Wireless Access Speed

[smacrory1980]

Hi

We have a teacher that requires the laptops in his classroom for a particular lesson. The lesson involves downloading a large flash file. It took about 30 mins for all the laptops to fully download the file ready to use through the wireless network. I've advised him to use the IT suite and wired desktops for this lesson but for various reasons he won't.

 

The WAP is in his classroom with a 1GBPS connection back to the switch. Any ideas how I could get things running faster for him?

 

Ta

[snagrat]

All will depend on the wireless access point and the wireless chipset in the laptop.

 

Can you confirm these please

[LeMarchand]

Where is the file, what is the speed on the WAP and laptops, how many laptops and what's the size of the file?

[Blue_Cookeh]

All will depend on the wireless access point and the wireless chipset in the laptop.

 

Can you confirm these please

 

This. Remember WiFi is a shared medium and unless you have some of the newest generation access points (with MIMO, beam forming etc) and client cards your devices are going to be fighting for air time on both receiving and transmission. This only gets worse with more clients.

 

If you're going to be doing this sort of thing in future I'd look at deploying these files ahead of time in the background using SCCM or whatever your $configmanager of choice is. I dare say if you looked at the graphs for your AP's uplink, it isn't even hitting 10% utilization during these times because the bottleneck will be the wireless portion.

[AlanD]

The WAP is in his classroom with a 1GBPS connection back to the switch. Any ideas how I could get things running faster for him?

 

Ta

The ignorance and expectations surrounding wireless never ceases to amaze me. Wireless is not anything like the kind of connection you get with wired. Assuming there is an access point in that room so that you are using 5Ghz (and assuming the clients support that) and you have an at least "n" capability (both client and access point) then you might get 100Mb/s of real world data rates - if you are lucky. And that 100Mb/s isn't - as in the cabled world 100Mb/s to every client - but its 100Mb/s shared between all the clients. Even if you have upgraded to the latest "ac" wireless standard you might not see any increase in data rates - because neighbouring access points in a classroom block are probably going to limit you to using 40Mz channels which the older generation of access points mange just fine. And while extra streams and higher QAM rates might give some notional connection rate of 800Mb/s on the access point or client they are never going to deliver anything like that - 20% is typical. And if you are using between 20 to 30 laptops you can lose twice that 20% gain in network collisions.

 

The 1 gigabit connection from Access point to switch and from switch to backbone is more than enough to saturate what your access point can deliver to clients. And that is assuming that your clients support what your access point is capable of delivering. I notice that many of the latest laptops with "ac" are single stream - and often single stream that don't support the widest 160MHz channels. These are often worse than 2x2 "n" clients in dense environments because you can't use the wider channels without co channel interference from neighbouring classrooms.

 

So no - you are not going to get large video/flash/multimedia files into every laptop in real time. Suitable USB sticks are only around £1 and can easily be pre-loaded with the data required.

[psydii]

With a good design and typical clients, you'd be lucky if you got 2.25mbit/sec at each client if they were all pulling a 1GByte file simultaneously.

 

A real-world throughput of greater than 100mbit/sec for all clients on a channel when all clients are busy trying to transfer is very hard to achieve.

 

This video explains why in some detail:

http://www.revolutionwifi.net/videos/

(Gigabit Wifi: Not so fast)

Edited May 21, 2017 by psydii

[AlanD]

I have received several private messages saying that my positing was too negative about real world data rates. It has been pointed out that the previous link to the video claims speeds in excess or twice what I was suggesting might be possible. Please note that those speeds were obtained using iPerf using a UDP (or one way) transmission. I am more concerned with file transfer - for logging on, loading data, etc and these required TCPIP which have "ack"nowledgement packets. And it takes almost as long to send an acknowledgement as it does to send a data packet with wireless (..and yes - I know these packets can be aggregated which helps of course). So yes - you need to divide these speeds by 2 to compare with speeds on a wired network. Another pointed out that MU-MIMO should give gains of up to 1.8 times in speed in ideal circumstances. And I suppose this goes to the heart of my posting - which is that quoting higher and higher "connection" speeds or "speeds of "up to ...." and "in ideal circumstances" need to be taken with a pinch of salt. I can't claim that my testing was exhaustive or scientific - but I though I was seeing gains of around 1.2 x using MIMO ...and for that you need a 4 stream access points (only a few of those around), one stream goes to manage the MIMO process itself, and the other streams relay on spatial separation - which might mean spacing the clients in a circle around an access point all around 15 feet away. Perhaps you can get rates of 1.8 x a single stream in that arrangement - but I didn't see it when they were laid out on desks in a classroom. So yes - MU-MIMO is worth having even at gains of 1.2 - but its not transformational. I had one message claiming much higher data speeds with an Apple Mac. Perhaps they have better clients - I will try. But I stick with my test results which is that "ac" is not generally going deliver a replacement for wired computer suites with 30 domain laptops. My experience is by the time you get to this number the "collisions" begin to have a very noticeable effect on total bandwidth. "AC" is great for BYOD tablet use....but they worked fine with "n" access points too.

[psydii]

I appreciate that you are trying to raise awareness of the intrinsic limitations of 802.11 due to the nature of shared mediums. However perhaps using the work 'ignorance' and appearing to apply it to the readers of your post might have been counterproductive to your objective.

 

That video basically does says the same thing you were, it just takes an hour.

 

I will note however that in a room with 24 2x2 ac wave 1 laptops doing heavy photoshop works (incl uploads of 100's of raw and jpg images) with a wave 1 2x2 AP with 80Mhz channel width in an environment where there is a lot of legacy abgn 20/40Mhz, we experienced (for the use case described) performance equivalent to 100Mbps Ethernet.

(note: 'experienced' not 'quantitatively measured', this was a qualitative observation/judgement)

 

Now, if all users tried to transfer all of the data in minute 1 of those lessons, there would have been a huge bottleneck and complaints as it might take 10 minutes + to complete. However the nature of the activity meant that the network load was spread out across the lesson and devices. Performance in *this scenario* is great on AC. The 10 minutes of airtime is consumed, but spread out into 5 second chunks around the users.

 

Back to the OP though, other things that make a huge difference:

• Modern OS's and their network stacks (at both ends of the connection), ensuring there are no b/g clients taking airtime within the RF broadcast domain.
  
• Correct encryption configuration on the Wireless. AES is needed to get n+ performance. If you have tkip or open wifi, you are limited to G performance.
  
• Channel planning, is there channel overlap (for example on the 2.4Ghz band there should be 1,6,11 only. Nothing should be on 2,3,4,5,7,8,9,10)? Is there sufficient space/isolation between APs on the same channels?
  
• Are there other things going on the wireless (e.g. AV/ Windows updates can involve gigabytes of transfer per client)?
  

[AlanD]

I accept your criticisms - but I do think that WiFi sales techniques prey on the lack of understanding of wireless technology and paint an overly rosy picture of what they believe "ac" can deliver by quoting these large numbers at is. I too have used class sets of laptops (although not with photoshop) and agree that classroom strategy is an important part of getting them to work well for any teacher. Booting the laptops well before the lesson is a good start - because you don't want updates (and gpo s and like) all going on the moment students turn them on....it needs to have happened before the lesson - or before the start of the day. And yes staggering logons and not doing something like saying "Each student should now collect a laptop from the trolley, switch on and logon" - because its just not going to be good experience for them.

 

 

When and where you can use 80MHz channels you clearly will get twice the data throughput to s single device - but in a dense environment you would get the much the same aggregate bandwidth by using two "n" access points...or indeed by allowing some clients to connect to neighbouring access points in the adjoining classroom. I'm not saying "ac" is not worth having ...but its not necessarily going to solve all your wireless problems or be as a good as a wired system.

 

And I'm sure (well - I know for a fact) that there are some pretty poorly set up wireless networks in schools - so yes its worth getting all that right - especially before you rush to buy new kit.

 

I might suggest using channels 1,7 and 13 rather than 1,6,and 11. Its no big deal and may not work well because not all client devices support channels 12 and 13 (although most modern ones do in my experience). Typically its not the default setting for managed access points - certainly not if you go for auto channel selection....probably because Americans forget we have slightly more generous channels. Using 1,7 and 13 give better frequency separation and slightly better data throughput when all channels are heavily used. This is of course tinkering at the edges...but it all potentially helps.

[psydii]

I would caution against any 2.4GHz channel plan that was not 1,6,11. While your proposal might be valid in theory in the UK, I would suspect that any possible benefit would only be seen in super dense, high airtime utilisation scenarios. Also and more crucially, in practice your neighbors will likely overlap and wifi does not react well when it encounters in use overlapping channels. (In my experience, 1 netflix stream on channel 7 is enough to destroy performance on channel 6).

 

Still intrigued as to whether the OP is going to share some more details on their use case, or whether 'distribute on USB/ Ahead of the lesson' was good enough to solve their issue.