pro's, how do you make a site survey from scratch?

[tsa]

ive made more than a few site surveys, but as soon as i go into a multi floor location i begin to feel that Airmagnet surveyor lack options.. or maybe its because i never saw a description on how to make a site survey with Airmagnet surveyor...

The problems i encounter:

One AP for a site survey is not enough, at least 6 APs are needed (2xch1, 2xch6, 2xch11) i need 6 to be able to spread them out and get a good overview, especially if the locations have 3+ floors.. if the building has more than 3 floors and the signal is able to travel through more than 3 floors.. 6 APs is hardly enough to make a site survey with Airmagnet surveyor, if you ask me.

Okay let’s say i begin the site survey in a building that has just one floor and more than 6 APs are needed to cover the whole location. i usually start by turning one AP on and use stumbler or Airmagnet to get a feeling on how well the signal travel around in the building. When i have a good idea, i turn all APs on and place them as I see fit. I make the survey and save the job and analyse it. (Is there any overlapping? is there any dead holes?) if i don’t see any problems, i continue, otherwise i adjust the APs and start over.

This is one of the problems I have. If im in a building where the 4-6 APs I bring are not enough, I have to reuse the APs. But as soon as one of the AP is used more than one time in the same site survey, Airmagnet surveyor deletes the old AP icon since it the software know the APs mac address cannot be 2 places at one time… so I have to stop and save and start over or ignore that the icon disappear and continue (surveyor remember the signal around the icon even it deletes it) anyway, in a multiple floor building… moving the APs are a really bad idea since you never know when that particular AP signal are active/needed on the surveyor once you continue surveying on floors above and below it..

The only multiple floor survey ive made where I KNOW that they got the exact right number of APs installed and placed perfectly was on a school that turned out to be so hard to “predict” that I made a survey on first floor and after the installation was done and installed, I then scanned the 2. floor for dead holes and was able to pick the exact spots where another AP was needed… but this was only because the first floor was all done with 17 APs installed and running… without them I would never be able to place the APs on the second floor and get it right without overlapping channels and dead holes.

A work around could be to “do the best possible job” and then call for a check when they are all installed.. but then again its not so great having to explain to the customer that there is actually 4 APs too many… damn a long post… I better not start on the 16 floor hospital I did some time ago. Such a job really teach you a couple of things about wireless and no software I know is gonna help you much unless you do the survey with 80 APs. Maybe Lanplanner is better? Ive heard some good stuff about how it is able to predict the APs with just a drawing but I think its very exspencive.

if you really read all this im would be very interested in hearing your experience with surveys, how many APs you use and software and if you have encountered any problems or know a better way to work around them.

thanks in advance

[M/Q]

My personal experience is on a smaller scale than some of the pro's here. But, I did not rely on the software as much as you are. I have used AirMagnet, but more for RF interference and getting a feel for the coverage area. The rest was a spreadsheet setup or just experience. I rather not rely on software for that aspect, as I have had times where it was not correct.

FlaWireless recommended this product in your other thread, have you looked at it? I value his opinion as well as Spiderbite and Golfnut. They all hopefully will chime in. As they have more experience with the big installs.

http://www.ekahau.com/

[tsa]

Thanks for your reply.

yes, even before i posted the first time i talked a friend of mine. he told me its a good tool, running on html and is alot cheaper than Airmagnets surveyor. That all i know so far though.

i agree with you that also for me a good site survey is alot about "feeling" rather than a step by step procedure but my customers really like a report and the surveyor (as you know) has a nice viewer. i usually give my customer the viewer and the site survey on a CD so they can see the colors and stuff.. they like that but it require a full survey with all the APs.

in the old days when there weren’t so many tools i did surveys with stumbler and 1 AP, those were the days!! the report was a print with digital pictures of all the spots and channel explanation. the survey was faster but the report making was slower :)

M/Q, the "active" and "passive" survey mode, what do you know about that? can that option be used to make a survey with fewer access points? it would be nice if its possible to place 1 AP and scan the whole building and when you have the signal strength recorded, then move the same AP to a new location and scan the building again and then be able to see both scans in one picture...hard for me to explain in English... i know its possible in airmagnets surveyor to gather surveys in the same view. but im not sure how great it is if we talk 40 single AP scans ina 6 floor hospital and how the overlapping will look on the drawing, if at all :)

anyway i love to hear if flawireless, spiderbite or golfnut or others has something to add, especially if its knowledge on how to make a 3+ floor survey and how many APs in use and the general approach

[spiderbite]

Ok dude, here’s the secret to the universe.

I still only use one AP at a time and I won’t do it any other way. For starters, I do not have the extra space and / or meat on my bones to carry the extra gear around with. Second, I will never, ever trust a piece of software to survey for me. I need to see it play out in real time with the equipment that will be in use in the environment.

Now, from your post I am compelled to ask you some basic questions about this particular project. What are the basic requirements for the WLAN in this hospital? I do a lot of hospitals and just to expedite this process, I’m guessing some sort of patient charting software whereby they scan a barcode on the patient and meds etc., possibly VoIP for the phones and maybe even guest / physician access to a separate network for internet, email etc..

How am I doing?

The problem that you will run into if you haven’t already is that your cells are too big. This will manifest itself in poor roaming times for laptops and phones, no load balancing across the network and a nightmare to trouble shoot.

Your problem is not the methodology per se, but you can’t use that methodology in an environment that will have that many users.

Scenario 1

Let’s say about 8 AP’s somehow lose their mind, and allow association to them but no network access from them. You have 8 black holes in the hospital. With your current layout you could have calls coming into the help desk from three floors in either direction presenting that the laptop “died”. All of this while the laptop next to it works just fine, because it’s associated to something else that works. The entire hospital is affected and you have to figure out why. This isn’t just a production environment. You could croak someone.

Scenario 2

Random complaints about slow network times. Some AP’s have many users while some have none. Or some laptops will hang on to a signal that is just enough to still associate, but not enough to send comprehendible data (or voice).

Scenario 3

Equipment failure occurs and no one notices, because the signal blasts all around from everywhere. How do you justify to the CFO that he /she really needed that gear when it’s been down for three months before anyone knew about it?

I hope I’m not hitting a nerve here, but since I probably am, these are all actual experiences I have fixed over the years because of installs like the one you advertise.

Make your cells smaller; strike a harmonious chord between bleed through and square footage. If I tell you anything else I have to charge you. It’s the greedy capitalist pig in me, I can’t help it.

Just for the record, I did the same thing for about the first three years of my career. But I saw the light and crossed from the dark side.

My personal opinion, depending on building construction from medieval to modern, there isn’t much in a hospital that needs to be more than 20mW with a 2.2 dBi omni on it. Most stuff I do is lower than that.

Yes, I use a bunch of AP’s and the surveys are painstakingly meticulous, but that upfront cost results in a quiet, happy hospital. All the users have the aggregate bandwidth they need for their various utensils, the network is balanced based on who is using something where and when something does break, you know exactly where. It also allows for migration to newer technology without enormous infrastructure costs… (Got that one from the brochure).

Don’t feel bad, because without you there would be no me. I would probably be driving a truck…For the last two years I have made a pretty good living out of fixing busted stuff.

Take this and think about it…I could be full of poop, so read everything you can about it else where. I’m not a genius but I’ve spent a heck of long time doing homework training with the Jedi’s. Plus I surf these boards…don’t knock it bro –

You’re in the right place….

Be well,

Spider-

[golfnut]

Spider and M/Q summed it up... It's about feel and product knowledge. We use Airmagnet but we don't depend on it.

Greg

[spiderbite]

To be honest we just started using airmagnet version 6 and the incognio spectrum analyzer thing...I havent even installed it yet...I'm going to have to get one of the guys to teach me this stuff.

[tsa]

Okay i might try (again) to work with smaller cells. There are 2 reasons why i stopped doing this some time ago. The first reason is that i didn’t see enough changes in the signal strengh/lengh when i tested it in open spaces. i mean if there is nothing at all to block the signal (big room and long halls), my tests between 100mW and 20mW showed little difference, though again when thin walls (even without iron in them) and wooden doors and such weaker blocks was involved i did see a big difference. however, i will try it again, this is a test i made long ago on a cisco 350. no one use that anymore and my next job is with Cisco 1200 APs or newer.

Another reason why i stopped is that i didnt trust the customer to adjust the mW in case they did a replacement. that could turn the wireless into a mess, right there, if i had left the mW at 20 and the customer 6 month after replaced the AP with a new one at 100mW. so for safety reason i became used to working with default mW.

But i see your points, especially for trouble shooting and bandwidth issues, small cells would prove much better.

i agree 100% with you about trusting the surveyor, god know how many times i told my boss that he should just give me the Airmagnet and 1-2 APs and i would do the job better than with 6 APs and the Airmagnet surveyor, but he was more concerned about giving customers a viewer and surveyor report on a CD. but now im the boss so that might very well change... the surveyor is in my opinion MUCH better to troubleshoot for overlapping and missing signal strength on a already installed wireless network. Just walk around and look.. So easy..

hehe well ..

Scenario 1 – I would tell the customer to triple check the security configuration on the pc (wpa, tkip, password, EAP/PEAP/LEAP whatever - check it even more) ive seen this so many times and i never failed when i said it was a client/security problem. i mean ive never seen a AP loose it mind in that way.

Scenario 2 - i would suggest the customer to replace his APs with new once that addresses load balancing, by taking and giving associated clients to neighbour APs or even better look into Airspace where you have a "Mother device" controlling all APs. I think this is a future in large wireless implementations now. A bunch of stupid APs - 1 box to rule them all - a piece of GUI to do it from.

Scenario 3 - in a perfect world even with 100mW cells i aim for little to no overlapping, so if a device fail, the customer hopefully notice  but you are right, i need to look into the options of small cells and focus less on the application and the nice report it gives.

Can I ask what kinda report you give to customers after a survey? Ive been using 2 different once, one I wrote my self with digital pictures and one I made from Airmagnet surveyor.

Thanks all

Almost forgot, yes about the hospital you did great in guessing, what ive been working with is called in Danish (EPJ) Electronic Patient Journal. And you hit the spot perfect when you addressed the separated guest network. i did that with 802.11a (EPJ was done with b/g) due to security reasons and options at that time.

sorry for all the grammar and stuff.

[M/Q]

Your grammer is fine, and see all you have to do is put more faith in yourself and your skill set. I can see where management thought the viewer and report were neat, but the customer could care less, they just want it done, done right and the best job for the least amount of money.

Like I said earlier I keep it simple with the reports as what do you do if you use something other than surveyor. I just keep it understandable and if there is a group presentation I might make it into a powerpoint show.

[golfnut]

M/Q is right on... The customer usually just wants to know that it's going to work and what's the return on investment. Some will attempt to take advantage of your expertise to install themselves so be careful.

We've only had one customer ask for a signal map and depending on the size/complexity of the project, we charge a fee for the survey.

We do make drawings with AP locations/signal coverage for our own planning.

As for cell coverage, we plan some overlap in all of our installations and we also monitor our networks remotely so we know if a device is offline. This keeps the customer operating if an AP goes offline.

Greg

[spiderbite]

ok here we go...


Scenario 1 – I would tell the customer to triple check the security configuration on the pc (wpa, tkip, password, EAP/PEAP/LEAP whatever - check it even more) ive seen this so many times and i never failed when i said it was a client/security problem. i mean ive never seen a AP loose it mind in that way.


Cisco 1200 AP's with IOS firmware will not allow you to associate if one of your parameters is incorrect. Unless of course it does....which we discovered was a bug in the firmware. Total ****age.



Scenario 2 - i would suggest the customer to replace his APs with new once that addresses load balancing, by taking and giving associated clients to neighbour APs or even better look into Airspace where you have a "Mother device" controlling all APs. I think this is a future in large wireless implementations now. A bunch of stupid APs - 1 box to rule them all - a piece of GUI to do it from.


All enterprise AP's address load balancing, however, how much better is your install if you can demonstrate physical load balancing? Cisco bought airspace and we will be using it soon, but why the heck do you go through the trouble of doing the survey if the appliance will massage the results through the course of the day? Suddenly, your channel allocations are gone, your cells fluctuate and interfere with stuff upstairs, downstairs and cause more problems than it fixes. Please seriously consider this important step. there is nothing like the feeling of demonstrating to the CIO et.al. that when you roam from cell to cell that you roam into the next one on the map and not one three floors away. That is what they expect.


Scenario 3 - in a perfect world even with 100mW cells i aim for little to no overlapping, so if a device fail, the customer hopefully notice  but you are right, i need to look into the options of small cells and focus less on the application and the nice report it gives.

Dude, if you have 100mW cells in a hospital, you have overlapping. Gobs of overlapping. I promise. If you do not, then I can safely surmise that you do not have enough aggregate throughput for all the utensils being used. P.S. how do you roam with no overlap...think about it...

Most people survey for what they expect to cover, but fail to survey to discover the edge of the cell. Which is how stuff gets three floors up...they dont check. Start aiming for about 5000 square feet or less with the least amount of bleed through between floors. Then aim for 20-30% overlap. Also keep in mind that overlap is not just linear. If three cells overlap one and do not touch each other, added together is a bit more than 30%...which is ok so long as same channel cells donot overlap (much)...


Can I ask what kinda report you give to customers after a survey? Ive been using 2 different once, one I wrote my self with digital pictures and one I made from Airmagnet surveyor.

After we discuss a price since I get paid to perform the survey, they get a map of where it goes and a report with landmarks etc..pictures are nice. The map contains the cells. The report discusses who is responsible for what and how much stuff they will need. They also get a quote to purchase gear and put it in.

After the install, they get a spreadsheet complete with AP name, floor, exactly where on the floor, power, antenna, mac addresses, serial numbers,ip addresses, switch info / ports etc...I also certify that the network will perform as prescribed.

[M/Q]

Hey Spider,

What you said about Phoenix, goes ditto here. Great stuff and it ain't in any book.

[tsa]

by overlapping i meant same channels, of course i have overlapping, other wise i have no roaming, heh. And yes, bugged firmware can alter any rule of thumb before you know it.

Speaking of roaming i had a interesting job some time a go, this company had wireless in big storage hall with trucks driving around real fast. Roaming between 3 APs in few sec sometimes. Another “special” thing about this place is that they used PDA with Citrix clients, a session demanded application. Meaning, you log on to the server and if you loose connection you have to log on again.

Now this place had an AP that didn’t work, so every time they passed this part of the hall they got disconnected, they didn’t notice right away so there was no indication of this faulty area, though i soon found it and got the AP to work. 2 weeks later they called again, "it helped us a lot but we still have some problems, please check it again". I was like... ill be damn... and went to see what was up.

They still got disconnected from the Citrix server now and then and only the wireless users driving trucks so i started walking around again with Airmagnet and a ping session towards the Citrix server in the halls. I then noticed, when i roamed, i sometimes lost 2-3 ping replies, sometimes more. Normally no one notice but if you are on a session based application as the one they used, once you loose connection you loose the whole session and have to log on to the Citrix server again.

I had found the problem but.. It then stroke me, when they roam as fast as they do.. or just roam at all, the switch get a little dizzy now and then because, as we all know this is what happens.

User is connected to AP1 - AP1 is connected to port 1 on the switch - switch Mac address table is showing that anything that is looking for this User has to get passed through port 1 to reach him.

User then roam to AP2 - AP2 is located on switch port 2 and now suddenly, for a short amount of time the switch has the Users mac address on switch port1 AND switch port2.

The Citrix server is trying to maintain the connection but when the packets arrive at the switch from the server the switch don’t know where to send them and the switch goes "screw it all i cant have the same Mac address on 2 different ports so i flush it from the table and relearn it"

User gets disconnected and log on again.....

i believe this can be fixed with multicasting? the AP will multicast to the upstream device (switch) that now user is roaming from AP1 to AP2. But normally this is not seen as a problem or configured, unless people use a fragile sessions based connection like this customer.

This whole problem i solved with hubs. i told him to put all his APs on a hub and into the switch. as we all know a hub does not maintain a address table per port, its a shared bus that broadcast everything so what ever AP the users was connected to, the packets from the server to the users always travelled out on the same switch port, the one the hub was connected to so there was never any duplicated Mac addresses or table flush, long story.. but it was fun the next time i had a customer saying.. "Dude, we have some problems when roaming.. Its really weird because…" ;)