Juniper Mist Wired Assurance with Sunalini Sankhavaram

0:09 um so distinguished Engineers as you can

0:11 see in the office the wireless C our

0:13 wireless CTO is Matt mcferson so I'm

0:16 bling between you and your lunch so I'll

0:18 just take a few minutes but we thought

0:20 you know it could be interesting for you

0:21 to know about us and and what we do um

0:24 so in this office of the wiress CTO we

0:26 work you know with the engineering team

0:28 we work with the product management team

0:30 but also what you may not know that we

0:32 also work with the R&D Departments of of

0:34 the last the large players out there and

0:37 you know that's what we call the device

0:38 ecosystem partners and that's you know

0:40 your key client and device ecosystem

0:43 vendors you the chipset vendors the

0:44 solution vendors um and what we do is

0:47 that we look at this Horizon you know

0:48 five to eight years we try to understand

0:50 what the problems will be in five to

0:52 eight years um and we try to find

0:54 solutions for for those problems so

0:56 somebody was joking about Wi-Fi 8 um so

0:59 that's what we're talking about you know

1:00 I've been talking about that all week I

1:01 don't have the access point yet but I

1:04 have you know about 600 pages of notes

1:06 so I'm sure if I concatenate this this

1:08 well enough will have an access point

1:10 out of it very good um so what we do

1:12 that we work in three different aess um

1:15 standards of course that is to say we we

1:17 design new Solutions um you know either

1:20 by ourselves or again with this device

1:22 ecosystem partners and we bring them to

1:24 these standard bodies um so of course we

1:26 are fairly active there um we have um

1:29 leadership position Triple E in the 11

1:31 bhbi Cox JTC the working group itself in

1:35 the wifi lines WBA and you know outside

1:38 of purely 11 we also look at you know BL

1:41 orb we are active in F which is the

1:44 Wi-Fi L equivalent for for White Band um

1:48 so we had um we contribute there I

1:50 counted before this meeting and we had

1:53 about 200 contributions just this year

1:55 in 2011 um so that's you know in the

1:58 standard but as you know in a 11 there's

2:00 a standard that describes the protocols

2:02 what the packets should be looking like

2:03 what the exchanges should be looking

2:05 like um but it don't tell you when it's

2:07 clever to use this Frame and when it's

2:09 dumb to use that frame so we we work

2:11 also with theem Partners to try to find

2:15 those you know Mis understanding that

2:16 can happen between the access point and

2:18 the client side engineering teams and we

2:20 try to fill these gaps um I was counting

2:23 against just this year we found about

2:26 147 gaps in the standard Wi-Fi you know

2:29 11be uh just for the implementation of

2:31 the different standard that we could

2:32 again um you know fill because we work

2:34 with these device ecosystem Partners so

2:36 you know that's the standard part we

2:38 also do what we call the standards plus

2:41 uh part which is where we work with the

2:44 device ecosystem Partners to go beyond

2:46 what is the standard today to try to you

2:49 know think of problems which may be a

2:50 bit specific a bit vertical specific um

2:53 and we try to develop Solutions together

2:55 for these problems uh sometimes we know

2:57 we do Prof of Concepts sometimes we

3:00 Implement those uh in in our products um

3:02 and then we proof case you know the

3:04 solution and we bring that back to the

3:06 next generation of the standards um so

3:08 many of us you know you know us well so

3:10 you know some of these names for example

3:12 fast lane you know you've heard about

3:13 that that's something we did a few years

3:15 back that became what we call in the W

3:18 Finance the oce Qs management R2 program

3:21 you also heard about the Spain plus uh a

3:24 program that we did for 11x case well

3:27 that was inserted then into 11be um as

3:30 the SCS QC Qs characteristic feature

3:34 that's now also in wifi 7 you know same

3:37 goes you know for open rooming you know

3:38 that we we brought that we build that in

3:40 house we gave it away to the WBA and

3:43 then we put in a 11 in the wifi Lin pass

3:46 Point program all the bricks you know

3:47 needed to make that work you know for

3:48 the entire industry so you know we we do

3:50 that a lot um and that allows us you

3:52 know to prove uh technology before we

3:55 bring it into the standards and then of

3:57 course you know in some cases as you see

3:59 the bottom some sometimes you know there

4:00 is no real need for industrywide

4:02 standards that's um what we call the

4:06 part specific standards where the

4:08 challenges are specific to a vertical or

4:11 to a type of device so much so that we

4:13 want to find solutions that work for

4:15 that type of device but that do not

4:16 necessarily require um a full blow Neo

4:19 integration into the standard for

4:21 example you know you have your Barcode

4:23 Scanner that thing can go to sleep and

4:25 then wake up and then it may lose its

4:27 connection quite often because it sleep

4:29 for well uh so there may need to be a

4:31 solution where we work with a a special

4:33 messaging to these devices so that they

4:34 maintain the connection you know for

4:36 hours or for days you know that that

4:37 kind of things so what we do is that we

4:39 design all these things we design we

4:41 bring into standards we do the proof of

4:42 Concepts we proof case the solution um

4:45 and then we bring the solution back to

4:46 the or the PM team and that's where you

4:49 know they they talk with you um to

4:51 determine which of these possibilities

4:53 have the most um uh benefit for the

4:55 industry and for our customers and then

4:57 that's how we integrate these uh these

4:59 features into our product all right

5:01 that's what we do roughly um you know I

5:03 don't have time to to go to all these

5:05 the details of what we do but we thought

5:07 could be interesting to give you you

5:08 know two quick examples of where what

5:10 we're working on now so that you have a

5:12 glance of where we're going um one of

5:14 them is what you see here which is

5:15 roaming again roaming yeah roaming you

5:17 know sadly has not been entirely solved

5:20 um yet you know there are plenty of

5:22 cases where roaming within a domain you

5:24 know Wi-Fi to Wi-Fi doesn't work well or

5:27 between domains which is a bit newer uh

5:29 between LT and Wi-Fi doesn't work well

5:31 you know you move from LT to Wi-Fi your

5:34 device you know turns to the Wi-Fi uh

5:36 path and then as you move away because

5:39 your mobile uh you don't get the right

5:41 connection to the AP and then you drop

5:43 the connection to the Wi-Fi before you

5:44 could establish fully the data paths and

5:46 that doesn't give you you know a good

5:48 quality of experience and also for you

5:50 know service providers on the LTE who

5:52 would like to switch the data path to

5:54 Wi-Fi in high density environments like

5:56 inside buildings it doesn't give them

5:57 this carrier quality so we're working

6:00 with a a bunch of device ecosystem

6:02 Partners you and device vendors um to

6:05 try to work on exchanges uh to be able

6:08 to provide information to each other and

6:10 guide this roaming experience though you

6:12 heard of FTM FTM is one uh technology we

6:16 have a um possibility now to have

6:19 location information to and from that

6:21 device so what we do is that we we

6:23 Define what we call this Vector based

6:25 roaming where we can establish a

6:27 displacement Vector for this device

6:29 which is you know the position and then

6:31 the derivative of that the movements and

6:34 then the derivative of that which is the

6:35 acceleration so we get that data and we

6:39 understand how the device is moving

6:41 compared to the position of the access

6:42 phones and then that allows us to you

6:44 know provide recommendations to this

6:46 device so for example you see the car at

6:47 the bottom that thing is moving fast

6:49 past the building so it may get in range

6:52 of the building access point but the AP

6:54 is going to tell that device you know

6:55 there's no point for you to join this

6:56 wi-fi because you're you passing too

6:58 fast to be able to work to be worth

7:00 using it uh the lady you see on on the

7:02 left side you know she's moving toward

7:03 the building for now she's on LTE there

7:06 will be a point where we'll be able to

7:07 say you're getting to the point where

7:08 you could be using the Wi-Fi in the

7:10 building she'll you know the AP will

7:12 help her connect that is to say find the

7:15 right time in the in the packet exchange

7:17 where she can switch a data path so that

7:19 there is zero interruption in in her

7:21 communication and then as she moves away

7:23 from the building she can also get from

7:25 the AP a time where it's it's the best

7:27 to switch back to the LT and of course

7:29 if you are inside the building we can

7:31 detect if you're moving or not moving so

7:33 this doesn't seem like much but this is

7:35 something we want to solve where we want

7:37 to get rid of this you know five to

7:39 eight second glitch that you have when

7:40 you move between the LTE and the and the

7:42 Wi-Fi next slide please um so that's a

7:45 small example but you know everybody

7:47 today is talking about new exciting

7:49 stuff like LM so of course in we looking

7:52 five years in the future we know that in

7:54 Five Years everybody will have LMS

7:56 everywhere not you know really probably

7:57 llms but probably slms you know small

8:00 language models that will be hypert

8:02 trained for some devices so we'll know

8:04 you have that in your network

8:05 infrastructures we know you'll have that

8:07 also on your our CL device um so what

8:10 we're experimenting right now with these

8:12 guys is to see how we can use these llms

8:16 to you know build the next generation of

8:18 what you're you're working on with with

8:19 Tony and M which is troubleshooting

8:22 where we want to be able to have these

8:23 LMS talk to each other dynamically and

8:26 automatically in the background Network

8:28 so that you know instead of pulling all

8:30 the kpis from those devices to get all

8:33 the information you need will be able to

8:35 detect first know withl um at one point

8:38 there is a d deviation from normal

8:40 behavior and when that point happens the

8:43 LM in the infrastructure on or the

8:45 client will be able to query the other

8:47 side and say can you give me the kpis

8:49 you know that are needed to troot this

8:52 symptom and this device will be able to

8:54 say okay you need this this this this

8:55 and that and we be returning the kpis

8:57 you need to exactly see what the problem

8:59 is without having to filter and get all

9:01 the packets so you know that's um that's

9:03 a couple of examples of course you know

9:05 I I should say none of that is committed

9:07 right we're looking into experiments

9:09 there are a few dozens of programs we

9:11 have running right now um you know some

9:13 of them will conclude successfully that

9:14 they don't go anywhere um several of

9:16 them will you know become slowly a part

9:19 of the of the protocol um and you know

9:21 some others will will get into our

9:22 products into this five or or 10 year

9:25 program and of course you know there are

9:26 plenty of other things like you know

9:27 we're talking location um FTM is here

9:30 there's a new generation called 11 a and

9:32 another one called 11 BK um we're

9:34 starting to see Fusion of location

9:37 information between um then ble um and

9:41 and and Wi-Fi so we're working all these

9:43 access to try to understand how you know

9:45 your Wi-Fi experience will be looking

9:47 like in in five years so this is you

9:49 know super exciting and of course you

9:51 know you know where to find me so if you

9:52 want to know more or exchange about

9:54 ideas about where where we could be

9:55 going please feel free to reach out but

9:57 I'm going to stop there so that I'm not

9:59 your L um the vector stuff was that

10:02 standard or standard plus I standard

10:05 plus standard plus yes for now yeah does

10:08 it account well I guess it probably

10:10 doesn't let's say we're leaving the

10:12 office and going to the car and parking

10:15 lot and we're on our teams call or WebEx

10:17 call excuse me yep is it gonna account

10:20 for that Weir trans absolutely

10:23 yes yeah even if you spin around the

10:26 building we'll be able to detect that

10:27 trajectory you understand that in in

10:29 fact you're sticking to the access point

10:30 domain so yeah it accounts for that too

10:32 yeah yeah so solve

10:36 that work unit