Extending AI to Enterprise Routing

0:06 hello and welcome to extending AI to

0:09 Enterprise routing I'm Bob La liberte

0:11 principal analyst with the cube research

0:14 and today I'm joined by some special

0:15 guests Bob Friday the chief AI officer

0:17 in CTO Enterprise of Juniper Networks

0:20 and Kena Atri the senior director

0:22 product management also of Juniper

0:25 Networks and we're here to discuss the

0:27 latest announcements uh around Juniper's

0:29 introduction of native AI onto Edge

0:32 routers and that would include both

0:34 uniper missed routing assurance and

0:36 Marvis for routing so welcome Bob and

0:39 Kena thank you for having us happy to be

0:42 here Bob thank you yeah absolutely well

0:45 let's get started you know when I think

0:47 about Juniper Networks you know really

0:48 from its Inception decades ago it's

0:50 really been instrumental in driving

0:52 innovation in the routing market and you

0:54 know when we think about Innovation

0:56 today that really means AI technology

0:58 and certainly we follow Juniper for a

1:00 long time and you've been pioneering the

1:02 use of AI Ops through the in the network

1:04 space and in fact I think it was just a

1:07 few months ago You released the AI

1:08 native networking platform which

1:11 delivers both extensibility and value

1:13 across the entire network so you know

1:17 when I when I think about why that's

1:19 needed and what's going on from an

1:21 analyst perspective you know I really

1:22 start thinking about the changes that

1:24 have occurred in the modern it and

1:26 application environments and that being

1:28 you know they've become highly dynamic

1:30 and also highly distributed right

1:32 applications are deployed across

1:33 multiple data centers multiple public

1:35 clouds and numerous Edge locations and

1:39 the unfortunate part of this is it means

1:41 the environment has become far more

1:42 complex it also means that the network

1:45 plays a much more significant role

1:47 especially across those wide area

1:49 networks right and making sure that

1:51 these distributed environments are able

1:53 to have a positive experience and

1:55 they're able to enable the business but

1:58 before we jump into a full-on AI

2:01 discussion I thought it would be good if

2:03 we could provide some context for the

2:05 audience and I believe it was just a few

2:07 years ago that Juniper Networks

2:09 announced its client to Cloud Vision um

2:12 and it really focused on providing great

2:14 experiences to end users so I'm

2:17 wondering if you could share maybe you

2:20 know the vision behind that approach and

2:22 how Juniper brings that Vision to life

2:25 yeah you know Bob let me jump in here

2:27 you know for me personally you know the

2:29 cloud AI ADV Venture started when I was

2:31 really at Cisco you know I was talking

2:33 to a bunch of large retail customers you

2:36 know and what they told me when I was

2:37 back then they said hey Bob if if we're

2:39 going to put like a connected mobile

2:41 experience onto this network you know a

2:44 business critical app they're like you

2:46 got to promise me that your controllers

2:47 are not going to crash you got to make

2:49 sure you can deliver code more than once

2:52 or twice a year and more importantly

2:54 you've got to guarantee that there's

2:56 going to be a great mobile experience

2:57 when we connect to that Network you know

3:00 and that is when I realized there was a

3:01 fundamental Paradigm Shift right where

3:04 we were going yes we have to keep AP

3:06 switches and routers got to keep

3:07 everything up and running in green but

3:09 what was more important was we had to

3:11 make sure that that user is going to

3:13 have a great experience and that's when

3:15 I realize that we're really going into a

3:17 day two real-time operation mode in

3:20 addition to day zero day one we really

3:22 had to make sure that day2 experience

3:24 was going to be great you know and

3:26 that's when suj and I decided to leave

3:27 Cisco because we decided that this is

3:30 really going to be an architectural

3:31 change in the industry you know this

3:34 really required a blank sheet of paper

3:36 where we could start from scratch and

3:37 building a new microservices Cloud

3:40 architecture on which we could do

3:41 real-time data processing you know and

3:44 that is why we started with the access

3:46 point because we were trying to answer

3:48 the question of why is that user having

3:50 a poor internet experience and it turns

3:53 out that the access point or the edge of

3:55 the network has a lot of the data you

3:57 need to answer that question and the

3:59 reason we built an access point wasn't

4:01 because we thought the industry needed

4:03 another one it was really around making

4:05 sure we get the data to answering that

4:07 question you know and since join Juniper

4:09 in 2019 what you've seen us do we

4:12 basically have extended that cloud AI

4:14 Ops coverage across the Enterprise

4:17 portfolio right going from the AP to the

4:20 switch to the router you know what we're

4:22 talking about here today is extending

4:23 that across the WAN router inside the

4:25 Enterprise these are these large uh

4:28 routers we see in the Enterprise so that

4:30 is really where the vision started you

4:32 know I think the other inspiration was I

4:34 tell people you remember Watson playing

4:36 Jeopardy I figured hey guys if they can

4:38 play Jeopardy we should be able to play

4:40 networking so that's how the vision got

4:42 started you know since join Juniper

4:44 we're basically just a continuing that

4:46 adventure and extending that cloud AI

4:48 Ops across the complete Enterprise

4:51 portfolio and now extending across SP

4:53 and data center

4:55 domains excellent yeah that makes that

4:57 makes a lot of sense and Kena I wonder

5:00 if you could maybe talk to us a little

5:01 bit about the announcement that you have

5:03 today obviously we're extending those AI

5:06 C capabilities into Enterprise routing

5:08 so maybe you could give us a little of

5:09 the details about what you're bringing

5:11 out absolutely Bob it's a great uh

5:14 momentous occasion for us to launch

5:16 missed routing assurance and this is a

5:19 classic story of 1 + 1 equal to 11 we're

5:22 already number one uh in terms of

5:25 innovation leadership in the routing

5:26 space uh with very very loved platforms

5:30 like the MX like the PDX serving up the

5:33 edge um and then we're already number

5:35 one in the AI op space with mist you

5:38 know Trail blazing in every imaginable

5:41 gner magic quadrant that there is what

5:44 happens when you bring them together

5:46 magic right so that's what we are

5:48 launching today is missed Enterprise

5:51 routing Assurance which is targeted for

5:53 our Enterprise customers uh who are

5:56 buying a lot of the routing uh layer

5:58 from us serving in different use cases

6:00 whether it's the private one Edge

6:03 whether it's the edge of the data center

6:05 whether it's their Cloud connect or

6:07 appearing in all those roles now these

6:11 customers can hook up their mxs acxs

6:14 ptxs all are Juniper routing gear right

6:17 onto the Mist infrastructure and benefit

6:19 from

6:21 that yeah absolutely I think you know

6:23 extending support to that routing space

6:25 to the W it really seems like a natural

6:27 evolution of where you're going and the

6:30 vision of what you want to do and being

6:31 able to I think as Ramy likes to say you

6:34 know just adds to that flywheel effect

6:35 of value the more domains you can bring

6:37 in the more context you can provide the

6:39 more value it delivers to

6:42 organizations um I'm wondering um you

6:45 know how how will Wan data complement

6:49 now saying that how will that complement

6:51 your endtoend AI op strategy Bob yeah

6:54 you know so you look at a great example

6:56 of what we announced last week so last

6:58 new week we announced something called

6:59 continuous learning for zoom and teams

7:02 you know and similar to you look how

7:04 open AI took trillions of words to train

7:07 chat GPT to predict the next word you

7:10 know what we're doing with our deep

7:11 learning here we're taking billions of

7:14 video collaboration data points and

7:16 training these deep learning models now

7:18 to predict the actual user experience on

7:21 video collaboration zoom and teams you

7:23 know and we're combining that Zoom teams

7:25 data with network features right so the

7:28 more Network features we get into the

7:30 model the better we can get to the

7:32 granular root cause so what the WAN

7:34 network brings is really a visibility

7:36 into the the WAN component you know when

7:39 you look at that client to Cloud

7:41 experience there's a couple of key

7:43 components here right we have that

7:44 client side the wireless link the land

7:48 and that Wan provider is the other area

7:50 that can cause pain to our customers so

7:52 what the WAN is bringing to us now is

7:54 visibility into that W connection so we

7:56 can really make better predictions on a

7:58 user experience in get to the root cause

8:00 of exactly why they're having a poor

8:02 user

8:03 experience yeah I think that makes a lot

8:05 of sense and if I can don't mind the pun

8:06 you're demistifying what's happening in

8:09 that in that W environment and being

8:12 able to provide context for users

8:13 especially when they're using I I think

8:15 you know you can honestly say zoom and

8:16 those other video collaboration tools

8:18 are mission critical and these days when

8:20 people are working in a hybrid mode and

8:21 working remote they need to be up they

8:23 need to be delivering a positive

8:24 experience so I think all that makes a

8:26 lot of sense Nique I want to go back to

8:29 you you know why are you doing this now

8:33 what are some of the trends that you're

8:34 seeing in this Wan Edge space that

8:37 really require Enterprise routing

8:39 architectures to be

8:41 changed absolutely so uh if we focus on

8:45 this particular uh customer segment uh

8:48 representing the

8:50 Enterprises they number one they

8:53 themselves are going through uh their

8:54 own digitization Journey that started

8:57 almost you know a decade ago and now we

9:00 see that you know even doubling down on

9:02 it especially accelerated by this AI era

9:06 around us where the Enterprises want to

9:08 consume more and more AI applications so

9:11 on one hand you know this with this

9:13 Enterprise digitization the van Network

9:17 their van Edge becomes even more

9:20 important and it needs to be very high

9:23 performance right so that's part one

9:26 second um the Enterprise they are not in

9:30 the business of selling networks for

9:32 them the network is an enabler helping

9:35 them sell what they do right whether

9:37 it's a retail whether it's a healthc

9:39 care whether it's a education vertical

9:42 they're not in the business of making

9:43 money using networks so what they really

9:45 want is a network that just

9:48 works and it's even more important

9:52 because there is a talent shortage right

9:55 uh in order to run complex ipms based

9:58 networks uh there is a talent shortage

10:00 right so to do that they really are

10:04 heavily relying on automation to help

10:07 them uh basically not have to even worry

10:10 about the network right uh load it shut

10:13 it forget it and then it takes care of

10:15 itself right and that is where AI Ops

10:18 and automation really help them in their

10:21 journey and the third thing from an

10:23 Enterprise perspective right the

10:26 sustainability goals have become uh have

10:28 come on the Forefront now not only is it

10:31 about meeting ESG mandates but even in

10:34 terms of the whole uh network operations

10:37 and the total cost of running their

10:39 entire infrastructure they need to start

10:42 thinking about optimizing that power

10:45 that space that longevity and the whole

10:47 TCO related to

10:49 sustainability uh and in doing so you

10:52 know uh Juniper is actually working on

10:55 all these three missions of performance

10:57 automation sustainability to help

11:00 reimagine that W Edge

11:04 architecture yeah I think that makes a

11:06 lot of sense you know a lot of those

11:07 challenges you brought up we see as well

11:09 that you know difficulty finding skilled

11:11 resources um that ability to overcome

11:14 the complexity and so those are a lot of

11:16 the things that AI Ops is specifically

11:18 designed to help with so to be able to

11:21 have the same number of resources be

11:23 able to manage a much larger and much

11:25 more complex

11:27 environment um and it's also great the

11:30 fact the other thing I like up from my

11:31 product management days I always had

11:33 something I referred to is that

11:34 principle of least astonishment so as

11:36 you continue your flywheel effect and as

11:38 you extend your AI tools and your um

11:42 Marvis assistant across all of those it

11:44 becomes that principle of least

11:45 astonishment if you have to work in

11:47 different domains you're going to be

11:48 extremely comfortable and familiar with

11:50 the network management tools and the

11:51 assistance that are available to help

11:53 you do

11:54 that um one of the other things I wanted

11:57 to drill down into a little bit you know

11:58 some same lines of looking at the trends

12:00 and what's happening do you see any

12:03 specific organizations that will benefit

12:05 most from having this AI Ops in the WAN

12:08 or is it Universal across all of them

12:11 yeah I mean I think what we're seeing in

12:12 the Enterprise space right now is that

12:14 our large campus customers right you

12:16 know when you look at higher ed

12:18 Healthcare large financial institutions

12:21 all these institutions have large Wan

12:24 routers complex back call systems in

12:26 them you know if you look what's

12:28 happening in inside these big

12:29 organizations right all those

12:31 applications that used to run in the

12:33 data sign all those applications have

12:35 moved out to the cloud what this is

12:37 requiring is very complex R landan

12:39 routing U pipes or underlays overlays

12:42 and getting traffic between data centers

12:45 between campuses so big large Enterprise

12:48 campuses are biding from this I've

12:49 talked to one Enterprise uh customer you

12:52 know and what he just wanted to do is he

12:54 wanted to make sure that he was getting

12:55 the data from those W routers back to

12:57 the cloud you know so before you even

13:00 get to AI half the value is getting the

13:03 data you need back to the cloud for

13:05 visibility and observability and that by

13:08 itself brings value once you get that

13:09 data exposed to a larger group I would

13:12 say the other class of customers we're

13:13 seeing that see value in this is

13:15 probably our tier 2 service providers

13:18 msps which to some extent look like

13:20 large Enterprises you know and what

13:22 they're really looking for is turn

13:24 reduction right they're looking at user

13:26 experiences that making sure there's no

13:28 turn reduction so I think we see our

13:30 large Enterprise campus customers seeing

13:33 value and bringing the WAN data back to

13:35 their team zoom models and then on the

13:38 MSP service Rider side we're seeing our

13:40 tier two seeing value and helping them

13:42 deliver better user experiences and

13:44 reducing turn in their

13:46 business excellent no that that makes a

13:49 lot of sense and so for all those

13:51 industries that are looking to to get

13:53 these

13:54 capabilities I was wondering if you

13:56 could tell us a little bit more about

13:58 the specific capabilities that will be

14:01 offered through mist and are available

14:02 to the customers at the launch today so

14:05 capabilities which products it will

14:06 cover Etc sure uh so we're launching

14:10 with uh support for our largest uh

14:13 hottest sellers like MX 204

14:16 mx34 and the access routers in the ACX

14:20 family like ACX

14:22 7024 um all of these uh will now uh

14:26 customers can use them on this very

14:28 familiar mist Qi what the use case that

14:31 we will be supporting primarily is

14:33 around the observability and insights

14:35 like uh you know Bob already mentioned

14:38 uh from here we're going to take it uh

14:40 towards delivering service experience

14:43 level metrics and really being able to

14:45 pinpoint you know um where the customer

14:48 experience is breaking and why it is

14:50 breaking and from there of course you

14:52 know we're going to leverage all the

14:54 power uh of Marvis AI Ops to solve W

14:57 routing problems for for these

15:00 Enterprises got it and and how do you

15:03 expect some of these early deployments

15:05 what type of benefits do you expect

15:07 those organizations to achieve by

15:08 deploying this have you spoken to any of

15:10 your early adopters and and what they're

15:12 seeing absolutely great question so let

15:15 me actually step back and break it down

15:17 uh from sort of where AI starts making

15:20 sense in the WAN network right the very

15:23 first step is to be able to spot

15:25 anomalies right uh AI models will be

15:28 able to tell for example oh there is a

15:30 service degradation somewhere right and

15:33 then step two is you start correlating

15:36 and now ai tells you I correlated it and

15:39 I actually saw service degration not

15:41 just you know where you saw the alarm

15:42 but I saw it in four other places it's

15:44 affecting 20 customers and by the way

15:48 after correlation turns out that this is

15:50 related to a particular link uh that is

15:53 you know servicing then AI comes in and

15:57 does a Diagnostics again some models and

15:59 turns out well there was some software

16:01 upgrade that happened and the

16:03 configuration was changed and that's the

16:05 root cause for this service degradation

16:08 and then the last part is the actions

16:10 where it would recommend in this

16:12 scenario fix the config and here is the

16:14 few lines of code that you need to push

16:17 would you like me to do it or do you

16:19 want to do a human assisted action here

16:21 right so when you look at the whole end

16:23 to end life cycle of where AI plays a

16:26 role it's really about day two

16:27 operations help helping customers spot

16:31 problems much faster so reducing that

16:33 meantime to

16:35 know helping customers diagnose and find

16:39 the accurate root root cause much faster

16:42 and much higher degree of accuracy and

16:45 then actually taking actions to close

16:47 the loop and fix the problem and do it

16:49 in much faster than humans much

16:51 accurately and over the course of time

16:53 learn what what those problems are and

16:56 actually you know make this whole

16:59 seamless towards a self-driving Network

17:01 so that's what the benefits we are

17:03 hoping to deliver to our Enterprise

17:05 customers uh uh through the Miss

17:08 platform yeah I think that's great and

17:10 one of the things you mentioned I just

17:11 want to touch upon for everyone who's

17:13 watching the closed loop system that you

17:15 have enabled in your solution I think is

17:18 a great way to make sure that humans are

17:21 an integral part of the AI and

17:23 essentially as you're going through and

17:25 doing this right it gives you the

17:26 ability to say yes this was right no it

17:28 wasn't provid feedback be able to

17:30 improve the solution be able to leverage

17:32 the knowledge that you have of your

17:33 individual Network to provide feedback

17:35 back to Juniper so it's constantly

17:37 improving so that's just a little

17:39 takeaway that I've I've learned from

17:40 working with you over the time and I

17:42 think it's an important part for helping

17:43 to adopt AI in in a faster time frame so

17:48 that's wanted to switch gears a little

17:50 bit earlier we were talking about that

17:52 sustainability aspect and I've been to a

17:55 lot of shows this this spring clearly AI

17:58 comes up and when I say AI referring to

18:00 those gen AI environments right those

18:02 large language models and the one thing

18:03 that's really clear is it's going to

18:05 consume a ton of power and so you've

18:08 talked about what you can do so clearly

18:10 organizations you know there a lot of

18:12 these data center environments looking

18:13 at are we going to be limited by the

18:15 power we have so every little bit of

18:17 power that can be saved is going to be

18:18 important I wonder if you could focus a

18:20 little bit on how your helping those

18:23 organizations to free up power in their

18:26 data centers in your Edge router

18:28 environment absolutely I would say this

18:30 is one of the most important missions

18:32 where we are investing and byy

18:34 intentionally designing our products to

18:36 meet sustainability goals um let me

18:39 break it up into few Dimensions at the

18:42 very core of it you know the way we

18:44 design the Silicon uh itself it means uh

18:47 with every new generation we can make it

18:49 much much more power efficient for

18:51 example our latest silicon uh is 77% uh

18:55 better uh lower power consumption around

18:58 65 % you know lower space required as

19:01 you build modular or fixed platforms

19:04 around it right um so the Silicon has a

19:06 big role to play second part is the

19:08 design itself that's where the uh

19:11 platform and the form factor and the

19:14 footprint all of these can contribute

19:16 you know to again the total cost uh

19:18 savings and the overall energy

19:20 consumption the other part is designing

19:24 platforms for the long run we call this

19:26 longevity you don't want to be investing

19:28 in platforms that you have to rep

19:30 replace every six years or five years

19:32 right so at Juniper we are very proud

19:34 even of our uh long-standing MX platform

19:37 that is in some networks been there for

19:39 15 plus years um so you know for us

19:42 really building that two times the cost

19:45 Savings in terms of sustainable power

19:47 efficiency space efficiency and two

19:50 times the life cycle and then the third

19:53 dimension is at in the operation stage

19:56 right how automation can help deliver

19:58 these sustainability goals so there our

20:00 thought processes you know not only can

20:02 we help identify where the network not

20:05 being used for example some links at

20:07 night are not getting enough uh capacity

20:10 can we turn them down dynamically bring

20:12 them up uh back up in the morning um

20:16 similarly can we Route traffic to a more

20:19 cost efficient path so these are all

20:21 some uh Innovative angles that we are

20:24 approaching sustainability with both

20:26 starting from the core of the Silicon to

20:28 to the system design to the whole

20:30 operation uh planed with

20:33 automation excellent that's great I

20:36 really I it's it's really going to be

20:37 important moving forward that

20:39 organizations continue to strive to

20:40 drive that sustainability and reduce

20:42 their power consumption and look for

20:43 ways to drive additional efficiencies

20:45 clearly you're taking a lot of steps to

20:47 get organizations and help them get

20:49 there so as we think about wrapping up

20:52 here um let's talk a little bit more

20:55 about the long-term evolution of mist

20:57 and how Jun uner AI will really help

21:00 deliver that self-driving network from

21:02 an endtoend perspective so Bob maybe you

21:05 could maybe you could give all the

21:06 viewers your take on how you see Miss

21:09 continuing to evolve yeah I see a couple

21:11 of vectors here on where Miss is going

21:13 to be involving Cloud aops uh the first

21:15 is around gen Ai and llm right we all

21:18 saw what happened when chat GPT came to

21:20 Market 2021 2022 um I personally believe

21:25 and I believed it since we started miss

21:26 that conversational interfaces are going

21:28 to be the next user interface you know

21:31 for networking and for other verticals

21:33 so I think we're going to start to see

21:35 that technology start to evolve more

21:38 we're already starting to see do magical

21:39 things uh you know at Miss Marvis we

21:42 started with a 2018 with natural

21:45 language understanding you know what

21:46 chat G PT really brings is a voice to

21:48 Marvis now so we're going to see that

21:51 extend over time I think the other thing

21:53 what we're going to see is these deep

21:55 learning models what we're seeing with

21:56 zoom and teams you know the same

21:59 transformation we saw them do inside of

22:01 the language space you know we're

22:02 starting to see that in networking right

22:04 models that can actually accurately

22:06 predict a user's performance that's what

22:08 leads to actually be able to get to

22:10 predicting and get to cause the problems

22:13 so if I look forward in the future you

22:15 know natural language is going to be a

22:16 big part of it deep learning is going to

22:18 be the technology that starts to disrupt

22:21 the networking space and bring more

22:22 functionality into Marvis and

22:26 networking sounds like an exciting

22:28 future I'm looking forward to seeing how

22:30 it all all comes out listen that's all

22:33 the time we have for today so thank you

22:35 for watching extending AI to Enterprise

22:38 routing um on the cube research for more

22:41 information on Juniper's AI native

22:43 platform and routing Assurance Solutions

22:45 please visit the Juniper website

22:54 [Music]