Wireless Access

Hi all,

I'm trying to get up to speed with a challenging situation for a school. The client has 20 new Lenovo, Windows 7 laptops which came configured with Broadcom 2.4 Ghz, 1x1 WiFi cards installed. I have a single IAP-135 AP. These are the only clients attached to this AP.

I am having difficulty in being able to deliver an acceptable experience with these laptops starting with logging into a Server 2008 Domain, with Roaming Profiles/Folder Redirection. The Roaming Profiles have been checked and they are in the 2-15 MB size range. Problems start at logging in with a group of 15-18 users in the same room as the AP. Signal is great, noise is clean. I do see a number of retries being shown in the console, but I don't have the number available to me, as I write this message.

From someone elses experience, would you expect this because of the WiFi cards capabilities?

Everyones help is immensley appreciated. Thanks.

Kurt

If can verify the following to make sure that you are not experiencing slow performance because of RF issues :

ARM history will show you if there's a lot of channel/power changes and why

ARM RF summary will show you the RF quality of the channels

You should enable to Drop broadcast/multicast :

What type of wireless authentication are you using ?

Do you experience this issue only when you have several users ? or it also happens even with just one ?

Make sure there's not packet loss between the IAP and your server ?

After the user  is able to login are you still experiencing wireless performance ?

Have you tried updating the wireless drivers ?

Do you have another laptop that has another wireless chipset to test with ?

Have try this :

http://support.microsoft.com/kb/981830

http://social.technet.microsoft.com/Forums/en-US/20983681-6470-4e71-895c-394b34da4e2d/slow-logon-and-roaming-profile-issues 

Try checking the channle utilization.

You can put the IAP in spectrum analysis(but you wont be able to serve any client)

If you got a fluke airchecker or something similar that would be great also.

The other important point.

2.4ghz and 1x1 :(

Even if you got an AP 135 the client with his 1x1 will take more air time than a client which could be 3x3 on 5ghz

your max speed per client would be 65mpbs and 30Mbps tcp/ip througput

Anywyas  the thing is that you are needing a lot of airtime in here, and its beacuse of the slow clients.

Let supppose that you need 7Mbs constantly for a few of minutes to download the profiles 

Assuming that you just got client capable of 65Mbps which is a tcp/ip throughput of 30 Mbps  and 20 clients at the same time logging in.

This means around 466% of airtime which means that yhou would need around  5 Radios to transmit 7Mbps constantly

I bealive that one AP is not enough with this slow clients  and the BW you needing and thats why you are getting those kind of trouble...

This would improve if you get more APs or you get the client to use better antenas for their laptops... for example if they would be running on 5ghz  3x3 client you would need just one AP.

The problem here could be your slow clients and the lack of APS in this specific scenario.(and even if you add more APS as you can just work on 2.4ghz as your client are just 2.4ghz which could be troublesome! as your just can work with 3 Channels and yhou would need to reuse)

In yhour sceneario it doesnt matter if it a AP 135 or AP 105 and so on. your problem  are the slow clients.

The first thing i advice my clients when they buy the solution to us

PLEASE buy 2x2:2 laptops abgn do NOT purshase 1x1 laptops bgn, this really impact your wifi performance...

Cheers

Carlos

Hi Carlos,

Thanks so much for your response. Would you be kind enough to explain in more detail:

1. The calculation to get 466% Airtime
2. The use of 7Mbs per client for downloading the profile to the client
3. If on 5Ghz, using 2x2 what values am I using for the claculation of Airtime
4. At this moment, there is only 1 AP that I've installed in the location, Am I able to do Spectrum Analysis and check the channel utilization?

Hopefully my questions aren't too dumb.

Thanks again,

Kurt

Dont worry, there arent dumb questions!

Question number 1

To calculate this you must know

The minimum BW you needing for your clients = which i took for instance 7mbs, it would be 1mpbs, you need to ttell me that.  At least you said you had roaming profiles from 2mbs to 15mbs.

You need to know the antenna and speed of your clients which in your case is 1x1:1 Which equal to 65Mbps theorically which equal to 30 Mbps tcp/ip real througput.

You need to know how many clients will be using the minimum BW  which in your case its 20.

Now you know all these numbers the formula goes like this

Minimum BW = MBW

Real BW of your clients =RBW

Numbers of clients =NCL

Porcentage of airtime =%Airtime

Forumula:

(MBW/RBW)*100 = %Airtime

(%Airtime*NCL)/80 = Real airtime for all the clients

So lets use your values

(7Mbps/30Mbps)*100 =  23.3%

23.3% * 20 =  466%

455%/80 = 5.8 Radios Which is around 6 radios.

Question Number 2

Thats a number i though it would be a minimum, if you can tell me the Minimum BW that its needed then i can recalculate with it.

Question Number3

If you were able to use 5ghz in your scenario and you had 2x2 Client antenas the real Tcp/ip throughput would be 150Mbps.

You can calculate the real tcp/ip trhougput  by just cutting the theorical BW by half.

Question Number 4

You are able to do that, but you wont be able to serve clietns if you change it to spectrum monitor!.

What im not sure how much info you can get from Instant spectrum analisys... but with a controller based scenario it would give you that information!

If you ever wondered the source for the formula i just used

http://www.aerohive.com/330000/docs/help/english/documentation/Aerohive_High-Density_Wi-Fi-Design-Config-Guide_330073-02.pdf

Yes its from another vendor but the formula its pretty handy.

There is another one from Motorola High Density guide, which is a bit different but at the end it give you the same numbers...

http://www.revolutionwifi.net/

Cheers

Carlos

If you did the same formula but instead of using 1x1 clients using 2x2 clients you would see a big differnece in radios that you would need! much lessTry it

It is possible to do the same test but with clients with 2x2 antenas? and 5ghz capable?

Cheers

Carlos

Hi Carlos,

I'll be able to spend some time with this client tomorrow, as well as absorb your answers. I'll get a meassage up here with progress results. I do have 20 USB based A/B/G/N 2x2 addapters coming to add to the testing.

When things start to fall apart, I do see a number of retries on all of the clients, but I don't have a point of reference to know what's tolerable or disasterous. What would be the best report/s to run, and what how much duration of time do the IAP-135s keep in their log? Am I able to post them back here for you experienced folks to look at?

Thanks,

Kurt

Hello Kurt

Be sure you got band steering on and that those clients are connecting to 5ghz :)  and have cahnnel bounding on(which is on for default on 5ghz)  Also have airtime fairness on(not on by default on controllers, but its on by default on instantsAP)

Well to see if something is not tolerable for me i use some rudimentary tools like ping. 

If i see to many ping drops 

Let say that you are logging on with your latops with the roaming profile and you got one laptop pinging to let say the swtich core? and you see a lot of ping drops, like 3 times out of 5  all the way, or  4 out of 10 pings this is really bad....

if i see 2 ping drop every 10  pings its bad but not that bad...

I would expect fo rexample out  of 100 pings 1% to 4% to be aceptable  more than that is not good...

Like i said its rudimentary, but it has been working for me just fine.

Maybe someone else can give you some MUCH better reference to know if its dissasterous or torable.

Hope victor can input you with more additional information.

Yeah forgot

GOOD LUCK Tomorrow!

Cheers

Carlos

Hi Carlos,

I've taken a look at the references you provided. Great!!!

I'm working through the training at Aruba, but real world business and opportunities are creating a trial by fire scenario. Thanks again.

Kurt

I DO understand your situation

I was on the same boat before...

Working on the traning but the real world bussiness creating trial by fire scenario :)

Cheers

Carlos

You can see the noise, channel unitilztion  and other info  ere:

The noise should never go under -80... if it goes under -80 then its too high

Check also how much channel utilization you got! before and when you are logging with roaming profiles!

Check also if the drops and errors increase when you are logging with roaming profiles

I think thats could be good info for you!

Cheers

Carlos

Also forgot that check at what speed you are connecting to the wifi.

I got a 1 ABGN Usb adapter... which is a linksys AE3000 which im using on a windows 8.1 and the draiver doesnt seems to work properly... i connect to the wireless network at the speed of just 54Mbps... but if i use it on a windows 7 it does connect at 300Mbps....

So its something witht he driver... and there are no drivers for 8.1 sigh... so  check that also! that you are connecting to the correct speed!

Cheers

Carlos

Hi Carlos,

I hope you had a great weekend.

I'm struggling a bit with the wireless adapers that  I've got go to prove out the client low connection speed problem. Even though all looks good as to noise, these TP-Link adapters, on 5Ghz, are dropping to 54Mbs whicj is certainly no good. My noise is better than -90, so I'm not quite sure what's going on.

I've got an IAP-135 on my own internal network for testing purposes and for gaining experience. What's also perplexing is if I've got one device connected, happens to be 5Ghz, 2x2, 270-300 Mbs connection, that I see a load of retries on both sides. At the moment, the IAP-135 is connected to a 10/100 POE+ Switch. Is is reasonable to conclude that the the connection to Eth0 being 100 is slower than the AP and the client which is causing that congestion (retries).

In the monitoring, is Retries synonymous with Frame Drops?

In your experience do you see a see the Instant monitoring not refreshing? What I'm seeing is the graphs will update for a couple of minutes, then freeze. I have to move to another item to cause a refresh. The FW is the latest off of the support site. It's sort of a hassle IMHO.

Thanks again for all of the great help,

Kurt

Hello Kurt!

Had a great weekend thanks for asking!

Guess yours wasnt the best one with this issue!

Anyways

If your noise is over -90 let say  -95 or something like this is really good... you should have no issue with noise then

Now i see you got the same problem i got with my AE3000 that i posted on a privews message.  At lest the issue with mine is that i dont have windows 8.1 drivers... and thats why it just reach 54Mbs...  if i plug it to a windows 7 machine then it works all good!

Your problem with your TPlink usb wirless adapter must be likely the same...

About your conclution regard the 10/100 port yes you are right, you got a bottle neck in there

Your IAP 135 is capable of 450 Mbson 5ghz and maybe like 200Mbs on 2.4ghz.  So thats a total of 650 trying to go through a port of 100Mbs...

About the monitoring tool to tell you the truth i havent noticed that.   Thats because when im looking at it, im just getting the information i need int he moment and then go to another  tab.  I actually dont stay in there watching the same  graphic.

I bealive that in the client you got 1 Gb port which is plugged to the IAP 135 right?

Does the client does not have some laptops with 2x2 cards? all they got are 1x1 cards??

About your question

Frame retries would be frames that were send again

And frame drops would be frames that never reached their destiny.

Not really exactly the same but something really similar

Cheers

Carlos

Hi Carlos,

A couple of answers:

Does the client does not have some laptops with 2x2 cards? all they got are 1x1 cards??

There are a group of faculty laptops that have Intel 2230s, single band, 2x2, but all of my testing group has been with the 20 student laprops with t Broadcom 1x1 adapters, with only them attached to the AP.

I bealive that in the client you got 1 Gb port which is plugged to the IAP 135 right?

Yes, at he client it is attached to a Gb connection, back to the Servers

About your conclution regard the 10/100 port yes you are right, you got a bottle neck in there

I've connected the IAP-135 to a Gb connection with a POE injector. I'm puzzled with regard to the frame drops and retries. The screen shots are of a MacbookPro, 5Ghx, 2x2 that is the only device connected to the AP. Why would there be this many?

I actually myselft dont trust that much in that graphic, as i see lot of times drops in it, and i dont see anything wrng happening or time outs or something like thatt, on the clients.

It would begood if someone from aruba can conirm that...

I remenber for example the goodput was not something really realiable on 6.1.x though...  on controller, but it seems that it was more trusted now on 6.3.x

It is ot possible to have 2x2 clients  i mean 20 and re run the test?  to see if it works much better? guess its really hard to do but i cant really thing on a way to test it without 2x2 clients...

Maybe someone else got a good idea?

Cheers

Carlos

Hi Victor,

Thanks so much for the input. Very good stuff. I'll be able to spend some time with this tomorrow. I've replied to Carlos, as well with a couple of additional questions so please chime in. I can't tell you how much I appreciate it, and how beneficial it is in understanding a good approach to troubleshooting, and design.

Thanks,

Kurt