Juniper Wi-Fi 7/6GHz and learnings, Dual 5 GHz Dual 6 GHz and RRM

0:00 hello everyone uh I'm Wes Pervvis and we're going to talk about

0:06 wireless can't start without uh talking about Wi-Fi 6E 6 GHz Wi-Fi 7 i wanted to

0:13 put up a a scorecard of kind of where you know what we see from our customers where we are um and so we are 2 and a

0:22 half 3 years into Wi-Fi 6E and of all of the uh Wi-Fi 6E capable

0:31 access points that we have shipped 50% of those APs have 6 GZ

0:36 enabled are those users who think they have 6 GZ

0:43 enabled but haven't enabled a WPA3 SSID no no this is those are radios that are shut off the Yeah the radios that Yeah I

0:51 looked at it the other way how many radios are enabled so these are 6 GHz radios that are enabled beaconing which

0:57 means they have WPA3 configured okay awesome thank you so there's a lot of room for improvement here although this

1:03 is even a year ago that number was 30% so there's there's been a significant uptick from a client adoption standpoint

1:12 we're in the 20 to 50% six gigahertz adoption depending on the environment so

1:17 if it's a BYOD environment like a university that number is uh usually

1:23 trends lower uh if it's a corporate environment and that uh with you know

1:28 corporate issued devices uh and that uh company has you know begun refreshing

1:33 with 6 GHz which many of them have starting several years ago uh we're seeing upwards of 50% or even higher in

1:39 some environments of of 6 gig clients from a Wi-Fi 7 perspective we're seeing

1:45 client adoption in the 5 to 10% range right now which is you know in you know first year of rollouts kind of what we

1:51 would expect that's what we saw with 6 gig So here's an example customer where uh

1:56 they deployed they have about 1,600 clients uh on the network about uh

2:02 34% so 530 at this point are connected on 6 gig and then uh 142 of those

2:09 clients are Wi-Fi 7 so just to you know this is where we are in the journey um

2:15 you know the six gig option is really uh is really picking up and then you know Wi-Fi 7 will take uh several years to to

2:22 reach a a critical mass that leads into all right how about

2:30 channel width right because channel width has a direct impact on the speeds that a user is able to achieve on the

2:35 network and so if we look at 5 GHz um our our default channel width you know

2:42 if you go deploy this network our default channel width on 5 gig is 40 MHz but the majority of our customers are

2:48 using 20 MHz on 5 gig which is great uh everybody or many people are changing the defaults which is always good to see

2:55 uh on 6 GHz the story is a little different uh worldwide we see uh

3:00 generally 80 MHz channel width uh that's our default uh in Europe the majority

3:07 channel width is 40 MHz because there's not as much spectrum i think over time as 6 gig

3:12 adoption increases these channel widths will shift left um so you know I think

3:19 you once we get appreciable adoption you'll kind of at least in the US with 1200

3:25 MHz if you do 20 mehz on 5 gig you'll you'll likely be able to do 40 mehz on

3:30 six gig in most circumstances uh and then in Europe you know probably

3:36 that means going down to 20 mehz uh in dense environments on 6 gig which some of our customers have already started

3:43 doing so then that brings up the question of how do clients actually handle 6 GHz like you know do they use

3:50 it do they connect uh and so what we see is actually a very strong affinity from

3:55 clients to 6 GHz they they generally prefer 6 GHz

4:00 um they're you know most of them are programmed to prefer wider channels but

4:05 even when 5 GHz and six gigs have the same channel bandwidth uh we actually

4:11 see very strong affinity so here is a a client that is roaming this particular client happened to roam 350 times over

4:18 350 times on this day and every single roam was on six was six to six now

4:25 that's not always the case you will see some some interband um but generally

4:30 speaking affinity is is pretty good and as an example here is a network that had

4:36 somewhere in the ne you know neighborhood of 26 27 26 or 27,000 clients in the past two

4:44 weeks so about uh 16,000 a little over 5 gig only right we we only ever saw them

4:51 connect on 5 GHz uh another 7,000 connected on five and

4:56 six GHz at some point in that two weeks so those are 6 gig capable clients but the most interesting stat here to me at

5:03 least is 3,000 over 3,000 clients connected on 6 gig only so in that twoe period we

5:10 only saw those clients on 6 gig they never uh they never connected on 5 gig

5:16 did you see did you see a preference for device type whether that's Android or uh Google

5:22 specifically Samsung iPhone did did you see a preference for one that that kind

5:28 of preferred preferred six only yeah so the the Apple devices from what

5:34 I see have the strongest affinity to six gig they you know you'll very you'll

5:41 seldom see them on on 5 gig in a in a like dualband SSID Um then Android is kind of next level

5:47 down and then Windows is usually kind of third um yeah sort of that's how I see um do

5:55 you have any stats on ML um not that I'm not in here yeah right

6:02 i'm just you know as you're talking about you know I wonder about lines that are coming on six and five six and two

6:08 four right right how do you how are you going to manifest how are you going to show roaming and Right yeah i'm not Yeah I don't have anything on MLO specific um

6:15 but yes you you know you have to show it all right and we will okay um okay so then but

6:23 kind of related the precursor to being able to do MLO is security right so

6:28 Wi-Fi 7 just like Wi-Fi 6E introduces some mandatory security features so Wi-Fi 6E

6:36 you had to do WPA3 right for any on 6 GHz you have to do WPA3 on a Wi-Fi 7

6:43 SSID that can be on any band there are a couple things that you have to do as well but one of those is GCMP 256 cipher

6:51 so you have to use GCMP 256 generally we're using CCMP 128 for most things today uh you have to turn on beacon

6:58 protection and you have on a WPA3 personal you have to turn on this SAPE

7:04 GDH or you know extended caps you'll also see it called Um and the way this

7:10 is manifesting and what you know actually what Wi-Fi alliance is recommending is generally speaking so

7:16 this is a you know beacon RSN element uh in the beacon you know I think most people are familiar with uh you know

7:23 with the AKMs the you know key management this is how the the AP the BSSID advertises hey this is this is my

7:30 security this is what I can do this I'm OWE I'm enterprise I'm I'm personal um but in the in the pair wise cipher list

7:37 uh You know typically that's always been CCM and you know at least for WPA3

7:44 but for Wi-Fi 7 you know generally the recommendation is to add GCMP 256 so you

7:51 will usually see both advertised on Wi-Fi 7 networks um it looks the same you know that you know whether it's you

7:57 know WPA3 personal enterprise OWE um you know you'll see multiple AKMs but now

8:03 you'll also see multiple uh you know ciphers as well so again every single

8:09 Wi-Fi 7 BSS has to advertise GCMP 256 has to advertise beacon protection and

8:15 then if it's has to advertise SE GDH so what does that mean for

8:22 roaming uh so uh now if you have a you know complete

8:28 Wi-Fi 7 network uh where you know you're going from Wi-Fi 7 AP to Wi-Fi 7 AP

8:36 security is the same on both sides AKMs are the same on both sides GCMP is there beacon protection's there very good

8:43 roaming so as you would you know as you'd expect you know roaming roaming is smooth

8:49 now I don't necessarily think that this will every customer deployment right

8:55 because many deployments will have mixed APs right there you know we don't yet

9:02 sell a Wi-Fi 7 outdoor AP um you know the in the real world we will have mixed

9:08 you know Wi-Fi 6 6E Wi-Fi 7 environments that's just the way it's going to be so

9:13 how does that look like well in let's say let's say you don't make any changes

9:19 right you you're you know you take your configuration as it is today with your Wi-Fi 6 or 6E APs and then you just add

9:26 Wi-Fi 7 APs with Wi-Fi 7 enabled it all works well with older

9:33 clients but the newer clients that are capable of GCMP 256 which there actually

9:39 quite a few of them out there now um you know because if if a client supports

9:44 WPA3 192-bit they likely already support GCMP 256 um and

9:51 so the problem is let's say I'm connected to a Wi-Fi 6 AP i'm connected at at CCMP 128 right i roam to the Wi-Fi

10:01 7 AP and as the client I say "Oh this can do GCMP 256 that's better

10:08 that's more secure than CCMP 128 i'm going to change to that." But the problem is it's not seamless you will

10:15 have a disruption there and and so what that means is really for

10:23 seamless roaming you have to actually enable GCMP 256 on your Wi-Fi 6 60 APs

10:31 for the most seamless experience and this doesn't matter security type because you have to turn on GCMP 256 for

10:36 every security type now this isn't all that different from you know what we went through with Wi-Fi 6E right you had

10:43 to turn on WPA3 you know you're probably not mixing just WPA2 and you know on your older stuff

10:49 WPA3 on newer stuff you know it's generally one shot so I don't think this is going to really be a big pill to

10:55 swallow um but just a it's it's a consideration to know about um that uh

11:01 you know you'll you'll have to turn on at least GCMP 256 on your older stuff um beacon protection doesn't really seem to

11:07 matter on the older stuff you it doesn't affect the roaming it It's really just the G GCMP 256 for the clients that are

11:14 capable of supporting it feels a lot like something I want Marvis to tell me and go fix yeah where the cloud will

11:20 just configure it yeah automatic miss this won't with MIS this won't really be a concern i just want to point out that

11:27 you know under the covers this is something that will that will happen and we've actually you know we've had you

11:35 know since Wi-Fi 6E with with WPA3 there's been a lot of discussions around you know client interoperability um and

11:41 we haven't seen interop issues you know granted we don't have massive exposure um to GCMP 256 but of the clients that

11:49 we've tested the environments that we've been in um turning on GCMP 256 seems to be a non-issue

11:56 do you only turn it on on the 6 GHz band nope you do it as well if you broadcast

12:01 you have to do it because Wi-Fi 7 you can turn on Wi-Fi 7 on any band yeah you're right so any any SSID that

12:08 broadcast Wi-Fi 7 that supports Wi-Fi 7 has to have GCMP 256 doesn't matter the

12:13 band okay but but you are saying that the clients that you've tested do support the GCMP 256 yeah the kind of

12:22 the newer clients um Yeah absolutely newer clients passion

12:28 basically any client that supports WPA3 192bit um which has been out you know for three or four years in mainstream

12:36 are these going to be the default values if I spin up a new new SS ID it's best interoperability correct and when will

12:43 these be the new values even if I don't have Wi-Fi 7 APs on my network um if you

12:49 there will be a we will tie it to a um value like the there will be a Wi-Fi

12:55 configuration option on the SSID if Wi-Fi 7 is enabled we will uh enable all

13:01 this other stuff okay now is there a standard name naming convention you're going to use or is has that been defined

13:09 by a group like like how it's going to look like in our UI wi-fi 7 enabled

13:17 disabled it's just like it's already there actually um it you know just like we had the Wi-Fi 6 enabled disabled box

13:24 it'll there's a Wi-Fi 7 enabled disabled per

13:31 WLAN all right that's roaming um so how do you actually migrate to Wi-Fi 60

13:37 Wi-Fi 7 so as I mentioned only 50% of our APs 6 gig APs have 6 GHz turned on

13:45 so there's a there's still a tremendous opportunity and you know when I talk to

13:50 customers like there's still apprehension out there I think mostly about WPA3

13:57 um you know there's some power um you know apprehension as well but mostly

14:03 it's WPA3 so like this is my you know high level blueprint of you know how can

14:09 you actually get to 6 GHz and now you know how can you also actually get to Wi-Fi

14:15 Well first and foremost try WPA3 you don't have to have 6 gig capable APs you don't have to have

14:21 Wi-Fi 7 APs even to get your feet wet with WPA3 and it you know it's it's

14:28 something that you should be doing today because you will have to go to WPA3 at some point now you don't have to do it

14:34 on every SSID and you can be strategic with it a lot of times what folks will do is on their1 access.1 accesss ID they

14:41 will they will turn on WPA3 for you know because that's been the lowest risk now

14:49 of course several years into into this right and there's been a lot of discussion on interop especially with

14:55 WPA3 personal largely not a concern anymore you know in most of our networks 99

15:02 99.9% of clients are you know supporting WPA3 uh personal um you know they're

15:08 still in the back of your mind hey am I going to have an interrupt issue but you know people that I've turned on just move on right if you you might

15:15 you know if you have a couple clients okay go deal with those uh and same for OWE uh we've had many customers now who

15:22 have turned on OWE or OWE transition for uh for guest networks um in particular

15:29 and uh it's been pretty smooth uh you know to the point where it's still on

15:34 right what what's your take on the IoT impact for WPA3 adoption

15:41 um I think so for IoT devices I think in many cases you're not turning on W

15:48 you're not turning on Wi-Fi 7 on that SSID um and the ones that you know the

15:54 ones that you are it's still like hey do they support WPA3 if they're not supporting WPA3 doesn't matter if it's

16:00 Wi-Fi 7 or not you know they're going to have they're going to struggle so you're thinking for the foreseeable future then

16:05 it's still advocating separate SSIDs for IoT largely because of WPA3 or lack

16:11 thereof support yeah and and it really comes down to device by device like you know there are many older devices that

16:18 certainly don't support WPA3 but I think we've even seen some you know recent devices IoT devices that aren't

16:25 supporting WPA3 uh and so it comes down to what's in your environment test it out see if it works largely you know

16:31 from our side we think it's not really you know a from working with you know a

16:36 lot of our customers not really a concern but you know give it a try see what happens uh next is design uh and

16:46 so what we have seen uh you know with the transmit power limitations of uh 6

16:53 GHz especially the asymmetry between AP and client uh it is most advantageous to

16:58 have a capacity based line uh where you know clients can you know have you know strong signal wherever they you know

17:05 wherever they are they don't have to be on the fringe of the cell um and where the the client TX power limitation would

17:12 be you know the seen the most um and then lastly is you know think about your

17:19 network what do you have from a switching perspective for PoE for for M gig um you know most of the Wi-Fi 7 and

17:25 6APs many of them need BT power for full functionality most of them have a AT

17:32 operating mode for reduced functionality we have many customers who use the AT operating mode no concerns there um you

17:38 know but think about the future if if BT is is appropriate as well as M gig um also not a requirement but a nice to

17:45 have now with three data radios wider channels um there are you know you can conceivably uh and MLO I should say you

17:53 can conceivably uh achieve more than a gigabit per second uh in real world conditions on

17:59 APs so this is the you know step one step two step three of you know how do you how do you actually get to Wi-Fi 7

18:05 how do you actually get to 6 GHz and so let's say you've completed this what you

18:11 know what happens to your network you know are there actually tangible benefits to 6 GHz um my answer is yes

18:19 and so it turns out we actually have a a actually a very good benchmark year-over-year so the same event a

18:26 one-day event held year-over-year uh so you know 2023 2024 2025 in 2023 the

18:33 first year that we did it the 6 GHz adoption was in the 10%

18:40 range and the the the graph here is the traffic uh by band so you know the the

18:47 six gigahertz traffic was you know very low you know I think even less than 10%

18:52 of the total traffic in 2024 the total amount of data transferred you know increased from 3

19:00 you know 3.8 terabytes to 6.8 8 terabytes and the 6 gig adoption was in

19:06 the 20 to 25% range and then uh you know overall traffic volume was you know

19:13 about a third on 6 gig this year traffic increased again to 11 terabytes right

19:22 more 6 GHz more capacity right traffic increases but most importantly the 6 gig

19:30 traffic volume is you know probably 40 45% where

19:36 and then 5 gig is you know 55% so the 6 gig you know contribution to

19:41 the network is is increasing you know maybe next year it will exceed it and this is this is a European event where

19:47 there's you know 480 mega spectrum I think you know results would actually look different in the US but it kind of

19:53 brings up the point of you know the way that we think about 6 GHz at least in Europe is it's really an extension of 5

20:00 gig you you've doubled your available channels uh and you know you can see

20:05 traffic profile matches so this is You know when I pulled this data it was very

20:10 encouraging to see wes for this event were you seeing airtime contention originally or yes so you've also

20:18 alleviated the airtime contention by using six gig as well right yeah that's one of the major benefits yes absolutely

20:25 wes it's great to see the overall RF traffic stats uh is there any

20:30 granularity as to you know what what is constituting you know the the up uh

20:36 swing in traffic like on a per application basis or there's there's um well okay not on it's this it's a you

20:43 know it's an event so it's mostly um you know email people posting pictures and

20:49 uh and whatever but um the you know there's more headroom for those people to do that stuff which is

20:55 why you know we're seeing the increase in traffic year-over-year here be interesting to see if how much

21:01 of it could be related to AI naturally i don't know um and and in the last two years

21:09 the six gig traffic volume increased 12 times do you have similar data for an event

21:15 that would be held like in the US or Canada where we could potentially have a lot more traffic not to this like this

21:23 is a it's a repeatable like it's the same thing every year you can compare yeah so it makes it very easy to compare

21:29 we do have US data but not to the same like Okay and then we're using 20 MHz Y channels on both bands um first two

21:37 years were 40s and then this year was 20s okay uh sorry uh 40s on six 20s on

21:43 five okay and then this year is 20s on both bands both bands okay

21:49 here's another example uh this is a large university customer who uh is

21:54 actually deploying about a thousand APs a month in session it's just incredible

21:59 uh but they have 60 70,000 clients on the network uh around the end of the

22:05 semester last year after Thanksgiving um uh they decided to enable WPA3 right

22:13 so they got their feet wet with WPA3 um and then a few weeks later at the end of the semester over winter break they

22:19 actually turned on six gig so when uh when the students came back after winter break this is the this graph is the

22:25 capacity SLE uh and so you know before 6 gig you know they're in the low to mid

22:31 uh 90% capacity which you know is pretty good for you know their size but by turning on 6 gig right away 25% of

22:38 clients connected and their capacity solely improved about 5 percentage points and at their scale that's a

22:46 significant uh significant improvement so there's tangible benefits to six gig

22:53 i'm a you know huge proponent of it you know the the roaming piece I think is sorted uh security largely has been okay

22:59 with especially WPA3 enterprise um and so yeah I want that next year i want that 50% number to be 70% and then keep

23:06 going up i would be remiss if I didn't talk about

23:12 2 x2 and 4x4 APS because uh you know this is what the industry has settled on

23:19 um and often you know the question comes up of why do you even have a 4x4 AP when

23:25 you have 2x two you know is 2x two all I need and so just as a you know refresher

23:31 this is uh you know 2x two transmit chains on the AP two receive chains um

23:36 and so when you look at you know the antenna polar plots often you'll see a line for each uh element each

23:43 transceiver and then you'll see an outer line that's the composite um that composite line is you know the um you

23:50 know usually would be with myo gain and depending on the vendor you know usually this is as measured hopefully it's as

23:56 measured some some people simulate but um these are from mistaps they're as measured uh and then 4x4 you have you

24:03 know four lines one for each radio chain and you know the composite line as well so the theory is you have um you

24:13 know higher you know more myo gain in 4x4 and you know better receive for MRC

24:19 um this is a MacBook client listening to the AP and so just by changing um the AP

24:27 from 4x4 to 2x2 in software right same AP same client location just by changing

24:32 config from 4x4 to 2x2 uh the signal as heard by the a as heard by the

24:39 client changed from neg 52 and a drop to neg 57 so the math would be you know usually

24:46 you would expect about a 3db change um going from 4x4 to 2 x2 in this case it

24:51 was five you know we live in an imperfect world it is what it is but you know this is this is just listening to

24:57 beacons right um and our AP is actually beacon across all chains which is why we see a difference here but you know this

25:05 is the initial piece of it okay signal you know we have a signal difference from you know we can have a more robust

25:10 signal with 4x4 how does that actually impact your data rates so if you look we looked

25:17 across um the missed universe we pulled uh you know data rate versus uh signal

25:22 strength on uh x-axis and the you know just looking at the trend right the uh

25:29 the trend goes up and to the right which means uh the data rates went up and

25:35 signal strength also went up for uh for a 4x4 AP same thing on receive side

25:42 where the trend is also up and to the right so there's a the biggest difference you know multiple multiple dB

25:49 difference in transmit and then you know slightly smaller you know one or two dB difference in receive um but uh you know

25:56 there's that has a actually a noticeable impact on the data rates again more robust link uh between AP and client so

26:04 does that actually relate to retries well it lowers the

26:09 retries which is good and then how does that actually relate to throughput throughput goes up um so you I'm not

26:17 getting you know super into the weeds here it's just meant to be a high level of um yeah there is actually a benefit

26:24 you have a more robust RF link and that does lead to speed improvements um you

26:29 know user experience performance on the network with 4x4 APs now does that mean 2x2 you know is useless no right it's

26:37 it's all relative like it's it's what you need on the network and certainly for the most demanding of of environments we would recommend 4x4

26:45 uh with this data would there be diminishing returns if you had a say 8 by8 or a 16x6 that is technically

26:52 feasible in the standard yeah you know you would see improvements but you know it gets to you know where does it start

26:59 start and stop mattering you know I I think with 4x4 we're kind of at that sweet spot of power price and

27:05 performance okay next up is uh the use of

27:11 directional antennas for everyday use uh and and so this is something

27:18 we've been exploring um we released uh the uh AP

27:25 47D this is a AP 47D um with directional antennas um you know built-in

27:31 directional antennas to the AP um and so the the reasons you know for

27:38 this AP are several and so we've seen a uh

27:43 increasing desire from our customers uh to have higher AP density in everyday

27:50 environments this is your offices you know maybe auditoriums in our retail customers as well where they need higher

27:56 AP density but if you're deploying omni APs even with you know our APs have 30°ree down tilt uh even with the down

28:04 tilt at some point you get to a point where you have too many APs your roaming sucks uh code channel problems it you

28:13 know you you really have to tune at that point when you you know when you've deployed too many APs for your

28:18 environment and so the solution to that is okay let's go deploy some directional antennas and typically we're doing that

28:24 with AP45E which works you know very well but feedback from customers is it's

28:30 kind of a pain you know we got to plug in the AP you know the connectors to the uh to the AP got to mount it it's harder

28:36 to make it look nice it's more costly to install and you know got to pay for the antennas um and so you know there's we

28:44 we've actually had like a a bunch of enterprise customers carpenter enterprise deploy 45

28:52 um in you know in their office environments right because they're they have such density such utilization in

28:57 the office you know that they need you know the directionality um that's why

29:03 there's a 47D for those types of environments um you know we think it's

29:08 going to be useful in those carpet enterprise in auditoriums it's going to be useful in uh even some of our retail

29:14 customers um you know to help support 6 gig uh density uh in those

29:20 environments this AP uh is specifically a 60x 60 uh square pattern um it you

29:26 know and so that gives the advantage of it you know it reduces go channel contention it reduces you know that the

29:31 AP can't hear as well so it doesn't there's other crap out there doesn't care as much about it and uh and clients

29:38 also don't hear as many APS which also turns out to be a big problem in in these really dense environments

29:46 any plans in uh having like a 40x 40 a 30 by 30 is that even doable like with

29:51 the physics no comment

29:58 uh you know this isn't a uh you know oneizefits-all

30:03 uh you know there's still absolutely a use for external antennas you know we but we released a AP47E with external

30:09 antennas uh you know especially in the more challenging environments like you know warehouse particularly we use a long narrow pattern uh style antenna or

30:16 high density where we don't have the 30x30 today and uh you know other other niche

30:22 environments um so as far you know as far as some testing here is a auditorium this is a

30:30 lecture hall at a university it already had 45 Z 45E antennas 45E APS with directional

30:39 antennas uh in this particular beam there's three APs we just took down the AP45 in the middle put up the

30:46 47D so from a a results perspective and you can't really make it out on the uh

30:52 on the projector but uh on the stream you should be able to see the color difference so the in the middle those

30:58 two APs in the middle are the 47D 5 gig and 6 gig and then on the left and right

31:04 on the top row uh are the existing 45e uh and so the the coverage pattern

31:10 between the 45E which is using a um you know on paper a like a 45x 45 antenna I

31:18 think um is uh you know the pattern is actually very

31:25 similar which is uh which was encouraging to see this particular customer was very happy they're going to

31:30 actually go forward you know all their autom going forward will be 47D it gives them the added benefit of they can

31:37 support low power indoor on the AP they don't have to worry about standard power um and then for uh for comparison the

31:43 the two bottom APs are Omni 45s that just you know permeate everywhere in the

31:49 room uh so there's you know absolutely benefits for the directionality um you know for this particular environment uh

31:55 so I just wanted to show some some real values real testing so why the 2x two for the 47D when it's expected to be

32:02 used in higher dense higher density spaces um they didn't have full uh full power

32:08 we didn't fully power it okay so the you know for this testing we didn't really

32:14 care about it um but yeah just to note the Oh so the the product is 4x4 correct

32:21 okay it just wasn't in this test correct and you know so you can you can account for that you know the the signal

32:28 strength doesn't match up exactly you know with the uh with the 45D you know mostly we we actually see about a 3db

32:35 difference which is what you expect going from 2x2 to

32:40 4x4 okay and then next piece um is dual 5 GHz and dual 6 GHz uh AP47

32:48 uh is capable of dual 6 GHz um as well as dual 5 on our previous APs we've

32:53 supported dual 5 as well and I want to get into a little bit of the uh nuance

32:59 that we've learned uh with dual 5 over the years so first is you know most

33:05 vendors out there and this is how we you know do our dual 5 have settled on this this sign of you know this kind of

33:12 architecture where you have one radio that uses the lower portion of the spectrum one radio that uses the higher

33:18 portion of the spectrum and you put a band pass filter in the middle so you you know you can achieve incredible interference rejection with you know

33:24 with the collo colllocated radios uh same thing on 6 gig you know we're using uni5 low band uni 6 uni7 and uni 8 as

33:33 high and then you know uni5 uni 6 as a guard band which means for Europe uh and

33:39 UK and other countries with 500 MHz spectrum dual six not useful right now

33:45 for you um however actually today 40 minutes ago the UK had

33:52 consultation uh for opening up the entire uh 6 gig band for um

34:00 uh you know for indoor use now this is consultation they're taking feedback um it just closed but that would be

34:06 actually very encouraging um for you know opening up the additional spectrum and you know they also want to turn on

34:13 you know AFC for outdoor use cases that would probably be longer term but if you

34:19 read actually phase one ideally by the end of 2025 so that's very fast in uh in

34:27 spectrum timelines regulatory timelines so we'll see you know actually comes back in the consultation but this is

34:33 what they want to do at least and would help with the you know dual six uh in uh

34:38 in UK and is said EU is still in planning do you do you know what they're Yeah so EU had a um I forget what the

34:45 MRD or they had they had their radio group cons you know consultation two

34:51 years ago the next one is in 2028 or something oh good lord um and so they

34:57 said we want to do it but we'll decide by the next one

35:02 thank you yep um so then what does that you know actually mean in terms of channelization

35:09 well on dual six sorry on dual 5 you have eight channels on lower and you

35:16 know depending on you know 16 17 on higher AP47 we support uni4 so there's

35:21 actually 20 channels on the upper that channel mismatch right is turns out to

35:28 be a problem it means that you can't just turn on dual 5 everywhere because then you've reduced the number of

35:34 available channels that you have right you run into channel reuse issues especially if you do 40 MHz you have

35:39 four APs before reuse issues on six gig not so much of an issue uh because the

35:45 it's more equitable there's a better balance now you know to just really drive home

35:52 the 40 MHz point uh you can you can see in here there's a lot of reuse so this

35:59 is um this is like an example of what not to do but it kind of drives my next

36:04 conversation which is um you know let's let's think about what are we actually trying to do we're trying to increase

36:11 the capacity on the network by having you know more available radios on on 5

36:16 gig or 6 gig so it turns out not all capacity issues are the same kind of comes down

36:21 to hey maybe I need to add some more APs like my clients are you you know using there's no other capacity there's no

36:27 headroom I need to add more APs or you know conversely it could be I have too many APs I'm I'm interfering with myself

36:34 or a neighbor's interfering with me and and so that leads us to how do we actually like let's actually bake that

36:42 into RM from you know we've had this auto conversion feature for many years um it's worked incredibly well we use it

36:48 all over the place where we can convert automatically the 2.4 radio into 5 gig into 6 gig now in

36:55 AP47 but what we have seen is in many instances the radios that we're

37:01 converting aren't actually the highest utilized APs that could benefit the most

37:07 from that additional radio in 5 gig or 6 gig that's what we want to solve so

37:12 today we have a density based approach where we're looking at the density of APs do I have enough density to support

37:20 um you know this auto conversion or auto cancellation what we are moving to is a

37:26 capacity based approach you know sounds simple but there's actually a like a lot

37:31 that goes into it um and so we we change our questioning from do I have enough density to can I do it do I need it

37:41 where do I need it right so can I do it is like are there enough channels enabled in my configuration uh how many

37:49 other APs do I hear like you know are there available channels from a capacity standpoint for me to do it um like do I

37:56 need it meaning do I actually have the util utilization characteristics to take advantage of of you know additional 5

38:03 gig radio additional 6 gig radio and then where do I need it right where where on the network do I need it so I I

38:09 on a previous MFP I've shown this slide which is we want to learn the behavior

38:14 of the site you know what is the 5 gig utilization 6 gig utilization what is the roaming how much dwell time you know

38:20 all these you know things and so this we are we are learning we are doing and that feeds into this new uh you know

38:28 reimagination of the auto conversion part of RM so this uh this plot uh on the left hand side is

38:39 um it's a uh you know it's an environment the blue APs are APS that

38:44 we're happy with there's no like they're happy as they are right they have plenty of headroom the red APS are APs that are

38:51 the high utilized APs that are having capacity headroom issues these are candidates for auto

38:58 conversion on the right hand side is actually what the algorithm did right so the first step is we have to identify

39:04 these high USAPs second step is let's go you know let's actually go convert so you know many environments especially in

39:10 retail you may not have all the available channels they may you turn off some of the DFS channels so this is with

39:16 15 available channels in their you know in this configuration um the black APs turned out to be the APs that we

39:22 actually converted um and so this is you know

39:27 okay so what does that actually mean well uh here is here's the outcome looking at um looking at one of these

39:34 APs the blue lines before and then the red and the green is after on both radios so blue before if the lefth hand

39:42 side is the number of clients versus the number of samples so there's you know you know mostly 15 to 25 clients on the

39:49 AP uh after the number you know we had actually good distribution between the

39:54 radios in terms of number of clients that decreased utilization is on the graph on the right uh utilization

40:01 decreased significantly after was it going to dual 5 or dual 6 uh dual 5 this

40:08 is five um this is a AP43 customer okay and

40:15 uh and and then the actual throughput in the AP went up as well so that's the

40:20 entire goal of this exercise is how do we identify the APs where we have headroom issues and then let's go figure

40:27 out if we can alleviate that and and going forward how are you going to make

40:32 a decision between where you go to a dual 5 or dual six so if you go yes we need more capacity are you going to look

40:39 at like how many clients on each bend or is this this is why we have to learn the

40:44 behavior of the site we can make those decisions we can make a bandwidth decision as well like you know do we do

40:50 we change the bandwidth do we change the band do we go five do we go six do we mix it yes that's why we need this this

40:56 is the foundation of understanding the behavior in the site then we can do all this other stuff and so this is going to

41:01 be sort of a a a learning thing that occurs after your miss network is up for

41:06 6 weeks you then have the option of doing the thing because we've learned enough about your network or is this

41:12 going to be a model that applies to all networks uh this will be a learning based on your network okay um and so

41:17 when you deploy you know we will do some stuff right away and then over time it'll get better so you know it's it's

41:25 not like you're sitting there waiting um but over time we will you know even today we you know if you stand up missed

41:31 APS you know the second day third day the RM even gets better because we have that reinforcement learning loop okay

41:38 and this would be part of that obviously okay absolutely I had a question if you make those

41:45 changes to the black ones how do you know it was a good change part of the reinforcement loop so

41:51 we actually we looked so today we're already consuming our capacity SLE and coverage SLE and roaming SLE's so we

41:58 make a change i should have put the slide in i was thinking about it uh so we make the change we monitor you know

42:04 these are representation of of client experience so if capacity goes down that's a bad change right if cap

42:11 capacity goes up if we're actually trying to improve our headroom so if the headroom goes up then that's a good

42:16 change we stay today we don't today like in today's algorithm we only go one way

42:22 um we will go two ways in the new uh in the new world is there something

42:28 planning to get rid of the other red ones um the problem is

42:35 uh in this particular case uh they als like like how so how do we get to dual 5

42:41 how do we get to dual 6 we convert a 2.4 radio they have 2.4 utilization so we

42:46 have to be cognizant that there's 2.4 utilization and not cause coverage gaps um and the the other problem is in this

42:55 particular environment they only have 15 available channels so if I convert more radios to dual 5 I'm not actually

43:02 picking up any capacity gains but but could it also recommend you should go to six gig i mean in this scenario you need

43:09 different hardware correct correct yeah that would so that is absolutely part of

43:15 uh you know we have to we have to recognize when we've done all we can and there's simply no more capacity

43:22 will customers be able to blend both a density approach and a capacity

43:27 approach to you know the optimization of the site yeah so we the capacity

43:34 approach consumes the density approach where we are still considering the density of APs to make sure there's sufficient density uh you know to

43:41 support any 2.4 utilization all right so so the reinforcement learning algorithm

43:47 you mentioned it takes into account capacity roaming and what was it for you mentioned free uh coverage coverage

43:54 because has that changed recently because did it used to just be capacity it used to be capacity coverage and two

44:00 or three years ago we introduced a roaming SLE for the purpose of feeding into RM