Expect the UnEXpected with Juniper

0:09 so my name is sanjo i am a product

0:10 manager for the for the wired and the

0:12 wireless portfolio accompanied by abi

0:15 who is uh the lead product manager for

0:17 wired assurance and rajat who is the

0:20 proud product manager of the just

0:22 announced

0:23 ex4100

0:25 and uh if you guys are like you know

0:27 some of us you may be wondering why are

0:29 we talking about a switch

0:31 in the mobility field day and and sujay

0:33 said the context right this morning uh

0:35 when he actually said that like you know

0:38 obviously

0:39 other than the switches powering up the

0:41 aps uh there is a role that the switches

0:44 play and if you have looked at sunolini

0:45 and bob and g-shang's uh like you know

0:48 the demos

0:50 there is a lot of like you know user

0:52 experience that is dependent on things

0:54 outside of wireless that all of you are

0:56 aware we talk about mean time to

0:58 innocence for for wireless networks and

1:01 that is where the switch comes into play

1:04 especially like you know when marvis is

1:06 at play

1:08 the user experience does not have to pay

1:10 and we are there to make

1:12 your

1:13 user experience great with a full stack

1:16 offering and 4100 is a step in the right

1:19 direction but before i get too ahead of

1:21 myself i wanted to put jeff's marketing

1:23 dollars into good use

1:32 [Music]

1:42 [Applause]

1:43 [Music]

2:23 okay great

2:24 uh

2:25 so now let's get actually into the bread

2:27 and butter of what the 4100 brings to

2:29 the table uh and along with wired

2:32 assurance and and marvis right

2:35 so the first and foremost

2:37 this is basically a switch born in the

2:39 cloud era uh so if you look at like you

2:42 know the mist

2:43 how we do our short end services

2:46 we are not there just monitoring

2:48 switches right so when there is an issue

2:51 around congestion or loops or when there

2:54 is an issue around port flaps we want to

2:57 be able to catch that right and these

2:59 switches born in the cloud era native

3:01 cloud native

3:03 they allow

3:04 us the efficacy of data

3:06 that gisheng and his team can then mine

3:09 to then correlate user problems how much

3:12 of that is contributed by wired or by

3:15 something else in the network right tied

3:18 to that is flow visibility right now in

3:21 the 4100

3:22 just like the 4400 which we announced

3:24 last year uh we have flow visibility so

3:27 which means that

3:29 dos signature attacks like you know

3:31 egress and the ingress monitors uh real

3:34 time supported in hardware

3:37 also act as a vector in giving more data

3:40 to cloud with which we can do a lot more

3:42 correlation

3:44 uh talking about dos signatures another

3:46 thing that we see in rfps all the time

3:48 is is support or requirement to support

3:51 maxsec for inter-switch communication we

3:54 have max x support on the 4100 natively

3:57 for

3:58 more security

4:00 the next thing that is coming up more

4:02 and more in in the campus networks is

4:05 how do you actually do an l2 stretch and

4:08 within that how do you do micro

4:10 segmentation uh the

4:12 ex4100 supports

4:15 evpn vxlan fabrics and we will get a

4:18 little bit more uh detailed in in the

4:20 demo uh but for now like you know this

4:23 has support and hardware for vxlan and

4:26 along with that we also support gbp that

4:30 is a tag in the header for micro

4:31 segmentation especially in the context

4:34 of proliferating iot devices in the

4:36 campus

4:38 you saw on the video plug about enhanced

4:41 poe fast and perpetual poe so now your

4:44 business critical and mission critical

4:46 wired devices like hvac or

4:50 lighting if they are connected to the

4:52 switch when the switch fails

4:54 the the devices are still going to be

4:58 receiving poe and that's a very useful

5:00 feature and then last but not least

5:03 wes did the expose on on wi-fi 6c

5:07 some of the wi-fi 6 eaps you see on the

5:09 market uh do require

5:12 poe plus plus right 802.3 bt power i

5:15 mean like you know many of the devices

5:16 like ours are probably right on the cusp

5:19 of like you know 30 to 31 watts at full

5:22 stream right like you know all four

5:24 chains all peripherals like you know

5:26 ethernet cranking at five gig so if

5:29 everything combined we are at 30 watts

5:31 which like you know technically is just

5:32 a wee bit above your 380 standard 4100

5:36 like the 4400 is a poe plus capable uh

5:40 switch that we are introducing in our

5:41 portfolio

5:43 um

5:45 so uh just a double click and i don't

5:48 want to spend too much time on this

5:50 particular slide

5:52 it is a very versatile switch family

5:55 right

5:56 you basically see on the left

5:59 the high performance multi-big multi-gig

6:01 poe plus plus

6:03 form factors the different skews and on

6:06 the right you are seeing the entry level

6:08 for the value conscious deployments some

6:10 of the fixed form factors and for fixed

6:12 form factors basically what that means

6:15 is that like you know the psu is is

6:17 fixed it doesn't take a redundant psu

6:20 right

6:25 so where does the 46 4100 slot in the

6:28 whole lineup of a versatile juniper

6:31 distributed enterprise and the campus

6:33 portfolio of of switches uh if you look

6:36 at the lineup today you have the 2300

6:39 fixed form factor going into the modular

6:42 poe plus plus

6:44 offerings

6:45 the 4100 f that we talked about slots

6:47 right above the 2300 at very attractive

6:51 price points and then the 4100

6:54 uh the m gig and the poe plus plus

6:56 queues uh slot in between the 3400 and

7:00 the 4300 and if you look at

7:02 the next gen deployments for the

7:04 distributed enterprise or for campus

7:06 networks we believe the 4100 and the

7:09 4400 collectively in access for all

7:13 types of deployments whether we are

7:15 talking about just your standard

7:17 deployments or with evpn vxlan in the

7:21 campus uh with different types of

7:24 architecture like you know it gives you

7:26 a wide range of deployment options from

7:28 small to large in the in the access

7:31 layer esm are these uh required to be

7:33 connected to the cloud or can they be

7:35 operated

7:36 standalone with no cloud connectivity

7:38 right

7:38 well great question so are they uh so

7:41 they are cloud native

7:43 but you do have the option to deploy

7:46 them standalone obviously understanding

7:49 that when you're deploying them

7:50 standalone

7:52 marvelous is not at play and your user

7:54 experience may pay so um any other

7:58 question on on the portfolio and by the

8:00 way during the break or after the

8:02 session rajat has actually a couple of

8:04 these units lying there if you want to

8:06 touch and feel have more questions like

8:09 you know please feel free to do that

8:12 so just follow up for sam's thing a

8:14 question um would the

8:17 different code between the two or the

8:19 same code and i just it's exactly the

8:21 cloud okay right yeah and and you

8:23 basically just need a wired assurance

8:26 subscription uh to basically call home

8:28 to the cloud and like you know get it

8:30 adopted and you'll be off to the races

8:34 yes describe the difference between a

8:36 cloud native switch and then a switch

8:38 that i already have but can be managed

8:40 in the cloud correct so uh as i kind of

8:43 said that we are we are not exactly just

8:47 monitoring right uh we are basically

8:50 providing assurance for end users so

8:53 royal the distinction that i will make

8:56 is that for all of these switches is the

8:59 new switches especially with the uh with

9:02 the hardware support for some of the

9:03 things like float telemetry data

9:06 these switches

9:07 will have a higher data efficacy with

9:10 which we'll be able to granularly spot

9:12 problems better and faster right now

9:15 pretty much the whole lineup of switches

9:17 over here actually do are supported by

9:20 wired assurance just that the 4400 and

9:24 the 4100 the ones that we released in

9:27 the last couple of years just have that

9:29 like you know extra oomph on them for us

9:32 to be able to manage user user

9:34 experience better

9:36 anything you'd like to add to be yeah

9:38 the elementary is key uh the the rate at

9:40 which you get telemetry the hardware uh

9:42 supports flow based telemetry you can

9:44 get better visibility to what exactly

9:46 it's happening not only from a security

9:48 standpoint but also the rate at which uh

9:50 you can actually see uh things happen on

9:52 the screen so telemetry makes a huge

9:55 difference in our

9:56 omar was to make better decisions and

9:58 and that's the key difference

10:00 that's uh that's a beautiful segue to my

10:03 to my demo uh so thank you for that plug

10:05 sonalini but before i go to my demo any

10:08 other question

10:10 how does uh this extract telemetry

10:13 impacts the actual performance of the

10:15 switch like do you see any cap when you

10:18 start pushing a lot of traffic to the

10:20 switch do you see any limitation in that

10:22 so there's a clear separation of data

10:24 plane and control control okay your data

10:26 plane is untouched your control plane is

10:28 where you get management telemetry from

10:30 so

10:31 your the amount of traffic you push

10:32 through the switch has no uh impedance

10:35 on on

10:36 what how much telemetry you would get so

10:38 they're fully separate okay

10:41 okay

10:43 so um let's basically just go into the

10:45 demo so what we are going to show in the

10:47 demo obviously this is a mobility field

10:49 day and anything without wireless is

10:51 blasphemy so what we are going to show

10:53 is exactly what sunilini said from a

10:57 cloud managed perspective the onboarding

11:00 experience right how do you basically on

11:02 board a switch and in this particular

11:05 example if wes has talked about our

11:07 wi-fi 60 portfolio we show an ap 34 and

11:11 an ex 4100

11:13 both of them

11:15 adopting to our cloud inheriting the

11:17 configuration and then you're like you

11:20 know basically rolling right so day zero

11:22 is really about

11:24 uh your provisioning

11:27 before you get into all of your

11:29 deployment and operations and that is

11:31 what we show in the demo

11:36 uh so over here like the intent of the

11:38 slide is to just point out that it is an

11:39 ep34 along with an ex 4100 that is that

11:43 is the focus of the demo uh we stitched

11:46 this together so if you are familiar

11:48 with the missed onboarding process we

11:49 give an installer a missed ai app

11:53 every single device that we ship like

11:55 you know gets shipped with a claim code

11:57 and

11:58 along with the serial number and all you

12:00 have to do from the app is scan the qr

12:04 code

12:05 and then you can like you know name

12:07 the switch that is what we are showing

12:09 over here

12:11 and then you are going to assign the

12:13 switch

12:14 to a particular site right so that is as

12:17 simple as it gets the installer can do

12:20 it the the ins there are there are

12:22 specific installer roles that allow them

12:24 to use this app seamlessly so at this

12:27 point of time

12:28 when i have actually claimed the switch

12:30 and assigned it to a site it

12:32 automatically appears on our site but

12:35 the status is disconnected and that is

12:37 because the switch hasn't yet been

12:39 plugged in and connected to the

12:41 internet so now i'm going through the

12:43 process of configuring a port and we

12:46 support dynamic port profile or dynamic

12:48 port configuration in this particular

12:50 example we are using uh

12:54 the port profile for an access point so

12:56 anything on lldp that that has missed in

12:59 it uh which is on on our mr access

13:02 points we are basically going to

13:05 provision the port for that and when

13:06 that actually happens uh like you know

13:09 you are pretty much

13:11 almost ready to get both the switch and

13:14 the access points connected uh if you

13:17 can roll on the video so here what we

13:19 are basically showing you asked about

13:21 cloud native role uh so

13:24 over here you'll actually see that on

13:26 the switch there is an led for the cloud

13:29 which is blinking and this follows the

13:31 exact same format as our access points

13:34 right on our access points we basically

13:37 have blink codes so depending on the

13:39 state of the access point with respect

13:42 to connectivity link up ip address and

13:45 all the nine yards that's the exact same

13:47 sequence that these switches

13:50 also will support right so there are

13:52 blink codes so now we are going to get

13:54 ready at this point of time the switch

13:56 is actually

13:58 still disconnected not been plugged in

14:02 every time

14:04 go ahead okay

14:06 one one key element about uh just

14:08 addition about the blink codes right

14:10 uh our mistake fees if you've seen uh

14:13 they're cloud managed that would mean if

14:15 they're up in the ceiling but they're

14:16 not working for whatever reason

14:18 it still communicates to its neighboring

14:20 ap to tell you that the ap is down

14:23 because it did not get an ip or the

14:25 ethernet click is down or it's not able

14:27 to get to the dns without you having to

14:29 make a site visit

14:30 exact same constructs will be pushed to

14:32 the uh for the push the 4100 world as

14:35 well switches are cloud managed but if

14:38 they're dead in the water somewhere in

14:39 the closet do you have to make that

14:41 20 mile drive or walk all the way from

14:44 uh to the closet or maybe not we can

14:46 actually communicate the status of the

14:48 switch to either a neighboring ap or a

14:52 neighboring switch which is connected to

14:53 the cloud and give the information of oh

14:55 this switch isn't up because it's the

14:56 same three blinky lights the ip

14:58 addresses in there

15:00 consistent

15:01 how do you learn things like management

15:03 vlan like uh some other competitors

15:05 require your management vlan to be one

15:07 in order for all this cloud stuff to

15:09 work do you overcome that or is that a

15:11 similar limitation

15:13 basic standard switching practices um

15:16 either you if you have the ability for

15:18 us to have a

15:20 management vlan as one for it to

15:22 communicate to the cloud that's okay if

15:24 not your upstream switch you apply a

15:25 native vlan having a dedicated dhcp pool

15:28 for that native vlan whatever that may

15:29 be and you can subsequently as soon as

15:31 it connects to the cloud you have full

15:33 flexibility for it to either get a

15:35 separate management vlan other than what

15:36 you have already and also if you don't

15:38 like bhcp management for switches you

15:40 can actually configure them to static as

15:42 well the only onboarding is just

15:44 either with static or native

15:47 yeah

15:51 so at this point of time i think uh like

15:54 you know we are at a stage where we are

15:55 going to plug in uh the the upstream

15:58 have this basically connected uh over

16:01 here you are seeing the are we seeing

16:03 the connectivity status here uh so this

16:05 is the ap

16:07 the ap is

16:10 yeah so so the switch just basically

16:12 got on

16:13 the port profile got assigned and you'll

16:16 basically see the ap come in magically

16:19 so the process that has seen the switch

16:21 insights right the switch went through

16:22 the process of booting up the switch

16:24 went through zero touch provisioning all

16:26 of these uh our providers insights

16:29 the stu the end goal of the story was we

16:31 connected an ap we connected uh we

16:33 brought up the switch

16:35 the deployment framework was for us to

16:37 just plug the device in the device went

16:39 through the process of the entire boot

16:41 up also understanding the connected

16:43 device was an ap automatically and

16:45 that's why the dynamic profile assign

16:47 came through the ap got its profile ap

16:49 was already provisioned with the ssid

16:51 and uh you will see that the ap is

16:54 beaconing the ssid with the right vlan's

16:56 plumb to it so trying and avoiding the

16:58 whole missing vlan situation with them

17:00 put profiles

17:02 so here is you're seeing the ap

17:04 beaconing mfd demo uh a test ssid

17:08 okay

17:09 um

17:10 avi um i would love for you to actually

17:12 talk about our evpn vxlan support while

17:16 i play this video

17:18 sure

17:19 uh

17:20 4100 not only uh support not only as a

17:24 practice just basic access switching but

17:26 also can be a part of your uvp and vxlan

17:28 framework evp and vxlan uh

17:31 from a wired assurance perspective we

17:33 support three broad classifications of

17:36 how you can bring devices on board uh

17:38 multi-homing uh crb or even ipclo

17:42 the most important part of

17:44 the crowd here is is the ip chloe is the

17:47 uh

17:48 form uh feature wherein you can actually

17:50 extend vxlan all the way from your core

17:52 to the access and bring in micro

17:54 segmentation and play and that is where

17:56 4400 is coming to 4100 come into picture

18:00 the the this particular uh example in

18:03 here shows a very uh a quick example of

18:07 us being able to onboard 4100s as to as

18:10 your axe layer and also associate them

18:13 to the fabric uh from from the point of

18:15 uh configuration it's pretty simple you

18:18 just add a simple click uh add the

18:20 existing device the 4400

18:22 and your for the configurations that are

18:24 necessary for the 4100 to be

18:26 participating in the vpn vxlan fabric is

18:29 all pushed down and now your 4100 is

18:32 part of the fabric

18:34 any questions so far

18:37 yes vxlan now is completely gui based

18:39 and not

18:40 cli uh you can actually uh import the

18:44 full blown deployment of evp extend with

18:46 ui that is correct

18:50 so the last one and a half minutes of

18:52 this video is going to focus on uh the

18:54 gene generally i think april early asked

18:57 about third-party data ingestion what we

18:59 do with third-party switches so

19:01 obviously everything applies to juniper

19:03 switches but if you have a

19:06 wireless miss deployment and you have a

19:08 wired third-party deployment you can

19:10 always get a marvelous subscription to

19:13 now see like you know how we can

19:15 basically give you the same insights uh

19:18 from whatever we learn through the

19:20 wireless network right so

19:23 here is an example like you know you see

19:25 the switches tab and a whole bunch of

19:27 different

19:28 metrics appearing up on top like you

19:30 know version compliance switch affinity

19:33 uh like you know poe compliance i mean

19:35 as a as a proof point uh one of our

19:38 customers with two thousand sites in

19:40 only two of them they actually found uh

19:44 a like you know a version mismatch which

19:46 means like you know my rest of my

19:48 universe is deployed on a particular

19:51 version

19:52 but on these two sites i have a version

19:55 mismatch and that is basically going to

19:57 get highlighted right

19:59 and and this is just us learning

20:01 everything from the missed aps and your

20:02 connected switches could be any vendor

20:05 and we can still tell you those switches

20:07 did not have the right version on on

20:09 there and that's that's the message

20:12 right and here like you know uh if you

20:15 have an audit level view like you know

20:17 just like in wireless you would actually

20:19 go through

20:20 all of the different sles for your wired

20:24 network and you can basically look at

20:27 like you know things that are performing

20:29 sub-optimally and you can hone in on

20:31 them so one of the things here that we

20:34 are showing in the juniper uh networks

20:37 campus uh with our wired and wireless

20:40 deployment uh the switch affinity so

20:42 what switch affinity means is like you

20:44 know if how many aps are actually

20:47 connected on a switch and where are

20:50 those ap so you have a location view as

20:52 well right so in an ideal deployment you

20:55 are going to salt and pepper your access

20:57 points on your switches why is that

20:59 important because you control the blast

21:01 radius for failure of a switch

21:04 now what happens when you so this for

21:07 example is a

21:08 is a good example where two different

21:10 switches all the ap is color coded are

21:13 connected to uh two different switches

21:15 sprinkled uniformly across the floor

21:18 right uh

21:20 if if something different were to happen

21:22 and if you can just roll on to the other

21:24 one in this example you show that we

21:27 have two different switches on a

21:28 different site and

21:31 all the all the the floor ap is covering

21:34 the the floor part of the floor are all

21:37 on one switch

21:38 the one the the the remainder of the ap

21:41 is on a different section of the floor

21:43 all on a different switch so what

21:45 happens then is that you have a problem

21:48 if there is a switch outage

21:50 you are going to create a hole in one

21:53 part of your floor right and that is why

21:55 these metrics are important these are

21:57 basically

21:58 the signals that we would give via

22:00 marvis for the network operators to act

22:03 on uh and if there is anything to be

22:06 rectified like you know take the right

22:08 action

22:11 the the third-party switch insights um

22:14 is is all that data or most that data

22:16 just coming from lldp are there other

22:18 features that that switch has to have

22:20 enabled mostly uh like you know i think

22:22 we we look at the lldp information and

22:25 derive all of these insights from lldp

22:27 today

22:28 okay so the more that switch is giving

22:30 the better that is correct yes that is

22:32 correct

22:34 the location the view and the placement

22:37 go ahead

22:39 yes so sorry um um um one thing i just

22:43 wanted to add to this entire discussion

22:46 is

22:47 um just within the fortune 10

22:50 you know we have you know three of the

22:52 largest networks within the fortune 10

22:55 and the smallest of them is is about 30

22:58 plus thousand access points

23:00 all of them are now standardizing on our

23:02 switching they started with our wi-fi um

23:05 and you know one of them now has a

23:07 thousand stores with us another one is

23:09 standardized on on ex switching another

23:11 one is doing a 10 000 retail deployment

23:13 with us with ex switching and and even

23:16 in the enterprise side if they if they

23:20 buy our wireless

23:22 they're almost automatically switching

23:24 to um the ex portfolio from a switching

23:27 perspective right so so what i want you

23:30 to walk away with is the reason we're

23:33 talking about switching in mfbe is

23:37 we see this as an access network

23:40 historically wired and wireless you know

23:42 just because one vendor was selling them

23:44 single throat to choke was what was uh

23:47 um was was the was the play for us now

23:51 it is the single ai engine that connects

23:54 wire and wireless there are only two

23:55 types of devices in our access networks

23:57 there are only two types of devices in

23:59 our networks that are either wireless

24:01 devices or wired devices and we provide

24:04 ai for both of them right and so that's

24:06 why we're excited about sort of bringing

24:08 the 4100 to sort of complete the

24:10 portfolio uh and bringing ai to wired

24:13 wireless all of that sounds great but my

24:16 moral of the story here is we are seeing

24:18 customer proof points brian ward is at

24:21 dartmouth running campus you know ex

24:24 switching plus missed wi-fi mit you know

24:27 is headed there you know a lot of the

24:29 retailers are headed there a lot of the

24:31 enterprises service now you know started

24:33 with wi-fi now going to switching and

24:35 and you know headed to sd-wan with us we

24:38 are seeing proof point of ai adding

24:40 value and customers seeing it and living

24:43 it and actually uh you know uh combining

24:46 this wired wireless portfolio so uh

24:49 that's why we are excited about this

24:50 piece of this sandra and i'll be i'll

24:52 turn it over to you

24:53 thank you sudhir for that commentary

24:56 yes april so i have a question to build

24:58 on that because it is quite exciting

25:00 bringing it together but your products

25:02 do flow control right they look at the

25:05 flow management have visibility flows

25:08 within the switch

25:09 are you starting to move towards an

25:12 end-to-end quality of service flow in

25:14 marvel so i can see over the radio link

25:17 through to the switch through the rest

25:19 of the network and start managing

25:21 end-to-end quality of service

25:24 uh so today with the marvis

25:27 marvis

25:28 story always has been client to cloud

25:30 trying to understand what where where

25:32 the problem was if you see today the

25:34 question of troubleshoot application

25:36 themes for example your team's call

25:39 moves from your your client your access

25:41 point through your switch and then when

25:43 and outside

25:45 throughout this portfolio if there is

25:47 congestion on the switch for example to

25:49 your account

25:50 the the telemetry that we get from the

25:52 switches can actually indicate

25:53 congestion and we point that out your in

25:56 your client to cloud view to say

25:58 uh the issue if and again this is for us

26:01 our two for the you know for the first

26:04 pers first personnel the layer one pack

26:06 that gets the tickets

26:07 for them it should be as simple as

26:09 asking about the person who complain

26:11 take the username and ask them why it's

26:12 why there is a problem

26:14 uh and we take data from wireless wired

26:17 van and if the issue in terms of wired

26:20 uh

26:20 is is congestion which is

26:22 a qos issue we can actually bring that

26:26 out to you and say congestion on this

26:27 port is the same port the client is

26:29 connected to and passing traffic to and

26:31 that's the end goal so uh taking the

26:34 end-to-end flow and identifying where

26:36 exactly the issue yeah and and if i may

26:38 just add avail one other point to what

26:40 what uh abby articulately stated uh

26:43 cloud is the place where we stitched

26:45 things together so whether it's your

26:47 routing policy whether it's your uh

26:49 security enforcement policy for

26:51 segmentation

26:52 cloud is basically the orchestrator so

26:55 all the signals that we are actually

26:57 getting into our cloud from these

27:00 different components of the access

27:01 network cloud is a place where we

27:04 converge orchestrate and write things

27:06 back whether you need to quarantine a

27:08 device or whether we need to go look at

27:10 a like you know a broadcast form

27:13 happening in the wired network uh things

27:15 of that nature right yeah i'd like to

27:17 push back a little if you wouldn't mind

27:19 because when you think where your

27:21 comment is we're very focused on the

27:23 client

27:24 client is running an application

27:26 client needs a certain level of quality

27:28 of service to successfully do that

27:30 congestion is important and a very

27:32 important first step but what i really

27:34 want to see is do i have quality of

27:36 service end to end for my user

27:38 experience sir today we don't do it

27:41 right um and and we we need to do it and

27:44 we want to get there i just want to

27:46 answer straight that

27:48 that that is that is a plot we need to

27:50 head to and we're getting there