Keith Parsons - Top Ten Tips for Wireless LAN Designs

[Music] [Applause] [Music] technical difficulties are always fun aren't they especially when your cto turns off his microphone pack well anyway we're going to move forward and in this next presentation we're going to be hearing from keith parsons and he's going to give us some great tips about wi-fi and maybe how we can avoid technical difficulties so as you know keith parsons is well known within the wi-fi space gives us a lot of great valuable information at his conference and also provides a great of a lot of great content online particularly with the podcast coming out of wlan professionals so what we're going to do without any further ado is turn it right over to keith [Music] welcome to the cwmp presentation on top 10 tips for wireless lan design my name is keith parsons and i'm going to be your host for the next 30 35 minutes or so that's the time frame so uh buckle up it's going to be like drinking from a fire hose just a quick slide about me you can find me on twitter keith r parsons on twitter i've been around the industry for a while if you uh need anything just ask i love con conversing with people via twitter or via email it's great you can also go out to the wm pro's website and see a lot of stuff there to get started we're going to look at a view that i put together it's a graphic on how to troubleshoot wi-fi wait i thought this session was about how to design yep this was about design but we need to understand the process and the flow of how it works together so if you look at the next picture here we're going to be focusing on the red side your network infrastructure local area network is the blue and the wan and the internet is the green but we're gonna be focusing on the red part of this at the end of the slide you can of this presentation you can go and download this graphic and look at it yourself i strongly suggest you figure out how wi-fi works it's like the core of what we're doing here and there's a graphic that helps you place things in in space so you can understand how they all work together well actually we don't start with design do we we first have to start with define what is it that you want there's a definition stage and personally when i'm going to a customer i'm going to spend half maybe three quarters of the entire design time and budget on the definition because if you can define it very precisely the actual design part's really easy so define then design and then someone does deploy now some companies the deploy is the same people in my firm we usually offload the deployment to some other group who pulls the cables and mounts the aps etc then there's a validation one of our rules we're going to cover here is about validation very very important to validate your design and then the last d here is diagnose i've got 10 tips for you today and yeah i had some extra ones so i put them in the front rather than the back all problems will look like a wireless problem the minute you fire up the access point so do whatever you can to make sure the wired infrastructure is working before you put that last component of the access point in place because as soon as you put the access point on the network it becomes a wireless problem and we know most problems aren't wireless problems so why would you make it even worse so i like to have in our installs we build from the core out to the edge rather than reverse because if you put the ap up first and then you pull the cable and then you configure the switch and then you configure the router and then you get back to the controller there's a lot of things that could go wrong i would rather have all everything except for the wireless working and tested before we get there another quick tip delegate we can't do everything this is wireless lan design tips so anything that doesn't have to do with that delegate it out somewhere else know your phi what's the phi the physical layer you need to know how a nic works now nicks are entirely different in wireless than wired we think of a nic as something that's just convert ones and zeros on one side into some form of media and send them on the other side but if you start down at the neck layer with wi-fi you'll realize that the radio tap header is something that the receiving nic adds to the frame that comes in meaning every single device receives a different radio tap header they may have different spatial streams they'll be in a physically different location they have a different antenna so even though an access point transmits a single frame payload that's identical every client that hears it is going to hear it differently that's a pretty big change what we're doing we need to understand rf media how it works the whole modulation encoding scheme and what happens when you change from a 20 to a 40 to 80 megahertz wide channel that's changing our network and don't get suckered i don't know a better word to say don't get suckered into thinking there's something magic about 80 megahertz wide channels they're obviously four times better no they're not four times better the things about the phi spatial streams modulation coding and the entire contention process are what's going to drive your design code channel interference wi-fi 802.11 is 100 times more sensitive to other wi-fi than to non-wi-fi it's built right into the protocol so don't ignore the parts that wi-fi is the number one enemy of other wi-fi it's going to have to come into your design process you need to also understand how a single frame is transmitted because if you don't you're gonna it's gonna come back and bite you in this picture the only thing that crosses the wireless to wired boundary is the little teeny blue sliver all the other parts take time and again in in wi-fi the thing we have least of is spectrum and so if we can take the air time that's being used and make it more efficient we can get more data to go over the same network so if we want to solve for capacity more devices attached in the same area what we're really solving for is speed if we can deliver faster data we can make each user use less air time and if they use less air time more users can use it so capacity is about speed it's not about adding more aps yeah ap salesman would love you to know that but that's not really what it's about there is a chunk of time in this example about 38 of our time is just wasted in arbitration that's because the protocol that runs under wi-fi uses this distributed coordinated function just if you take those words apart you go what we're conor we're coordinated but we're not we're distributing coordination yeah what we're doing is we made up a system that didn't require things to talk to each other in order to work and it does it by using a whole lot of time to get there there's a whole lot of overhead for rcscts and then we have our little teeny payload and in front of the payload is some more overhead 802.11 overhead and then we have a sif sack at the end to get a little block act that comes in out of all the time we spend in the air a very small portion of it is actually going across the wired boundary all the rest of the colors other than blue they're in the air but they're not on the wire again if you understand this you'll also understand why it's really tough to get to a position where you need m gig to support aps now you might want to use mg nothing wrong with mg i like m gig please but that's not the reason you don't go to m gig isn't because wi-fi demands it just not showing up let's show another example now of the differences when we're looking at two different things the first one up is going to be 20 versus 80 megahertz channels and the second is what happens when we take the best possible wi-fi and compare it to the worst possible life first up is 20 versus 80. the air time the contention to get the airtime exactly the same for both of them rtscts exactly the same for both of them and in this color scheme this is using a different tool green is the payload and you can see as we went from a 20 megahertz to 80 megahertz yeah the the payload went faster all the other pieces didn't so the actual net savings in time is very small and i'm not even showing here the downside of going to 80 megahertz channels is you're going to lose 6 db of snr additionally you're going to have a whole lot of the primary channel versus secondary channel and the obs has overlap all sorts of other things for this little teeny game now your end users when they go to a 80 meters channel they'll see this bigger number but that number is only for the payload we really need to be looking about the entire airtime now the bottom of this slide also very important the bottom of this slide is showing what if we took a very poorly run wi-fi network it was designed with only six meg for the management frames 24 meg is the max we'll allow to go through for the data rate we're not going to allow it to float faster we're going to lock it only at 24. 20 megahertz wide channel single spatial stream mcs of only three that's a really poorly performing wi-fi and then we're going to compare it to some place where we put the control management frames at 24 meg we allowed it to float all the way up to full mcs9 two spatial streams and you go yeah it's faster really not that much faster the overhead is still there yeah the control frames went a little faster and it took a little less time but it's not 4x 5x different it's less than 1x differential between the fastest and the slowest wi-fi again it's all about air time and the reason rule number one is no the phi is because that's where all these little things happen another thing that happens in your phi is how does a client join there's an entire process of how a client goes through association probe request probe response dedication request authentication response association request association response and then the client can get on the network and then at that point there's either open psk or dot one x that's the authentication portion and then there's an encryption portion and then we have port control then we hit the upper layers to get dhcp dns vlans etc once you have all of that opened up and you can actually hit the lam then you hit the captive portal so if you have the captive portal that means all of the 8011 things are already working a lot of our end users think the captive portal is the beginning it's actually the end so you should really understand how this whole flow works and if it you're troubleshooting where each of these things happen i like to design with a open authentication no encryption no captive portal yes so i can prove it's not the wi-fi test your wi-fi separately and then add each of those other functions in that are on top of wi-fi 802.11 is the blue 8011 association all the other things are on top so designed to make sure that works properly the other piece is actually more about other things not the actual 8011 itself there's a whole lot of requirements and we usually get a requirement from our customers who come in and say something like we just need the wi-fi to work everywhere fast yeah it doesn't work that way not to mention we can't design wi-fi for all wi-fi clients just as example iphone one had wi-fi now if you were gonna design to make it efficient for an iphone one today you're designing for 2.4 gig not for five and that's just one thing had that antenna the bad chipset years go by and those older chipsets slower yeah use more air time yep more airtime means less density so we can't hit capacity if we're designing for those older devices so we need to know which device we're designing for which one device we're designing for now if we pick the right one hopefully all the others ones will come along so you need to know all your requirements now i put on the slide a couple of groups here the group in the green box are the things that you can put into your design or validation tool primary career coverage secondary coverage some people call that overlap yeah i don't do that anymore and if you want to look it up there's a fallacy of channel overlap a white paper i wrote probably a decade ago now about how we don't do overlap in percentage we do overlap by dbm also what goes in which frequency what goes into four what goes in five gig if you allow your customers to answer this one they will eventually if you give them enough help get to the point where they realize the only thing that should be into four are two four only devices you get the use only two four they stay together the only thing that should be in two four are the two four only devices you don't want a five gig capable device over in two four he'll have a bad day one going to two four means he'll have a worse day than he had in five gig not to mention he'll be taking away a slice of capacity from something you can't go anywhere else bad bad don't go that route another thing we should look at is coaching inference we should design for we should measure for it we should validate it it's the killer of wi-fi i can just throw up ap after ap after ap but if you have two aps on the same channel and they see each other you have the capacity of one ap actually a little less than one ap because they're gonna be sharing so coaching reference you definitely have to put it as a requirement you design to it you validate it and you make sure it stays as far away from your network as possible device to radio ratios there's not a lot of devices that say this when you buy a chromebook it doesn't say no more than 17 chromebooks per ssid or per radio or per ap they don't say that the things that really need that are usually voice over ip handsets and they do part of the reason they do is because the way they use the phy in qos so figure out what your device is what it specifically needs and make sure you give it what it wants both primary secondary code channel interference data rate snr all of the things now i'm not one to to dwell on terms like is this area high density or low density or medium density i don't care about density what i care about is what's the requirement for a school the main requirement might be all the classrooms if there's 30 devices in the classroom or 50 i don't care that's the requirement so that's normal density what i care about is special density areas in a school it might be the auditorium in a hospital the normal rooms the hospital rooms that's normal but if you go to the nurse's station and nurses congregate there now that area is special so it's more about designing to make sure you meet where the special density is a stadium that's normal density for a stadium you go out into the walkway or to the tunnels that's a special density because you have a different density than where people are so think about those now the ones on the right jitter latency end-to-end qos there's a whole bunch of things over here we don't measure these with our echo or tambo soft or air magnet they're more of a vendor specific they're still valuable they're still very important you need to know them they're a strong requirement but we're just going to measure them using different types of tools so this one it's all about the requirements it's not about the green on twitter i don't even know who they are but there is a wi-fi hulk and whenever someone mentions green there's a little twitter message about being the hulk because people say it's green it's not about the green i can make any wi-fi look green you just adjust the values i can make it look purple or blue whatever i want just being green doesn't mean it's good green doesn't mean good now an example here this is from a high school we did a survey in high school we put a bunch of aps up we did a bunch of testing and i mean i'm going to tell you about the testing but let's just look at these two little heat maps the one on the left is from a mist 43 ap by the way both aps are at the exact same transmit power the one on the right is a cambium they're both wi-fi six the missed one on the left happens to be a four by four the one on the right happens to be an eight by eight now if you looked at just the green now again these both have the exact same scale and they were collected with data from the exact same survey and you look at the cambium on the right and go oh that's really bad look how look how quiet it was neg80 all the way out that 100 meters by the way this is on a football field we but we stopped at 10 meters 25 meters 50 meters 75 meters and 100 meters and then you compare the data when you look at the data the missed ones on the left is in green pretty decent at 10 meters doing 120 meg data rate yeah good stuff now this was a combination that where the data came from as a combination of macbook lenovo with an ax chipset a wi-fi six uh ipad and a wi-fi six iphone this is the average data and you can see the one on the right the cambien outperformed the one that was more green it's not about the green it's about what are you actually going to achieve so strongly strongly recommend you test what you're doing how do we do these tests the last slide i showed you some test data yeah we wanted to do testing testing's important but we all we didn't want to use like really expensive exeterior kind of things or a faraday cage in a lab so instead we went out to a football field we got access to a high school football field about they're 100 yards long with two 10-yard end zones but instead we switched it up to meters and we counted out how many meters it was so we could be more international we set up some aps on one end they were up about 1.4 meters high off the ground we did surveys lots of surveys we walked with an ipad with an iphone with a gps with a stop and go with a continuous mode all sorts of things we did a back and forth horizontally following the lines on the field but also vertically going north south we did it with the device on the hip on the head in different locations we followed around with a drone we got some drone pictures to show you of the site uh we did a time lapse video of while we're there so you could see what we're doing the goal was to go on site and use normal equipment we just took some aps that we had in our labs normal laptops just some ones we had and then we took a pair of wlan pies and wmpi is just a little handheld device we happened to have one with us on site and it was our local server we put another one back at the school district's corporate site so that was our remote server and then we did different kind of testings so testing local was client to ap directly to the wlan pi we were using a html5 test deck so we weren't doing speed testing out on the internet we were controlling it all on our site and you can run the same tool on your own equipment anytime you want we wanted all of these tests to be totally repeatable and so you can test them yourself so in this test we did test local we also did tests remotely so we could tell the difference between a wan link and a wlan link uh the examples i just showed were what's the wlan link i want to know what how i'm doing from client to ap directly to the test i don't care about the backhaul now we've done so we did some testing on other back calls as well but if all you're trying to do is compare ap to ap or client to client you want to have it have it local there four follow best practices now how do you know what best practices are first of all you can go to the vendors whichever vendor you're beholden to whichever one you're installing or testing you're using and they have best practice guides now some of the best practice guides don't always match each other but if you try to meet what your vendor says to do it helps a lot you could also learn from industry leaders from other people who are presenting at this conference people present at the wlpc conference other vendor conferences there's lots of presentations you can get we have tons of videos on the wm pro's website now in our industry we have a term bad fi i don't know if i like the term but do your own testing just because someone said well the best practice is never hang an ap on the wall like a clock i've said it myself i think it is a best practice to never hang an ap on the wall like a clock have i broken my own rule sure there are times when you need to break the rule for certain reasons so don't be afraid to break best practice rules in fact i encourage you to break best practice rules as long as you know why you're breaking it just because you saw somebody posted on the internet and there's a cool oh look you can do this with an ap yeah wi-fi is so extremely resilient that even when you do it wrong it still works so just because it's working doesn't mean it was the best practice to do use your best judgment and adapt use the best practice until you can show that there's a reason you change the rule you put an ap on the ceiling rather than on the wall not just because it was a best practice but because you tested it you chose to not put aps in the hallways not because keith told you so but because you can see that there's an actual benefit to doing it best practices are just that the recommendations if you follow them usually most of the time you'll have a much more efficient wifi yeah this is one that i really yeah i just gonna say it right out don't do it don't follow marketing ratios they're not engineering they're purely marketing one ap per 2500 square foot well maybe on average in that building at that time no that's not engineering that's not design that's just throwing stuff on a map i even know some vendors who just tell you oh yeah high densities every 2000 square feet uh low densities every 250 square meters they just have these throw these numbers out it's the people who are trying to sell you things that tell you that don't do it use the tools use math like real math free space loss math etc to figure out how to do it now here's an example i show this example a lot of times but hopefully it'll help you understand if you're going to design a little dog house you go down to your dyi store you buy some wood and lumber you might sketch it out on a napkin and you hammer it together it's really actually easy you don't have to have a super design and i doubt you would do a snow load analysis on that little roof or how much wind it can take on the side or whether or not if the dog bumps into it it's going to collapse the size of the screws you need to put it together it's build it wi-fi is a lot like that when wi-fi is small scale you put an ap in your house done you get a bigger house you put two that's very simple at the small scale but as wi-fi scales larger and larger it gets more and more complex like this bigger barn instead of dogs if we had horses they're a lot bigger and so now we need to build the barn bigger it's still made out of wood but now we have to deal with wind snow rain how much weight can it hold what kind of gap is it and if i exceed a certain size i have to go from a truss to a glue lamb to a steel frame yeah that's all engineering wi-fi is like that the larger you get the more complex it is and so don't use a ratio 180p per classroom yeah i know you can make one ap per classroom work you can make one ap per 2 000 square feet work you can make wi-fi do anything you want that's not the hard part the hard part's making it efficient and that means using engineering tools always always always always validate what you're doing now there's a lot of different ways you can go through validation you could do an ap on a stick when you do an ap on a stick you actually put an ap up on a tripod now you can do this before you design as a technique to measure rf attenuation in the walls to measure how if the maps you were given match you can do an ap on a stick after you design to validate the design you're not validating the install you're validating does the design still meet my requirements now i like to do validations post install now in the post install validation ram there's two big categories active and passive you always have to have a passive in fact echa how makes it so you can't do an active without having a passive passive the most important passive means i'm listening to everyone neighbors mine my ssid somebody else's i want every all the information i can gather together compared with an active and active is when you have your client device associate to an ap and you're either pinging or sending iperf data realize the active survey data you receive is going to be extremely flawed because think every time you're near an ap and you walk away the signal strength gets weaker and weaker and weaker your ping times get higher and higher as you walk away and then all of a sudden you're right here and you're next to the next ap as soon as you associate to it it's amazingly fast and so you get with active surveys is this degradation as you walk away from one ap a little break and then really fast as you associated to the next ap if you did an active survey clockwise looked at the data you would see it skewed there would be a tail going away from every access point in an active survey away when you went toward nap you're not going to have any active data because you were on the previous ap now there's nothing wrong with active data it just doesn't tell you what you think it tells you you think the active survey is going to say you a a user here will receive this totally false you're walking some weird path that normal people don't do you're collecting data from aps you might skip three aps because you roamed over them just realize active survey data suspect and every time you roam you'll get a drop and so if you're plotting out ping times you will get drops every time you roamed that does not mean that this location is a bad wi-fi you might have fantastic wi-fi there it just happens to be where you did that little wrong stop and go versus continuous i love continuous i hate stop and go stop and goes for losers stopping goes for people who don't mind throwing away their data because when you collect data with a continuous every time you walk click when i start click when i stop click when you turn very simple rules it's all you have to need to know and you go everywhere all the data in between every one of those clicks is being collected you flip into a stop and go mode you click you wait takes longer time you then when you're done with the stop the go part is your walking and you know what happens as you're walking you're throwing all the data away you're collecting data the data is actually being collected right then and you decided since i can't walk in a straight line and i don't know how to click when i start when i stop when i turn i'm gonna willingly throw all that data away all you have to do to see this is look at your data set and you will see that stop and goes use continuous mode now in continuous mode we have other choices we can go autopilot we can do gps we can just walk and click you can use an iphone an ipad there's all sorts of ways you can get around these always choose the ap's before you start designing the reason you need to do it before is there's such a huge difference in access points there is no generic access point you can design for and then install and have it match just doesn't work that way so one i wish there was never a generic ap allowed because you can't buy one so why would you design for it now in the example here on the slide in the upper left you see a 10 db is the transmit power that's leaving the rate that's going into the radio the radio then sends it through its antenna and what we get out is eirp in this case this ap had a negative gain on its antenna totally possible a lot of phones and handsets do but this is an access point by the way all six of these access points they're all sold into the exact same market and if you pick one and it's the wrong one you might have more aps or less aps the one on the bottom right we put 10 db in and we get 18 almost 19 out it has a positive 9 db and then you go well how can they do that well if you look at the antenna you can see the instead of wasting energy going up and down it's only focusing in a certain direction and you get a better gain yes this means you need fewer aps this is just one example i can show you tons of examples when you go on site put it up in a football field walk it and you can see there is a huge difference between the aps the previous one i showed you isn't always about rssi there's other things that aps have spatial streams better mcs better firmware that allow you to change design for your clients you have but also choose your ap before you do any other design eight is meet all your requirements not just primary coverage not just secondary or coachella interference there is an entire list of things you need not just the list i show you there is clients need to see a certain throughput the rom times have to be down but the number one cause i've seen in my career for remediation calls yep co-channel interference coaching interference causes capacity to drop you think adding another ap adds capacity most of the time it actually drops unless you can prove that it's not in the cochlear interference zone of anything else it's a good time to talk about cabling cabling i thought this was wi-fi yeah let's just do a little side jump over here to cabling for a minute if we look at the evolution of cabling silver satin cat 3 cat5 cat5e and then finally cat6 what did we do to make our cable go faster it's literally a thousand times faster the copper is the same we use the same copper we use the same rj45 head the distances are the same what we changed were things like uh pin outs twist ratios dielectric that went inside all those combined show what we were really changing was we changed the interference self interference a cable interfering with itself and by getting rid of self-interference we need to go a thousand times faster what do you think we need to do to make wifi go faster lower interference what's the number one cause of interference for wi-fi other wi-fi back to the coachella friends design around it and you can make it go fast as well understand how capacity really works you can't just go draw a little box on the screen and say there's 27 clients in here and here they go yeah all of that software whether it be andrew vanage's capacity planner or any capacity planner built in any piece of software they all make one huge fatal flaw they make too many assumptions they assume that an ap has a certain amount of capacity aps don't have capacity the rf has a capacity so if i have two aps on the same frequency they're sharing the capacity and yet all of the capacity planners mess that up so understand what's really there realize what each of those assumptions is you can still work on capacity but it is not nearly as easy as you think it is it's not just a quick calculation we have to understand the design and whether or not there's code channel interference and we have to minimize that first before we can ever get to bat capacity yeah there's a cya document everything document for yourself documented for the future i really like documenting all the requirements so when they call back three years from now and they say this doesn't work and you go really it doesn't match the requirements we designed it for well yeah but we got new stuff oh so you want to change the requirements and i love handing over to my customers when they ask for things like uh we want you to put them in hallways we want them below the ceiling we want we want and we want to whatever they ask for and if i know it's going to make a bad design i just haven't assigned a document yes keith please design me a poorly performing network i've actually had people sign it usually they get a little gun shy and say well what do we really need to do to fix it if you need more of these and more of the tips or any of the graphics come out to wmpros.com we'll see in the community all right keith well that was a great presentation and um well virtual conference right so i know where i am i'm in north carolina but where do you happen to be and how are you doing today well i know where i am too i'm in elkridge utah about oh an hour or so south of salt lake and i heard something about elevation the other day on twitter i think with you what is your elevation there where you live uh 5300 or something so yeah people were talking about mountains and i'm like what yeah oh yeah my girl was going over a pass in washington and it's half the elevation where we're at so yeah i grew up in west virginia we like to talk about our mountains and that's why i i put in that twitter thread i said well you know west virginia mountains are beautiful not big that's funny and they are beautiful yeah and i wouldn't want to walk over them so no definitely not although there are some nice hiking trails well um we had a couple questions come in during your presentation and you were really right on top of it as usual and answered a couple of those already in there so if people want to see the the quick text answer you can see that in there but i want to give you the chance to address it um here live with everybody as well first of all uh people were asking about the results of those tests that you did in relation to the football field and wondering if they'll be available can you let them know what's going to be available when and where they can go to find it yeah so there was there was a bunch of different tests we did some were comparing different techniques of surveying stop and go continuous gps on a segway walking those kind of tests so we have that's one group we had throughput tests uh on different clients and against different access points so we'll be showing those as well um we tested spatial streams two four and eight and there's a test there we test wi-fi five versus one by six uh throughput to a cellular versus super pro wi-fi we even did a can you have one iphone talk to another iphone 100 meters away so each of those different tests will have a little uh video and we'll be releasing those on youtube channel and the wm pro's website throughout october okay and what do they search for on youtube to find that channel uh wireless land professionals awesome excellent uh the other question that came in were thoughts on mgig and 6e and do we need it what do you think well one like a lot of new technologies mgig is fantastic i love it i i don't people like you don't like m gig no what i wanted to show was in wi-fi most of the air time is overhead and only the payload is the part that converts from wireless to wired and goes out the ethernet port so just because we might have three four up 10 11 gigs some of the things that chuck talked about a little earlier today on the wi-fi side doesn't mean we're going to get that same amount on the wired side so love mgig it's a fantastic technology i think everyone should be headed that direction and if i was going to refresh my switches in fact i refreshed the switch just a couple weeks ago and i went to mgig of course i wouldn't not put in m-gig just don't think that because wi-fi is faster you need m-gig there are two decisions so i would put in mgig especially when you go to wi-fi six and then when you go to wi-fi six e uh even more chance that you might run into that but we still have a lot of time to go just don't rush out and jump on your uh you know update your whole infrastructure for the reason that i'm going to be putting in ap sometime in the future exactly all right awesome thank you very much for being with us anything else you want to share with the audience before we go no just glad to be here uh lots more sessions to go we just barely got started so hang in absolutely thanks for helping us out right up front keith appreciate you being here your professionalism always participating and adding value to the community so thanks a lot thanks [Music] you

Channel: CWNPTV

Views: 1,883

Keywords: cwnp, wifi, 802.11, wireless, wlan

Id: Ou21RfvdQZk

Length: 40min 30sec (2430 seconds)

Published: Tue Oct 06 2020