Voice Wireless Deployment help - High Utilization w/ no users

Answered Question
Feb 22nd, 2008

“We” have a highly dense AP deployment going on right now in our new HQ building to support wireless voice services.

Unfortunately, the AP's were hung only about 35 ft. from each other. There are about 20 AP's per floor in a three floor building. All three floors have one side without a wall. What I am seeing is that all of the AP's power down to level 8 (the lowest), and each of the AP's has around 30 Rx neighbors. If I leave all of the AP's turned on, the utilization per AP is above about 60% - and up to 100% without any clients associated to it. On the middle floor, I've tried turning off every other AP to see what effect that would have, and noticed the utilization dropped significantly (each AP below 50% utilized), but the power levels never increased on the AP's (I'm assuming because of the gazillion neighbors they see). So, I tried to manually increase the power on every other AP to power level 6. Now, of course, the utilization is really high again.

Is this typical for a voice deployment?

Should the AP's be more like 50-60 feet apart?

Any suggestions? (don't put AP's on the 2nd floor - just 1st and 3rd)?

Please see the attachment.

Thanks!!

Attachment: 
I have this problem too.
0 votes
Correct Answer by john.preves about 8 years 9 months ago

They will stop laughing when it works. I promise.

One of the things I do is set 'mandatory' the data rate 11Mbps, and every thing above that is 'supported'. Everything below 11 is turned off if I can help it. The worst device you have will transmit data at 11Mbps...it doesn't need to go slower. Make it roam to a new AP with the data rate setting and the power setting. (some phones need the 5.5 so try it and see if you can get away with it. This keeps the cells as small as you can get away with (11Mbps).

250 people per floor is a lot. That brings you back up to 11 AP's just for the user density (using a 1:3 ratio of active calls to users)i.e. 250\3 = 83.3 (number of users out of 250 who could realistically be expected to be making an active call at any given time) 84\8 (number of active calls Cisco supports) = 10.5

And Each floor is 250 people! You will be using at least 11 AP's per floor whether you want to or not - now you have to figure out how to shoehorn them in there.

If the clients can hit multiple floors from where they are standing - you have no load balancing anymore. All your math goes out the window. If the floors are identical, I would place the AP's identicaly as the floor above and below. Even with low power, there will be some bleed through. I will put my strongest signal (new radio) right on the strongest bleed (radio from upstairs) giving the client the best shot at associating to the one I want him to associate to. Remember to do your channel allocations vertically as well or else this approach will hurt you.

If any areas are more dense than others, keep that in mind as well.

This deployment will be a good trick. I eagerly await the outcome...

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 4 (6 ratings)
Loading.
john.preves Fri, 02/22/2008 - 18:06

Did the site survey require this many AP's?

In hospitals, 20 per floor is not unusual, but this is due to the usually dense building construction and crowded environment.

If you do a site survey based on signal strength, signal to noise ratio, and square footage you may find you have too many radios for the environment. At some point, too many radios will upset the environment.

I am a huge believer in small cells and lower power. I do not use Auto RF. I manually configure my channels and power settings based on my survey.

In your case, I would turn everything off and start over. For my needs, I need a -65dBm signal strength with at least 25dB (usually 30dB) SNR. I attempt to achieve that with small cells (2500 square feet or less, depending on user density) to roll out voice.

The square footage numbers aren't set in stone but I don't deviate from the signal strength /SNR.

I use the Cisco Site Survey tool and a continuous ping with a -l of 300

(ping x.x.x.x -t -l 300) and I get more accurate results than anything else I've used. (Airmagnet, Ekapuke. Ideally, using the phone to survey would be the way to go, but I rarely have that luxury.

Figure out how many users you need to support and work your way out from there. Usually, in hospital environments, you end up with lots of aggregate throughput just because you need many radios for coverage. but you should not have more than 20 - 30 % overlap. Set your settings manually and turn auto RF off.

This usually affords me the seamless roaming, coverage and aggregate throughput with physical load balancing...since the cells are so small.

Don't forget to check through the floors either. If the signal is just as good upstairs as it is down stairs - the power is too high.

You can try different antennas, but right now work on the survey.

Also, the phones need to be at a lower power as well. Or else you end up with the same situation you just fixed in reverse.

netwrkgod Mon, 02/25/2008 - 05:15

Unfortunately, I wasn't the one who did the survey to begin with - if one was done at all :( Now people just moved on to that floor, so I have to take care when making changes.

When you say "lower power", are you talking 8 ?

Also, this is a wide open floor plan - there are no walls the entire length of the floor and we are primarily using 1131's.

I'll go over there and take some measurements to see where we're at.

Thanks for your input.

john.preves Mon, 02/25/2008 - 14:17

I feel your pain...lower power is always in relation to square footage and user density. You and I do not know what the power setting will be until we keep testing them to determine what will be sufficient for the area and current and future density.

Is this an atrium type situation? How is it you are missing a wall?

How many users are we talking about? with 14 AP's turned on at 8 users per AP- that's 112 active calls. the probability of 112 active calls is slim in a room with 112 people in it. It goes up if the room has 300 people in it.

If you do not have that many users and you have open space with no wall on one side you need to start tearing a few of those down.

Dont be surprised if it turns out you only need three or four.

see if you can get a power setting that gives you a nice cell without bleeding through the floor - much. this may be impossible with the big hole where a wall should be, it will be fun to try...

john.preves Mon, 02/25/2008 - 14:27

incidentally, did you notice how many AP's you were near that were better than -50dBm?

At some point, only having three channels to play with, you are going to step all over yourself.

If I have a channel 1 cell (example) and I wish the 11Mbps data rate to end at -65dBm ( and start looking for a new cell) Than my next channel 1, in a perfect world, should be somewhere near 30dB away...I'm paraphrasing so it isn't wireless according to Hoyle, and we never get what we want anyway. This is where the power settings, antenna selection and AP placement come into play.

netwrkgod Tue, 02/26/2008 - 04:57

Yep, there were always >3 AP's better then -50 in range.

Yes, this is basically a 3 floor atrium. Each floor on this side of the buiding will house about 250 permanent people.

Thanks again for the input, I just didn't want to get ridiculed for turning off AP's after someone else already hung them.

Hopefully this afternoon I get over there and determine the cell sizes and clean things up.

Correct Answer
john.preves Tue, 02/26/2008 - 22:06

They will stop laughing when it works. I promise.

One of the things I do is set 'mandatory' the data rate 11Mbps, and every thing above that is 'supported'. Everything below 11 is turned off if I can help it. The worst device you have will transmit data at 11Mbps...it doesn't need to go slower. Make it roam to a new AP with the data rate setting and the power setting. (some phones need the 5.5 so try it and see if you can get away with it. This keeps the cells as small as you can get away with (11Mbps).

250 people per floor is a lot. That brings you back up to 11 AP's just for the user density (using a 1:3 ratio of active calls to users)i.e. 250\3 = 83.3 (number of users out of 250 who could realistically be expected to be making an active call at any given time) 84\8 (number of active calls Cisco supports) = 10.5

And Each floor is 250 people! You will be using at least 11 AP's per floor whether you want to or not - now you have to figure out how to shoehorn them in there.

If the clients can hit multiple floors from where they are standing - you have no load balancing anymore. All your math goes out the window. If the floors are identical, I would place the AP's identicaly as the floor above and below. Even with low power, there will be some bleed through. I will put my strongest signal (new radio) right on the strongest bleed (radio from upstairs) giving the client the best shot at associating to the one I want him to associate to. Remember to do your channel allocations vertically as well or else this approach will hurt you.

If any areas are more dense than others, keep that in mind as well.

This deployment will be a good trick. I eagerly await the outcome...

netwrkgod Wed, 02/27/2008 - 05:39

Roger that! What you just posted is more inline with what I was thinking, so I'm glad I haven't completely lost my marbles. I will post back with some results later.

netwrkgod Mon, 03/03/2008 - 08:40

So far, I've ended up turning off about every other AP so the remaining AP's are ruffly 45-50' apart from one another. The 3rd floor AP's are set to power level 6 and the 2nd floor AP's are set to power level 7. There are 13-14 active AP's per floor now.

The utilization is much better on the AP's - averaging around 35% utilized. I did a walk through with Airmagnet and the channel utilization is 30% or less on each of the 3 channels and there were always at least 2 AP's better than -65 dbm, 25 SNR, and -90 dbm noise.

So now it's wait and see for the performance results.

Thanks for the help!

john.preves Mon, 03/03/2008 - 12:41

30 % utilization before users is still way high..Do you have the airmagnet with the cognio software on it and can you perform a spectrum analysis with it?

netwrkgod Tue, 03/04/2008 - 05:20

I did a walk through with Cognio running a week or two ago and did not notice any interference.

john.preves Wed, 03/05/2008 - 18:41

Hmmm...are you experiencing significant overlap even with the AP's turned off?

This sounds like a do -over.

I would turn them all off and then turn them on one at a time and see where it gets out of hand.

Hey John,

Love the survey tool humor. Are you really that down on AirMagnet? Have you used their Planner tool and combined it with onsite measurements? I've used the hodgepodge of NetStumbler/Ping/IOS AP linktest in the past but don't we have to evolve?

On a side note, do you think RRM will ever be modified so as to not change channels on an AP when any kind of clients are connected? This is the showstopper for hospitals, and something for which the competition has the answer.

john.preves Sat, 03/15/2008 - 10:36

Greetings-

First, let me say I read all of your stuff and realize that you are way more evolved than me based on your posts. I have some of your answers on file for "future reference" along with a few other folks.

Anyway - Airmagnet as a survey tool (laptop)is still kind of buggy. If I want to lock a data rate and use this as a baseline to measure with, I need it to stay there. Cognio is a wonderful utensil and the rest of Airmagnet is good for other stuff (sniffer traces, utilization, etc.)

Airmagnet surveyor...where do I start? It makes pretty pictures for customers and then when it doesn't work properly, contractors have something to point to that shows coverage.

Ekahau- if you took a survey and then went back and modified your results based on actual signal etc...you would have a very good mathematical representation of your environment. Spectralink does that. I don't like the 'survey one cell' people and then clone the green spot all over the map.

The free Cisco tool tells me the name and mac of the AP I'm associated to, when I trouble shoot. Perfect for trouble shooting. That and a ping test will perform a survey that will stand on its own once installed. Unless, I can actually have the device (phone, vocera toy)I'll use that and measure every cell back and forth and make adjustments as I go. It takes a while, but I have never had to give a paycheck back because the finished product did not perform as advertised.

As far as RRM -geeebus- even if the 'throwing the client off the network feature' were modified (ha)then you still have the '80% overlap in order to throttle down to less than 20dBm' issue. Can you think of a situation where you took one AP and placed three or four more (depending on the date of the documentation you read)at the edge of a -65dBm cell? If I have to do this to allow auto RF power 'figure outer' to work, then I don't really need RRM do I? Pico cells are further away than that.

Our facility is moving away from Cisco and we have been incorporating Meru as we go along.

It's kind of like wireless for dummies, but you can not argue with results. It's very difficult to screw up - but on the flip side - none of your troubleshooting tools work. Fortunately, I haven't needed them.

You didn't used to work for HCA in NOLA did you? Take care-

Hey,

Thanks for the kudos. I really enjoy your contribtions to the dialogue here as well.

I was a Wireless Valley fan but it doesn't incorporate measurement data well, and doesn't scale on cost and engineer adoption. AM appears to have the suite to go with in general. I'm hoping they continue to improve their products. AFA pretty pictures, people like them, but we know they're only snapshots -- to paraphrase a famous philosophers line, you can't step into the same RF environment twice.

Yes the 19dB co-channel separation is a real challenge that software may never be able to achieve, especially while dynamically changing channel and power. Its like asking a robot to make a bed that's rotating and flipping around -- even if you slow it down or stop it long enough, it takes a lot more finesse than it appears. In the end, the clients have to be involved, and be at the center of the decisions, as they have to sleep in it. We'll see what 802.11k/r does for that.

Going with a SCA? Interesting. Very Zen. How do the medical device manufacturers like having a single AP MAC - they've argued with us that they need their own BSSID so as to discard frames from other SSIDs.

There's real value in simplicity, and if the HCF isn't cutting it there's good reasons to give more deterministic polling-style access methods like SCAs deploy a try. I was a fan of HCCA but it got dumped by the industry in favor of EDCA. Check out this doc for the possible reasons why. Although lately I hear it may make a comeback.

http://www2.informatik.hu-berlin.de/~wolter/publications/epew06b.pdf

john.preves Sun, 03/16/2008 - 04:42

Very Zen indeed...hahaha! That was good.

Ummm... I attempted to read your document and it might take about a month to absorb and re-read and Google etc...but I'm thinking the gist of it is the time interval is pre-determined either way (HCCA or EDCA)

Meru has described their time division multiplexing as almost like token ring...nobody talks until your finished.

I might want to look into that further since they were probably over simplifying.

I haven't heard anything about the medical equipment people and BSSID's...I will ask about this...thanks. So far we have voice and data no RFID yet or medical pumps, etc. but that's all in the pipe.

I can however, use VNC which is like a server that hosts huge files like a movie and then use a wireless laptop or ten and connect to the server and view the movie across the wireless, while roaming even. I'm not sure how that works or if VNC does some kind of compression, but it is fairly impressive none the less.

john.preves Sun, 03/16/2008 - 04:57

Actually we are going to keep the Cisco for things like remote sites and guest networking. I would prefer a single platform myself, but that fight has been lost and I resign myself to learning both platforms and remaining quite confused on both.

This reason and God punishing me for an incorrigible childhood that lasted well into my late 20's is why I have no hair.

I just hope the fallout from having Cisco and Meru AP's scattered around in the same room is going to be as agreeable to them when it's finished as it is to them now. We shall see.

john.preves Sun, 03/16/2008 - 22:28

I am very concerned. This is the argument I lost. I brought up that very article during my dissertation. The powers that be have decided that as long as the Cisco that gets whacked is the guest network, then so what?

If nothing else, I get to have a 'production-lab' and see what happens. I will keep y'all posted.

john.preves Mon, 03/17/2008 - 21:43

The hospital and all the boss people above me are pushing the Meru. Actually, it has been pushed already. I have almost one hospital completely Meru now. I have finished surveying one other hospital, currently awaiting deployment, one building of a third (also awaiting deployment), half of one other hospital, and two more in the pipe. Add the guest networking component, another hospital that is on the fence, a whole bunch of doc-in-the-boxes that will most likely remain on Cisco and the occasional break-fix like today and it's a lot to get done. Plus, I'm learning things like switching, trunking and routing from the regular network geniuses. I also get to teach one minion person wireless principles and technique while he helps me put all this together. Truly a righteous gig. I am blessed. (even though its cold eight months out of the year).

Just to clarify - we are still primarily concerned with voice and data. Guest networking is third and somebody is probably talking to the RFID/ medical device people, but I have yet to see a piece of gear for that. I'm sure that is going to happen faster than I can get to it.

Which is why I'm pro-actively doing surveys for facilities that haven't spent money yet.

We never seem to roll out one whole hospital at once. Coverage gets chopped up depending on who's department is paying. I want to have the whole thing drawn out so when they pick and choose who gets what, it will all be kosher in the end.

I hope to have some kind of guest networking up and in test mode this week. It will go in a lobby area where it should get quite a workout with two AP's.

Results will be posted forthwith and breaking developments as they occur.

Peace-

Thanks John,

You may want to include location in your current surveys, but with the proviso that some of the APs deployed will need to be receive-only (peripheral APs). Its probably better to consider this an overlay on an existing data deployment. I would make sure you be diverse in the placement of APs going forward as preliminary preparation. Avoid horiztontal and vertical stacking, offer at least some measure of variation in RSSI for location piloting.

Interesting impetus for Meru. It doesn't sound like you have to deal with pumps yet. We've found these vendors to be unwilling to adopt any non-IEEE spec (or they're hiding behind this so as to not have to re-engineer for CCX compatibility). Hence their own SSID to avoid bcast/mcasts from others (God knows a pump CPU is probably half of what's in my PDA). Single channel will certainly facilitate a client that doesn't have any roaming logic.

I assume you can have multiple single-channel zones via subordinate controllers? Devin Akin of CWNP has been a spokesperson for Meru of late, and I believe/assume he uses the term "stacking" to convey this capability?

We have the same funding model, and it makes a lot of sense for you to do the building in advance to maintain continuity in the deployments. Just make them pay for whatever covers their area.

I come from the wire-side, but spend a lot more time in the air now.

Good luck and keep sharing!

Regards,

--Bruce

Actions

This Discussion