn't slow down the rest of the network

 

8x8 MU-MIMO should allow 4 2x2 devices to work at the same time, if they're spaced apart.

 

Also seems to like having the clients support 8x8 CSI feedback in the driver http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.309.5245&rep=rep1&type=pdf