The Cloud-Ready Data Center

Introduction

0:00 [Music]

0:10 thank you

0:13 so following Sadir is always an

0:15 experience I feel a little bit uh

0:18 marketing kind of didn't set me up right

0:20 they said business casual Studio doesn't

0:22 even have shoes on

0:24 um so we'll just

0:25 kick the shoes off does that give you

0:28 more technical street cred with this

0:29 audience I got to go sockless I will

0:31 it's a little gross

0:33 um my name is Mike bushong I I run the

0:36 data center business here at Juniper

0:37 Networks we talk a lot about AI in

0:39 action we've talked a lot about the

0:41 campus we've talked a lot about the

0:43 branch we just got into the wired side

0:45 it's not AI in campus it's AI in action

0:48 we've got to carry that over to the data

0:49 center I want to talk a bit about how we

0:52 do that

0:54 so

0:55 it's language we've always done it that

0:57 way is it's some of the most pernicious

0:59 language in it organizations anywhere on

1:02 the planet right every time you want to

1:04 go and make some change right we know

1:05 that technology is not stagnant it

1:07 doesn't move at the pace of whoever the

1:09 incumbent is things come along now and

1:11 again a little bit like X-Men style a

1:13 mutation comes in and then leaps us

1:15 forward a couple of years

1:16 when that happens collectively and as

1:19 individuals we have to have the courage

1:21 to do something different now many of

1:23 you are using myths today

1:24 and in that missed decision you took a

1:27 risk you said I'm going to try something

1:29 that's new and that's not just company

1:31 risk that's personal reputational risk

1:34 now there's a saying in our industry

1:36 right no one gets fired for buying IBM

1:39 in the networking space it gets a little

1:40 bit grosser no one gets fired for buying

1:42 Cisco but I'll tell you what you don't

1:44 get promoted for buying Cisco either

1:46 we've got to have the ability to go and

1:49 make change what I want to do is argue

1:51 the case for why what we're doing in the

1:53 campus side and how we're transforming

1:55 with AI

1:56 is going to carry its way into the data

1:58 center so if you like what you see on

2:00 the campus then you have to have a

2:02 conversation with juniper about data

2:03 center

Juniper’s Data Center Business Growth

2:05 now we have a data center portfolio that

2:07 Services the service provider space

2:09 we've got customers in the cloud space

2:11 we've got the Enterprise covered and

2:13 these you know we feature the largest

2:16 companies across the globe in all each

2:18 of these markets

2:20 now that portfolio it's based on

2:22 primarily broadcom silicon the takeaway

2:25 here is if you're familiar with like

2:27 switches and common form factors just

2:29 presume that we have a familiar form

2:31 factor that does what you need

2:34 but of course data centers aren't made

2:35 with switches they're made with Fabrics

2:37 they include things like management and

2:39 operations and so from that Foundation

2:41 we've built software on top that allows

2:44 us to change the way operations happens

2:46 in Data Center

2:49 we acquired a company a couple years ago

2:51 called appstra and in the same way that

2:53 mist has transformed the wireless space

2:55 and what we're doing in the campus side

2:57 abstra has transformed what we're doing

2:59 on the data center side you see the

3:01 numbers here we've doubled our data

3:03 center business in the last two years

3:04 now if you're sitting there who cares

3:06 about whether we've doubled the business

3:07 that's a great Juniper thing what it

3:09 signals is that something is going on

3:12 there is a change that is afoot and that

3:15 change means that there is a better way

3:17 of doing things and people are betting

3:19 with their wallets to take advantage of

3:20 it

3:22 now there's another event that's going

3:23 on right now right there's like a Cisco

Cisco’s Customers Struggle with Complexity

3:25 live or something everyone's afraid to

3:26 say the Cisco word for some reason I

3:28 didn't get the memo Cisco there was

3:30 column so Cisco's market share this is

3:33 what negative seven percent keger looks

3:35 like

3:35 don't mention this because I want to

3:37 take take shots at Cisco's market share

3:39 the point is that it isn't the market

3:41 Rose and so Juniper doubled our business

3:43 because you know a rising tide floats

3:45 all boats now there's a sinking ship

3:48 and as they go down what this means is

3:50 that people are transitioning from

3:52 incumbent ways of doing things and the

3:54 reason they're doing that isn't to eke

3:56 out incremental advantages in pricing

3:59 but we've talked about all day today

4:02 this is about fundamentally changing

4:04 what the point of interaction is and

4:06 what's your what you're capable of from

4:08 an operations perspective

4:10 and that same transformation it's

4:11 happening on the data center side

4:13 and frankly Cisco's loss is our gain

4:16 so why is that happening

4:19 look I'll be honest all right Cisco's

4:21 gone through some execution challenges I

4:23 don't think we have to pick on that in

4:25 2013 they talked about ACI ACI was

4:28 actually and I'll be honest I was like

4:30 at an sdn company I was there because I

4:32 was going to make sdn millions I drive a

4:34 Hyundai it didn't work out

4:36 um

4:37 yeah it's too soon it still hurts my

4:39 wife is pissed

4:41 um

4:42 moment of silence hold on

4:47 stings

4:48 um

4:49 when I saw ACI the idea that you could

4:52 converge operations around applications

4:53 in data center that's actually a

4:55 profound thing to go do they were trying

4:57 to make a difference in the operations

4:58 side the challenge

5:00 was that the solution fell over under

5:02 the complexity that they were

5:03 introducing to make it work

5:05 a bunch of people went out and bought a

5:07 bunch of Nexus switches tried to deploy

5:09 ACI and they said it doesn't quite work

5:10 that way

5:12 here's a Professional Services contract

5:15 see how that works

5:16 no it's not free we'll charge you for

5:18 that every year

5:19 and so the the as that fell over then

Changes in the Network Engineering Workforce

5:22 certainly Cisco was going to have

5:24 challenges but if we think that the

5:26 transformation that is going on in the

5:27 industry is limited to execution

5:29 challenges I don't think we're doing the

5:30 industry Justice

5:32 and frankly since I have a bunch of

5:33 friends at Cisco so I make fun of a

5:35 little bit but it's a bunch of smart

5:36 people that work there they're not dumb

5:38 it's not purely an execution challenge

5:40 the bigger issue the Strategic issue

5:42 that is at its core is a change in the

5:44 workforce altogether

5:46 now take Network Engineers right I'm not

5:49 a network engineer but I used to be a

5:50 trainer so I guess I'm kind of close

5:52 Network Engineers sort of entered the

5:54 market in what like mid 90s early 2000s

5:56 as the internet stuff was going on

5:58 if we're being honest Network Engineers

6:01 we're kind of Aging out a little bit

6:03 right not not me but like you all

6:07 none of you none of you right people

6:09 over at Cisco alive are aging out

6:13 um but as they age out right and then

6:15 there's like the great resignation or

6:16 the great reshuffling call it whatever

6:18 you want right there's a change in

6:19 Workforce the rise of remote work means

6:21 that people can work for Amazon without

6:23 living in Seattle so that changes stuff

6:25 up some people are just making work-life

6:27 balance because of the or decisions

6:28 because of the pandemic whatever the

6:30 reason there's a wholesale change in the

6:33 network engineering Workforce now this

6:35 is terrifying because the bumper crop of

6:37 network engineers

6:39 it's not the same group that came up

6:41 before they didn't major in networking

6:42 and do four years at some major

6:44 university they're not trained on Cisco

6:46 press and and they don't have like rote

6:48 memorization of you know vendor-specific

6:50 syntax that you can use and some

6:52 certification that says I memorized the

6:54 commands they don't have all of that

6:57 the next crop of network Engineers

6:59 they're learning on AWS and Azure and

7:02 gcp which means the point of interaction

7:05 the point of familiarity isn't the CLI

7:10 it's now going to be the way we consume

7:11 Cloud it's probably closer to terraform

7:13 than it is to Cisco CLI

7:15 and what that does is that means that as

7:17 you try to fill your Workforce

7:20 with people who don't have as much

7:22 experience and haven't gone through the

7:24 memorization process then your mops and

7:27 stops don't work

7:29 and when that happens

7:32 you either break everything or you

7:34 introduce a new way of operating the

7:36 reason we've been winning

7:38 is because we introduced a new way of

7:39 operating

The Network Automation Journey

7:41 now Network automation has been a thing

7:43 for I've been in the network automation

7:45 space for like 15 years is that right oh

7:47 my God it's 15 years in network

7:48 Automation and if you look at the

7:50 millions of people hours that we've

7:52 poured into Network automation

7:54 like we are literally nowhere it's

7:56 embarrassing

7:58 and we have to rationalize with all that

8:00 effort why is it that things are not

8:02 more automated

8:04 it's not because people are too dumb

8:06 it's because Network automation is too

8:08 hard so devops bore out if you're

8:10 particularly cynical right to make error

8:12 as human to propagate error to all

8:14 server and automatic ways devops

8:16 now if you don't believe that ask the

8:17 folks at Facebook and and tell them to

8:20 to recount the story when they had to

8:22 saw into their data center cages so they

8:24 could get access because their badges

8:25 didn't work

8:27 that's automation

8:29 right now the reason that stuff like

8:32 that happens and I love this other one

8:33 right we do this not because it is easy

8:36 but because we thought it would be easy

8:37 right automation doesn't make good

8:40 automation makes more if you've got good

8:42 it makes more good but if you have bad

8:46 it makes more bad and that's the way to

8:48 break things at scale well we have to

8:50 solve around operations and automation

8:52 we got to look at this differently right

8:54 the old approach isn't working

8:57 so what do we do

8:58 I'm putting up really pretty slides

9:00 right um okay so we acquired a company

The Experience-First Data Center

9:03 called abstra abstra is like the literal

9:05 inventors of intent-based networking

9:08 right and with Appster we say let's go

9:11 after day zero day one and day two but

9:13 what if we presume rather than thinking

9:16 that people are too slow

9:18 we think that networking is too hard

9:21 now to be clear networking should be

9:22 kind of hard right we do complex things

9:24 but it should be really hard to make

9:26 mistakes and if you start with a lens of

9:29 reliability

9:31 then having a different why not just

9:33 speed but reliability that drives a

9:36 different what

9:37 so I meet with customers all the time

9:39 and I talk about Network Automation and

9:40 their operational journey and big words

9:42 like transformation and I was meeting

9:44 with a large retailer in Europe they've

9:46 got a large e-commerce site I mean this

9:47 is like one of the the biggest

9:49 e-commerce sites in Europe and I asked

9:51 them I said how long does it take you to

9:52 turn up a new server I want you all to

9:54 think in your head how long that should

9:55 take

9:56 because I was probably thinking the same

9:58 thing you are a couple hours maybe a

10:00 couple minutes if they're really

10:01 operated if they're really automated you

10:03 know maybe a couple of days if they're

10:04 kind of slow yeah six to eight weeks

10:08 like what

10:09 how can that be and they're like well

10:11 because we have these blackout Windows

10:14 right so we want to make a change what

10:16 we do is we schedule the change it has

10:18 to queue up until there's a review

10:19 meeting they have a review meeting

10:20 there's always a take two on the review

10:21 meeting they go to the take two of the

10:23 review meeting they get to the scheduled

10:24 maintenance window the scheduled

10:26 maintenance window only succeeds about

10:27 50 of the time they go to the second

10:29 maintenance window and then they get

10:30 their servers up

10:32 now this is interesting because it means

10:34 the problem the speed problem is not how

10:36 long it takes to configure stuff I'm

10:38 going to give you some useful advice you

10:39 guys can write this down

10:41 if your biggest problem is how long it

10:43 takes people to key in stuff or if

10:44 that's if you're really trying to make

10:45 it faster and it's in keystrokes are the

10:47 biggest problem that you have and you're

10:49 worried about people fat fingering you

10:51 know in your in your networks and write

10:53 this down the best return on investments

10:56 you will get is send your team to a

10:58 typing class

10:59 like they will they will type faster and

11:01 they will make fewer mistakes and you'll

11:03 be good

11:04 if however the biggest problem you have

11:06 isn't the typing but the elapsed time

11:08 right it's the coordination between

11:09 organizations it's the coordination

11:11 between tools if that's the actual

11:13 problem

11:15 then there's other ways to do it if I

11:17 could guarantee that a change that you

11:19 were going to make in your data center

11:20 today if I could guarantee

11:21 mathematically

11:23 that it was going to work that it was

11:25 known good

11:27 the thing I would save is not the typing

11:29 time we can automate that too by the way

11:31 the thing I would save is all the review

11:34 time it takes to get that change to the

11:35 system

11:37 that's where the difference is going to

11:39 be made

11:41 automation is nowhere not because people

11:43 aren't capable it's because we've been

11:44 trying to solve the wrong damn problem

11:48 if you want to drive super fast on the

11:49 freeway go for it right the reason you

11:51 don't is because you're going to get

11:52 into accidents you want to automate

11:54 today you know code some stuff up the

11:56 reason you don't do it is because you're

11:57 going to get into accidents take that

11:59 problem away and you unlock speed

12:02 and when you do

12:04 you get real value that comes back in

12:07 return

12:08 now we do all of this through intent

12:10 based networking if you're not like an

12:11 intent based networking person let me

12:13 describe it fairly simply intent based

12:15 networking says I want the language of

12:17 networking to be what you want not how

12:20 to get it now think about how you get a

12:22 network today right you take hundreds or

12:24 thousands of distributed devices you set

12:26 hundreds or thousands of knobs on them

12:28 and when you get it just right you

12:30 slowly back away and you don't touch

12:32 anything

12:33 I started my career I did research for

12:36 Boeing I worked on jet engine design it

12:38 sounds really cool it's kind of boring

12:40 um

12:41 and I always joke I'm like if we built

12:42 airplanes or we build Networks

12:44 I would walk everywhere and I would

12:48 constantly be looking up for falling

12:51 airplanes right we need to change that

12:53 and that means flipping what we do now

12:55 when you do that when you work at the

12:57 level of what do you want not what not

12:59 how do the devices work then the tool

13:01 translates what you want into device

13:03 syntax into device config when we do

13:06 that by very nature the product is

13:08 multi-vendor we are the only

13:09 multi-vendor game in town

13:11 our lead customers are using us over the

13:13 top of ACI and over the top of Arista

13:16 and over the top of Dell and yes over

13:18 the top of juniper but by being

13:20 multi-vendor we give them a different

13:21 level of control

13:24 and then if you build a network and

13:25 you're not thinking about security I

13:27 don't know what you're doing we'll talk

13:29 a bit about security here in in on the

13:31 next session but in the security space

13:33 if we know what this what the network

13:35 looks like then what we can do is start

13:38 highlighting things like overlapping

13:40 security policies so you can flag if

13:42 there's going to be issues or likely

13:44 issues

13:45 and it's not enough to do this at Day

13:47 Zero what happens on day two

13:50 so we take that and then we make a model

13:52 of it we look at all the network State

13:54 and we say this is what a known good

13:56 Network looks like and when we push it

13:58 to production

14:00 then what we do is we say here's what's

14:02 actually in production and if those two

14:03 things are the same

14:05 we give you a big green thumbs up and if

14:07 they're different

14:09 we tell you what's wrong

14:11 what's the number one cause of downtime

14:12 right it's like fiber seeking backhoes

14:15 DNS probably the top two once you get

14:18 into the networking side it's human

14:19 error

14:20 what I want to do is provide an

14:22 environment where we drive human error

14:25 to zero

14:27 when we do this we see return you see

14:30 the numbers in terms of payback and Roi

14:33 it's not just about what do you get out

14:34 of the tool it's about what do you what

14:36 is required to make the network work day

14:39 after day after day and when you do this

14:41 you can start rationalizing tools and

14:43 getting rid of that

14:44 that allows you to streamline things it

14:46 lets you put people on a smaller number

14:48 of tools regardless of what their

14:50 experience is

14:51 if I'm not waiting for a ccie or some

14:54 certification to prove that somebody has

14:57 demonstrated they can do something then

15:00 I can start to hire and change my

15:01 Workforce up it means I can start moving

15:04 workloads around a bit so that different

15:05 people can do different things

15:07 and if I can do that over a multi-vendor

15:09 environment

15:10 then it means you can allow the physics

15:12 or the economics or maybe the features

15:14 to dictate your next purchase

The Proven Business Case of Juniper Apstra

15:18 and our customers they make those next

15:20 purchases I'll start I'm not going to go

15:23 through all these you can read them or

15:24 is the font big enough you can look at

15:25 it and see big white blocks if you're in

15:27 the back T Systems was our lead customer

15:29 right large data center provider in

15:31 Germany they actually came to us before

15:33 the acquisition they said you need to

15:34 acquire this company we're using it to

15:37 cap ACI we would like to build Juniper

15:39 but you have to have a management story

15:40 if you can unify that then you give us a

15:43 credible strategy to move from where we

15:45 are to where we want to be I said sold

15:47 let's go do that we did that and then

15:50 they and they were able to cap and grow

15:51 it gave them flexibility in their vendor

15:54 choice where their operations previously

15:55 didn't allow it

15:57 um the one in the the upper right that's

15:59 a rifison they're like a large a large

16:02 Bank

16:03 and Ernest at rifles and these guys he's

16:05 a great engineer listening to him he

16:07 gets so he gets super geeked out about

16:09 the technology and he will go for like

16:11 days on the technology bit but the party

16:13 was actually most excited about his

16:15 procurement team who he usually has to

16:17 fight with to get things through his

16:19 procurement team became fans of the

16:21 technology because it meant they could

16:23 put every RFP out to bid sincerely

16:27 and they got uh then they could put

16:29 things like lead times anyone dealing

16:31 with supply chain no one right that's

16:32 all solved

16:34 um they could deal with like lead times

16:35 they could put in uh pricing and then

16:37 they could force everyone to to bid

16:40 the part that's cool they're not on this

16:42 screen we have a Canadian uh service

16:44 provider that uses App Store over the

16:46 top they had an environment where we

16:48 couldn't meet their requirements on lead

16:50 times

16:51 and so they swapped to Dell I have a

16:53 happy Customer because they were able to

16:56 make a vendor Choice and then after that

16:58 then they're they're putting Juniper

17:00 back in and subsequent builds it gives

17:02 you freedom and flexibility

17:05 uh Tigo they have uh data centers in

17:09 like Columbia and Guatemala Belize

17:11 Honduras Panama they had issues because

17:15 they had an uneven Workforce people that

17:16 acted semi-autonomously but they needed

17:19 to deploy the same data centers

17:20 everywhere that showed up as like

17:22 quality problems for them because their

17:24 procedures and their architectures are

17:25 always like similar but kind of

17:27 different we rolled abstra in there

17:30 it solidified the workforce they're now

17:32 stamping that out in every location and

17:35 all the issues they had those issues are

17:37 gone because when you standardize the

17:39 architecture and you streamline the

17:40 testing what you get is better results

Intent-Based Networking & AI

17:45 now I got all the way through this and I

17:46 didn't even mention AI like so we'd have

17:49 to be idiots to look at all the stuff

17:52 that you see in Marvis and Ai and say

17:54 you know let's let's keep it in campus

17:56 of course we want to bring these things

17:58 together in a data center there's some

18:00 things that are deterministic they're

18:02 going to be you want them to be the same

18:03 way every time for that we use abstra

18:06 there's some things that are stochastic

18:08 probabilistic we things like

18:09 troubleshooting when we talk about you

18:11 know bringing the the tickets down to

18:13 zero and helping identify where the root

18:15 cause is figuring out where you know the

18:17 point of dmarc is the problem is going

18:19 to be somewhere between the access port

18:21 and and then DCI or it's over the WAN

18:23 link or whatever

18:24 bringing AI to bear on that problem

18:27 it supercharges differentiation that we

18:30 already have

18:31 the reason we're at AI in action is that

18:34 if we can take these things and bring

18:36 them together into a common uh into a

18:38 common platform

18:41 then the things that you're seeing on

18:42 the missed side

18:44 you start seeing those on the data

18:47 center side

18:48 we've been working at this for quarters

18:50 now and that bringing these things

18:53 together is the key to transforming

18:55 operations going forward

Demanding More from Your Network

18:58 and so I want to leave you with this

19:00 right when we talk about you know things

19:02 like TCO or we talk about things like

19:05 agility it's not only some advantage

19:08 that a cruise at the business level

19:11 right agility is and I have 11 year old

19:13 twin boys right for me doing things

19:16 simpler knowing that it's going to work

19:18 that means that

19:20 I can go to a soccer tournament on a

19:22 weekend because I'm not worried about

19:23 you know staying after if my production

19:25 change doesn't doesn't go the way it's

19:27 supposed to go it means I can spend an

19:29 extra dinner with my family where I

19:30 otherwise might have had to work

19:32 the differences that we make they're

19:33 more than just technical they're more

19:35 than just the percentages these are

19:36 personal differences

19:38 now I'm telling you that that's on the

19:39 horizon and whether you're in data

19:41 center or you merely work alongside

19:42 other teams that are in data center it's

19:44 incumbent upon you today to challenge

19:46 the status quo demand more from your

19:49 data center demand that when you go to

19:51 RFP

19:53 that anybody could legitimately bid for

19:55 it

19:56 I dare you to put out an RFP that any

20:00 supplier can bid and not be worried

20:01 about the operational implications

20:03 I dare you to hire a new college grad

20:06 because you shouldn't have to have a

20:07 damn certification in 25 years of

20:09 experience to provision a pod that's

20:11 already been deployed hundreds of times

20:13 in your environment

20:15 or better yet I dare you to push to

20:18 production on a Friday

20:19 I'm Mike Bouchon where's Juniper Data

20:22 Center and this is what AI in action

20:23 looks like in Data Center

20:31 now networking may start with switches

20:33 but it certainly includes security I

20:37 give you Kate Adams

20:40 [Music]