ZKast with Sudheer Matta of Juniper Networks

Introduction

0:00 [Music]

0:03 foreign

0:06 caraval from ZK research and I'm here in

0:10 person at Juniper's Enterprise analyst

0:13 and influencer Summit and I'm joined

0:15 today by Sadir Mata who's a group VP for

0:18 uh Juniper's Cloud product Mist right

0:21 yeah I do an Enterprise product yeah

0:22 yeah so just give a quick buy one to

0:25 yourself and tell us what you do here

0:26 first of all Aziz thank you so very much

The future of networking

0:28 for the opportunity and thank you for uh

0:30 being part of the uh Enterprise Summit

0:33 for us here at Juniper uh so hello

0:35 everybody severe Mata iron product

0:37 management for our AI driven Enterprise

0:39 portfolio which is Wireless switching

0:42 sd-van and soon to be announced uh our

0:45 Netflix access Assurance uh um product

0:47 portfolio as well that's a lot of stuff

0:49 it is so normally when I do these zcasts

0:51 I I interview the guest but we're going

0:54 to flip things around a little bit I

0:55 understand there's a few questions you

0:56 want to ask me first before we get going

0:58 that's right so fire away that's right

1:00 so Zs obviously you spoke you speak to a

1:03 lot of customers vendors partners and so

1:08 cloud is top of mind for a lot of them

1:10 it is so so talk about sort of you know

1:14 customers moving to the cloud where they

1:16 are in in especially when it comes to

1:18 network where they're in their journey

1:19 and and some of the challenges you think

1:21 yeah it's it's funny I always say if you

1:24 want to understand the future of

1:25 networking look what happened elsewhere

1:27 in I.T and that'll happen in networking

1:29 you know five six years later right and

1:30 I think I think that's planning out

1:32 again with Cloud obviously Cloud's been

1:34 a big part of Enterprise strategy for

1:36 the better part of what two decades uh

1:38 it's relatively new to networking last

1:41 half decade or so and I I'd say part of

1:43 that is just because of the relative

1:45 importance of the network right we don't

1:47 really get a second chance if if you

1:49 mess up the network so there's been a

1:50 lot of conservativism from Network

1:52 buyers but I think a lot of early

1:55 adopters did move and part of your

1:57 customer base is that but I think what

1:58 the the pandemic did was it ushered in

2:01 this new era of networking where if I

2:03 don't have uh my it Pros in the office I

2:06 know my users in the office and I'm

2:08 trying to manage This Global Network I

2:10 got to do things differently and so the

2:12 old way of sitting there and hunting and

2:13 pecking at a CLI while I'm directly

2:15 attached to the router the switch

2:17 doesn't work anymore right so I do think

2:19 it's uh it's something most customers

2:22 are looking at I think they're looking

2:23 to really change their operational model

2:28 and automate more and that's been a big

2:30 change too because frankly three four

2:32 years ago when I'd say automate to it

2:33 Pros they go whoa whoa whoa you're

2:35 taking our job but now they understand

2:36 automation is a tool with which you do

2:39 their job better but still in networking

2:41 it's it's less than half I'd say

2:43 probably under 30 percent of companies

2:46 have actually migrated their management

2:48 to the cloud yeah and and for those that

2:51 have not moved why do you think that is

2:54 I just think it's a lot of conservatism

2:56 and uh not wanted you know that

2:59 expression if it ain't fixed don't broke

3:01 it or any broke don't fix it but I think

3:04 in a lot of cases it is is broken they

3:06 just don't know it right if you look at

3:08 a lot of the the challenge run running

3:10 networks now human errors still accounts

3:12 for the number one cause of downtime

3:13 right so that's something we got to get

3:15 rid of uh you know here's another

3:17 interesting stat you know that three

Legacy infrastructure

3:19 quarters of trouble tickets are opened

3:21 by the user not the I.T department and

3:23 what that does is it puts us in reactive

3:25 firefighting mode right and then there's

3:28 just uh I I think there's a lot of

3:30 Legacy infrastructure out there right

3:32 and so that stuff's got to get removed

3:33 and modernized before we can actually

3:36 move to more of a modern Cloud

3:38 architecture you know for our Network

3:40 yeah so so you pretty much sort of

3:42 summarized

3:44 um sort of our premise that the network

3:46 is going to move to the cloud and so

3:47 that's how we started so I'll let you

3:49 actually now carry on with your part of

3:52 the interview and the questions yeah so

3:54 uh Juniper missed right it's a lot of

3:57 momentum for you guys in fact when I

3:58 watch your earnings calls Enterprise is

4:00 growing like crazy for you right it's

4:02 now uh uh I think bigger than it's our

4:04 number one word yeah which is which is

4:06 amazing I would never have thought that

4:08 for Juniper years ago so but talk about

4:10 Mist in the cloud architecture and how

4:12 it works yeah so so thank you so so

4:15 um for those of you that that may not

4:16 know what a mist is uh mist is a company

4:19 that got founded in 2014 uh by sujay

4:22 hajela Bob Friday we get started on uh

4:25 in Wi-Fi and then when we merged with

Mist

4:28 juniper uh we became the Enterprise

4:30 portfolio of juniper so now you know we

4:33 came in with Wi-Fi then added switching

4:35 SD van and now you know access Assurance

4:38 which is the NAC portfolio that we just

4:40 launched and so uh this is the portfolio

4:43 all you know the founding principle of

4:47 mist was we're going to be Cloud native

4:49 and AI driven from day one right

4:51 everything we do Cloud native using the

4:54 elastic sort of scalable public Cloud

4:56 architecture and so that's what uh

4:59 that's why this is the our Enterprise uh

5:02 business unit internally we call it AI

5:04 driven Enterprise we just we want it to

5:07 be ingrained within our name that we are

5:10 AI driven Cloud native and that's what

5:12 that's what the portfolio is yeah and

5:14 you know one of the reasons uh one of

5:16 the things I liked about Miss was that

5:17 you did start with Wi-Fi and I've always

5:19 said that within the confines of

5:21 Enterprise networking there is nothing

5:23 harder than Wi-Fi troubleshooting right

5:26 in fact I've talked to some

5:27 organizations where they have Engineers

5:29 that spend you know a full day a week

5:32 doing nothing but Wi-Fi troubleshooting

5:33 right so if you can solve that problem

5:35 yeah you can solve a lot of the other

5:37 ones and so I'll give you a simple

5:39 example on that I used to run a school

5:42 district Network for my local school

5:44 district with the IIT team when I was at

5:46 Cisco and you know they were having

5:49 issues students not able to connect and

5:51 so they said hey can you come and

5:52 troubleshoot literally we took eight

5:54 Engineers from the partner you know four

5:56 of us from Cisco four from uh the

5:58 customer you know a massive team of

6:01 people we went classroom to classroom

6:02 opening laptops or you know trying to

6:04 capture Why students can't get on the

6:06 network

6:07 and no one knows and nobody knows

6:09 because obviously what happens when you

6:11 go to the doctor it works right and so

6:13 whenever whichever classroom we went the

6:15 laptops were working fine so anyway long

6:17 story short we are essentially from that

6:21 experience uh essentially you know a few

6:24 days later that that I quit Cisco and

6:26 joined Miss because I like wow there's a

6:28 better way to build networks than people

6:29 going to classrooms opening laptops out

6:32 of that experience came a feature we

6:34 call Dynamic packet capture so never

Cloud management

6:37 ever do you have to send 15 St you know

6:39 people to classrooms trying to you know

6:42 figure out what what is happening so it

6:45 truly cloud enabled a way to capture

6:48 failures in the Network store them if

6:52 you need them but maybe even react and

6:54 fix the network uh as you see it yeah

6:57 and that's great for chronic problems

6:58 because those are the ones that seems

7:00 networking you know I'm a former network

7:01 engineer myself and Wi-Fi you could you

7:03 just hope it went away that's right

7:05 that's right so now there are a lot of

7:06 cloud management tools today I mean

7:09 Cloud management is not unique to

7:11 Juniper so how do you feel you're

7:13 differentiated in amongst this ocean of

7:15 many Cloud management tools great

7:17 question so so uh you know uh in the in

7:20 the networking space you know Meraki was

7:22 definitely the one that started first uh

7:25 you know with a with a cloud driven

7:26 architecture then came a Ruba and then a

7:29 bunch of other folks but really missed

7:31 had a clean sheet of paper opportunity

7:32 so the difference fundamentally is you

7:37 know Meraki was basically a set of

7:39 servers in A Shard running in private

7:42 data centers right because Meraki was

7:44 founded before the iPhone was invented

7:46 yeah and so it was it was Private data

7:48 center based right and then a bunch of

7:50 other companies tried to move their

7:52 controllers to the cloud and all this

7:53 kind we had you know born in 2014 we had

7:57 the the wind in the sale of us of AWS

8:00 and Azure and gcp and so we use the

8:03 elastic scalable public Cloud so number

8:06 one elastic scale availability right our

8:09 largest customer today has a hundred and

8:12 eighty thousand access points in one

8:14 instance that's a big Network and and

8:16 then they doubled down and bought 25 000

8:19 switches adding another 30 000 switches

8:21 this year right there isn't a Meraki

8:25 Cloud that can manage you know 250 000

8:28 devices like that in one you know

8:29 instance uh the some of the other folks

8:33 haven't even hit scale like that right

8:35 so a a hyper differentiated you know

8:38 elastic scalable cloud is number one and

8:41 then it's a very micro Services

8:43 architecture right basically you know we

8:46 took what existed in the wireless LAN

Cloud native

8:49 controller World completely piece parted

8:52 the functions and and built micro

8:54 Services stateless microservices so we

8:57 can update them at will without causing

9:00 downtime right and so that's the second

9:03 so elastic scalable public cloud and

9:06 microservices core or differentiation

9:08 for us yeah and I'm glad you brought

9:10 that up because not all clouds are

9:12 created equal right so uh and it's a

9:14 it's you know for everybody watching

9:15 this it's this is important to

9:17 understand picking up a legacy workload

9:19 like a wireless controller yeah and

9:23 putting it in the cloud doesn't make it

9:25 Cloud correct it just makes it Cloud

9:26 resonant right not versus Cloud native

9:28 and there's a big difference between

9:30 Cloud resident Cloud native I think

9:31 that's what you're saying yeah yeah and

9:33 it you know there are easy ways in which

9:36 you can you can you know figure this out

9:38 right so for example

9:40 um for a cloud native architecture like

9:42 ours we are able to release new features

9:45 to the cloud right in fact yeah in fact

9:48 you do a bi-weekly uh update yeah yeah

9:51 and uh most of the cloud management

9:53 tools they actually have to shut down

9:54 that's right right so how are you how

9:56 are you doing that yeah so so so most of

9:59 our competitors you know uh have to take

10:01 down time on their Cloud I mean the the

10:04 largest competitor you know uh you know

10:07 a Rubik as an example takes down time

10:09 probably 8 to 12 hours every quarter to

10:11 upgrade their cloud in our case we've

10:14 not had to take down time at all right

10:16 so so and we upgrade production networks

10:20 while they're running while they're

10:22 changing configs while their user using

10:24 the net that isn't scary customers

10:26 initially it scares one of the one of

10:29 the customers actually uh they were like

10:31 what are you doing to us so you're going

10:33 to change my network every two weeks

10:36 without permission from me

10:38 fast forward to two years into their

10:40 deployment they say it's the single best

10:43 feature they've had from us because they

10:46 get new features and architecturally

10:48 they experience no downtime right from a

10:51 user perspective no downtime but they

10:53 get new features it's it's an

10:54 interesting architecture so the question

10:56 was how do we do this right and so the

10:59 the the technicalities actually if I

11:02 explained it is very it becomes very

11:03 simple so because our micro services are

11:07 stateless so we have let's say 12

Inline updates

11:09 instances of a micro service running I

11:12 can update the instance one the

11:14 remaining 11 are carrying the load and

11:16 then when it's updated it joins the

11:18 queue again and so I can you know in a

11:21 round robin fashion do the upgrade none

11:23 of this you know instances are have

11:26 state in them and so they can seamlessly

11:28 pick up where they leave and then they

11:29 continue and so we are doing inline

11:31 updates across all the micro services in

11:34 the cloud hundreds of them

11:36 with zero downtime and zero scheduled

11:39 outage yeah that's amazing that you can

11:42 actually do that and yeah you know when

11:44 you think about other uh it

11:48 um parts of it that move to cloud like

11:49 application development it's very common

11:51 to update an application on the fly

11:53 right right in the consumer world that

11:54 happens all the time you look you get a

11:56 new version of some applications yeah

11:58 yeah but in the in the networking space

12:00 we've resisted that and I think

12:02 the downside of doing that again if you

12:04 if you want to think of the if it ain't

12:06 broke don't fix it you're not getting

12:08 the security updates you're not getting

12:10 new features that are going to make your

12:11 network run better right which is why

12:13 when you look at a lot of large Legacy

12:15 networks out there they're running 10 20

12:18 30 40 maybe even 100 different versions

12:21 of operating systems they've got all

12:23 these different features it's very hard

12:25 to keep things in sync so things will

12:27 work out well in one box and not another

12:29 and then you wind up with the mess that

12:31 we have today yeah yeah and actually

12:32 you've touched on one of the most

12:35 important benefits of this a lot of

12:37 people don't come to Cloud because they

12:39 think from a security perspective oh my

12:42 God it's outside my uh my network and

12:45 how do I secure it in fact it's the

12:48 opposite Cloud gives you agility to

12:50 apply security patches you know when

12:53 when the crack virus uh crack a

12:55 vulnerability got announced right we had

12:58 a release on the same day the most

13:01 recent the log 4J yeah you know

13:03 vulnerability that was announced we

13:05 remediated on the cloud globally for our

13:08 customers without them lifting a finger

13:10 versus everyone else had to upgrade

13:13 their own network because the controller

13:15 had vulnerabilities right and so

13:16 security is a great outcome in fact the

13:19 cloud is more secure and more resilient

13:23 than your on-prem networks yeah right

13:25 it's it's very counter and you would you

13:26 wouldn't have said that a few years ago

13:28 either right yeah you couldn't yeah you

13:30 could so now you've used this term scale

13:31 over and over

13:33 um when you say scale what do you mean

13:36 by that how do you think about that and

13:37 how does mist adapt uh scale up yeah so

13:41 so let's talk about scale up a little

13:42 bit right scale up from a the size of

Scale up

13:44 the network perspective and and the

13:46 operations perspective so I'll talk

13:48 about the operations first when Gmail

13:50 first came out Google had a distinct

13:54 goal that any page in Gmail should load

13:57 in 100 milliseconds less than 100

13:59 milliseconds right it doesn't matter how

14:01 much email you have in your email inbox

14:03 they want to load in in less than 100

14:05 milliseconds I get a lot of email yes

14:08 we had a similar goal that it didn't

14:11 matter what size of network you're

14:13 operating with juniper Mist we want you

14:15 to load in under two seconds right so if

14:18 you are Fortune you know 10 company with

14:21 a scale of Fortune 10 Network we still

14:24 want your network to just be buzzing you

14:26 know you be fast and and so you know a

14:29 lot of the Legacy

14:30 vendors if you do on-prem you know DNA

14:34 Center Cisco Prime blah blah blah you

14:37 are responsible for scale up in our case

14:39 we are responsible so that your

14:41 experience is always delightful that's

14:43 number one that's from a it experience

14:46 perspective from a user experience

14:48 perspective we think of scale as we just

14:50 want it to be elastic scalable right we

14:52 do not want you to have to worry about

14:54 because think of all the things that

14:56 current controller-based customers have

14:58 to worry about oh my controller has a

14:59 capacity of 5000 oh I need 200 more APS

15:02 oh God I need a new controller oh it

15:04 doesn't support my old AP image I mean

15:06 it's all of that magic poof goes away

15:08 right elastic scale and that's what the

15:10 my the public Cloud got for us yeah and

15:13 uh okay so now I'm going to flip a

15:15 question back at you asked me before why

15:17 aren't customers with the cloud what are

15:18 customers telling you why haven't they

15:20 moved the cloud and how do you convince

15:21 them that something they should do yeah

15:23 so so I would say you know exactly like

15:25 what you said is it it's it's either

15:27 inertia because don't fix it if it's not

15:30 broke or it's fear it's really just

15:33 those two feelings for them to go to the

15:36 cloud the the honest truth is all of the

15:39 vendors are going to the cloud if you're

15:40 Cisco you're going to the cloud with

15:42 Meraki if you're a rubber you're going

15:43 to the cloud with Central if you're Mr

15:45 coming to midst right everybody's going

15:46 into the cloud so so that's not a

15:49 question anymore whether your network is

15:51 going to the cloud or not all vendors

15:52 are going there so let's get on with it

15:54 now the the thing that actually you know

15:57 people struggle with is is is is just

16:01 you know how do I think about security

16:03 and here's the proof point I want to

16:05 give you right the world's largest

16:08 Enterprises including the most stringent

16:10 Enterprises like for example we've we

16:13 have won

16:14 um two of the largest five banks in the

16:17 in the world right literally the world's

16:20 largest financial institutions for the

16:21 wired wireless networks are coming to us

16:23 right and so they've done the reviews on

16:26 on security the top retailers in the

16:29 world are running on us in every

16:30 continent the number one retailer is

16:32 running on us

16:33 um and then we just most recently

16:35 launched our fedramp Cloud so so

16:38 basically U.S federal customers are

16:41 jumping on Australia Federal customers

16:43 other Federal customers are actually

16:44 looking at this as well and so

16:47 I want people to to think and assume

16:50 that hey the cloud just like you

16:53 validate security when you go to a

16:55 Office 365 type of cloud you know do all

16:58 of the checks but the cloud is secure

17:00 and other customers have done the due

17:02 diligence and and and the largest you

17:05 know customers in every vertical are

17:06 coming to us yeah now um at the

17:09 beginning you describe this as an AI

17:10 Cloud so what's the relationship between

17:14 ai ai Ops in the cloud yeah so so so

17:17 it's it's it's a very symbiotic a very

17:21 necessary relationship right you can

17:23 almost argue the AI and Cloud you can't

17:25 do AI without the clock you cannot do AI

17:27 without the cloud you absolutely cannot

17:29 right because you know we have we

17:31 measure user experience every user every

17:34 minute for all the users on the network

17:36 and that's what we use as the the data

17:39 for the foundation for the AI we needed

17:41 an elastic scalable club and that's

17:43 where you know the first employees

17:45 missed hired had nothing to do with

17:46 networking these with Cloud Engineers

17:48 they built you know literally the first

17:50 three employees never never built a

17:53 network never never were in networking

17:54 they were elastic scalable Cloud

17:56 engineers and that became the foundation

17:58 so if we need our elastic scalable Cloud

18:00 for AI yeah I think uh just an easy way

18:03 to think about it is AI requires lots of

18:06 horsepower right most companies don't

18:07 have that on their own Prem because lots

18:09 of data Sciences that's right to be able

18:11 to train the models and things like that

18:13 and even if a company had those things

18:15 I'm not sure that their data set would

18:17 be large enough to train it with right

18:19 right and so that's um you know and

18:21 that's what you get from the cloud right

18:23 or at least the cloud vendor like

18:25 yourself I do think an interesting uh

18:27 data point that you guys published those

18:29 the efficacy of your of your AI and

18:31 right uh because let's let's be honest

18:33 it's not perfect right right but it gets

18:35 better with time correct and I do think

18:37 that sometimes when I talk to network

18:40 Engineers they go we tried Ai and it

18:41 didn't work right because it made one

18:43 error and I might I've always said don't

18:45 wait for Perfection that's right as long

18:46 as is better than people that's right

18:48 right and as I mentioned human error is

18:50 still the largest cause of downtime so

18:52 that's right you know

18:53 I I think the you know AI is going to do

18:56 a better better job of that I'll give

18:57 you a simple example of this I think in

18:59 the in the spirit of what can AI do for

19:02 you if you're a a wired Wireless

19:04 engineer can AI impact you here's the

Can AI impact you

19:07 simplest example we deploy the most

19:09 complex networks in the world I mean

19:12 large robotic facilities and warehouses

19:14 to higher education you know a large

19:17 University uh around the world

19:20 we deploy our APS obviously you design

19:23 you know you we deploy our APS and

19:26 basically

19:28 um you know essentially we ask the

19:30 engineers to step away from the keyboard

19:31 right we let the AI self-optimize better

19:35 than any ccie has ever been able to

19:37 optimize the network every missed

19:39 Network around the world is built that

19:41 way where we trust our AI to to build

19:45 the most efficient Network

19:47 better than humans can in every

19:49 environment and that's been the best

19:51 outcome I'll give you a couple of

19:53 examples right so give me some customer

19:55 examples in the business value that are

19:57 brought yeah yeah so so Jim Pike at

19:59 Newport News schools right so

20:02 um they were 3 000 AP Aruba shop lots of

20:05 you know fine-tuning the school district

20:08 Network school district is a heavily

20:10 used Network Jim says after they

20:13 deployed the 3 000 missed APS AP to AP

20:16 they replaced no additional APS no

20:18 additional cables exact same location

20:20 same amount of rapes

20:22 zero tickets to an entire school year of

20:24 testing zero tickets right and then in

20:28 the higher ed space Dartmouth College 70

20:31 reduction in user tickets coming in and

20:34 off the 30 of the tickets that still

20:35 come in you know 80 of them are solved

20:39 by first line Half tasks because of the

20:40 visibility AI has impact on the

20:42 operations right AI has an impact in

20:45 deployment you know we displaced 24 000

20:48 Meraki APS at the Gap with 24 000 missed

20:52 APS no one API additional AP added no

20:54 additional cable pull

20:56 85 reduction in the number of site

20:59 visits right number of site visits

21:01 across and those are expensive expensive

21:03 site visits around the world right so AI

21:05 has an impact yeah and so uh one last

21:08 question and uh I we can't talk AI

21:11 without the topic of chat GPT and uh

21:14 General AI coming up so what are you

21:16 guys thinking about there how how will

21:17 that uh work into what you guys are

21:20 gonna missed we're looking at Shops

21:22 because we love this we we you know you

21:25 know Bob Friday uh I mean almost very

21:28 Visionary four years ago he said we went

21:31 from CLI to dashboards and we're going

21:34 to go to conversation assistants 40

21:35 years ago you know our competitors sort

21:37 of mocked us saying you know they're

21:39 losing their dashboards

21:40 um but you know we started Marvis

21:43 conversation assistant four years ago

21:45 we've we've had a very mature uh

21:48 conversational assistant for uh for

21:50 helping with you know solving problems

21:51 on the network

21:53 um what we have already had is NL you

21:56 natural language understanding and we

21:59 have all the AI models for anomaly

22:01 detection to find marvelous actions and

22:03 stuff what we are super excited about uh

22:06 you know with with GPT and chat CPT in

22:09 specific is is nlg natural language

22:12 generation yeah so what you saw on May

22:15 17 that the mobility field they launched

22:18 is that we are now natively integrating

22:20 chat GPT into marvelous conversational

22:23 system so chat CPT the power of chat TPT

22:26 is now in The Marvelous conversation

22:28 assistant that's impressive so uh all

22:30 right so here uh I think we'll run up

22:32 against the clock here I want to thank

22:33 you very much for your time on behalf of

22:36 the year of Zs caraval from ZK research

22:38 saying uh don't forget to hit the

22:39 Subscribe button and I'll see you next

22:41 time on another zcast thank you very

22:43 much

22:45 [Music]

22:50 foreign