Networking for Change - The Future of Data Centers

0:01 [Music]

0:08 businesses are transforming at record

0:10 speeds and connecting centers of data is

0:12 fundamental to success

0:14 gartner recently said that businesses

0:16 are three times more likely to fail in

0:18 their digital transformation initiatives

0:20 if they don't transform the network

0:22 first

0:23 i'm emily wong and i've brought together

0:25 some brilliant colleagues for discussion

0:27 on the future of data centers

0:29 gentlemen would you like to introduce

0:30 yourself

0:31 i'm onsu karam

0:33 previously

0:34 founder and ceo of abstra currently vp

0:37 products at juniper

0:39 and my name's nick davey i'm the product

0:41 lead for control networking here at

0:42 juniper uh scott sneden i lead the data

0:45 center specialist team at juniper

0:47 networks great it's amazing how we have

0:49 a chance to be here in person

0:51 in this special times right so to kick

0:53 things off i'm sure you have a lot of

0:56 conversations with your customers can

0:58 you share what are their top of mind

0:59 what are their challenges these days

1:03 um

1:04 well you know the there are a lot of

1:06 things that are that are

1:08 kind of

1:09 driving the conversation in a variety of

1:11 ways and it's always different with

1:13 every customer we've talked to

1:14 but in your intro you talked about

1:16 digital digital transformation and sort

1:18 of transforming the network that that

1:20 comes to us in a lot of different

1:21 flavors you know there's

1:23 supply chain issues that are kind of

1:25 forcing customers to think about

1:26 multi-vendor in new ways there's the

1:29 great resignation where people are

1:30 scared of losing their skills and talent

1:32 and

1:33 the single source of truth that exists

1:35 in the mind of the engineer all of that

1:37 is is

1:38 a big worry um but really

1:41 the the lead thought is trying to think

1:44 of ways to make the network contribute

1:46 to the business instead of being an

1:48 afterthought that is just an expense to

1:50 the business so that especially with

1:52 enterprise is sort of a shift that's

1:54 happening that mindset's existed in

1:56 service provider for quite a while

1:58 because when a service provider thinks

1:59 about the network the network is their

2:01 product and so of course the network is

2:03 front of mind

2:04 in their business thinking but for

2:06 enterprise the network's always been

2:08 an afterthought and expense something

2:10 that i have to band-aid um but you know

2:13 the big shift that seems to be happening

2:15 our customers are really starting to

2:16 think more forward about the network as

2:18 being a contributor to the business yeah

2:20 correct i mean think think of everything

2:23 we do today

2:24 we utilize the network as its foundation

2:27 whether we're talking to uh

2:29 to our family you know

2:31 over geographical distance or if uh

2:34 we're shopping online or if we're doing

2:37 a video conference um

2:39 that's this is why i like this garner

2:41 quote right

2:43 every business is digitally transforming

2:45 and you can't digitally transform if you

2:47 don't transform your network first

2:49 so the net network has become this

2:50 critical foundation

2:52 for everything we're doing

2:54 and when you think of how fast we're

2:56 going

2:57 you know the the challenge is how to

3:01 scale the network how to operate the

3:03 network at scale

3:04 at the speed of the business and doing

3:06 it reliably yeah right reliability is

3:09 key here you know there was an ad from

3:11 the 80s

3:13 power is nothing without control and

3:15 this is living up to the expectations of

3:17 customers as well

3:19 the the migration to public cloud and

3:20 hyperscalers has changed the expectation

3:23 about infrastructure provisioning like

3:24 you said it has to move at the speed of

3:26 the business people aren't willing to

3:28 wait anymore for provisioning or

3:29 firewall changes their expectation is

3:32 they can define their applications click

3:34 the button and then have those

3:35 applications launch and be ready for uh

3:37 to use correct

3:39 and like you mentioned a lot of the

3:41 activities have gone online right from

3:43 working remotely shopping kids taking

3:45 classes at home how can it teams ensure

3:48 this smooth customer experience

3:51 the um

3:52 i mean nick touched on one about the

3:55 expectation of of

3:57 infrastructure being delivered instantly

4:00 that's become the norm for the

4:02 application teams

4:03 your application teams can click get

4:06 their machine and they're up and running

4:07 and their app has a place to live the

4:09 network teams haven't quite followed in

4:11 that same motion or at least they're

4:13 they're oftentimes just playing catch-up

4:14 trying to accommodate that motion or

4:16 that activity um and and so

4:20 what we really want to see our network

4:22 customers doing is is thinking about

4:24 delivering a network as a service or

4:26 delivering what they offer as an

4:28 on-demand service you know and if if

4:31 they can integrate directly into the

4:32 tools that those server teams use

4:35 all the better so automation is

4:37 absolutely the key but

4:40 you know to some answers point about

4:41 going fast just because you can strap a

4:44 rocket ship to your car doesn't mean you

4:45 should um

4:47 you know so you need to be able to be

4:49 fast and deliver those services quickly

4:51 and on demand but you have to do it

4:52 right if you make a mistake you've

4:55 slowed everything down or even worse

4:57 exposed yourself to a security flaw that

4:59 you may have never seen coming

5:01 you know and again

5:02 to scott's point

5:04 you can't you know security is top of

5:06 mind for everybody

5:08 it used to be that well you know we we

5:10 we enter commands manually and then that

5:12 operator leaves and then the next

5:13 operator comes in it's like i wonder why

5:16 these commands are there what's the why

5:17 why why is the network configured the

5:20 way it is

5:21 you can't have that today and you know

5:23 you need to know why network is

5:25 configured the way it is

5:26 you need to know that your network has

5:29 the right security posture otherwise

5:31 you're exposing yourself to security

5:33 vulnerabilities which can be

5:34 catastrophic

5:36 yeah but that goes back to the way that

5:37 we were creating networks before it was

5:39 artisanally crafted cli we were

5:41 sculpting uh yes creation exactly

5:44 exactly

5:45 so

5:46 um we can't do that anymore because it's

5:48 not uh like the network and the

5:50 infrastructure is not ours to sculpt and

5:52 control we are exposing that now through

5:54 a set of interfaces to our application

5:56 owners and the the folks who depend on

5:57 that infrastructure so if you're going

5:59 to put power into the hands of the users

6:02 there needs to be like you say control

6:04 around what you're exposing and there

6:05 needs to be um like an architecture

6:08 templates and frameworks that you use to

6:09 make sure that the correct config gets

6:11 applied every single time

6:13 and is there a way we can do a control z

6:17 if oopsy sometimes happens that's a

6:19 that's a that's a really good point you

6:20 know when you talk to network operators

6:22 and they have ptsd they call it the like

6:25 you know the pinky ptsd which is that

6:27 they click on enter

6:29 and then

6:30 like the whole network

6:33 freezes well that's the most terrifying

6:35 thing that you can hear in any operation

6:37 center is a space bar followed by three

6:39 other space bars louder and louder that

6:41 means that you have a problem

6:43 and so yeah rollback is the most

6:45 important consideration when deploying

6:47 um especially when uh when you're uh

6:50 handing the control over to to users and

6:52 application owners they need to have a

6:54 safe and reliable way to roll back to a

6:56 known good state as well yeah and who

6:58 doesn't love that it's like if i make a

7:00 mistake i don't have to worry about

7:03 getting yelled at i can quickly just fix

7:05 it before anyone can notice that i

7:07 actually made a mistake then that's

7:09 great it's like in the plane you have so

7:11 many layers of operating a plane of

7:13 safety right so like the software needs

7:16 to ensure that every step that

7:17 everything you're doing is reliable and

7:20 at the very end even when you

7:22 commit all of it then you have a way to

7:24 get out of it right so like this is what

7:26 you need to think about having all these

7:28 layers of safety yes absolutely so we

7:31 touched a little bit on security but

7:33 with the current geopolitical

7:35 environment we're living in

7:37 um what are some suggestions for dc

7:40 teams and security teams

7:42 i

7:43 you've gotta you've gotta kind of

7:45 approach security a little more

7:46 holistically than we ever did before i

7:48 you know i can't tell you how many times

7:50 and nick you've seen this in kubernetes

7:52 deployments and things where you you

7:54 have an application environment that's

7:55 very very dynamic and changing and since

7:58 the network team and the security team

8:01 don't really have the facility to change

8:03 at the same pace they'll compromise and

8:05 they'll say well okay in your little

8:07 corner over there we'll just leave that

8:09 wide open and you can do whatever you

8:10 want to

8:11 and maybe you build a perimeter or a

8:13 little moat around that castle but if

8:15 something happens inside that castle all

8:18 bets are off and and so you've got to

8:19 approach security

8:21 in the same way that the server teams

8:23 are deploying it it's got to be

8:25 very segmented it's got to be um

8:28 isolating

8:29 like things from other things that

8:31 aren't like things

8:32 but you've also got to make it

8:34 it's got to have the ability to change

8:36 and be dynamic you can't let a network

8:39 engineer just shortcut a security policy

8:42 by saying okay well over there i'm going

8:43 to let them do what they need to do

8:45 because i can't keep up with what

8:46 they're doing you've got to be able to

8:47 keep up with what they're doing so

8:49 infrastructure in the past has always

8:50 facilitated the app owner right so we

8:52 try and make as permissive a set of

8:54 rules as possible so that the

8:56 application works but not too permissive

8:58 such that we let the bad guy in

8:59 and i think the issue comes or the the

9:02 the friction comes from the interactions

9:04 between those two teams so one of the

9:06 best ways to address security is to

9:08 embed it with the app itself right the

9:10 folks who understand the flows and

9:12 mechanics of all of those micro services

9:14 running around the cluster are the

9:15 application owners themselves and so

9:18 security needs to be another resource

9:20 defined with the application just like

9:22 storage just like load balancing just

9:24 like network segmentation security has

9:25 to be embedded into the application

9:28 manifests and actioned directly at the

9:30 point where the application touches the

9:32 network and one of the biggest things we

9:34 can do to help facilitate that is to

9:36 give that network engineer the gift of

9:38 time

9:39 you know good

9:40 powerful automation that makes their

9:42 day-to-day easier lets them have a seat

9:45 have the time to have a seat at the

9:47 table with those application owners to

9:49 make sure that it's designed in from the

9:50 beginning so

9:52 you know we we all talk about automation

9:54 might eliminate jobs no automation is

9:57 going to give those engineers the

9:58 ability to participate and to contribute

10:01 at the front and not just be putting out

10:03 fires at the end or trying to find that

10:06 security flaw that was introduced

10:08 because they didn't do security

10:10 proactive instead of reactive exactly

10:12 right right i mean the only thing i'll

10:15 add

10:15 is that

10:17 you want a a foundation which which

10:20 allows for that level of security you

10:22 know we talk about the single source of

10:24 truth you know the knowledge of what's

10:25 in your network right the knowledge of

10:27 why the network is configured

10:29 the way it is

10:31 the ability to react quickly you know by

10:35 deploying the right security policy

10:37 again you can't do that without having

10:40 the proper layers of software the human

10:42 brain cannot make up cannot just uh have

10:46 doesn't have the ability to do it you

10:48 know at the scale that we need to with

10:50 all of the parameters that one has to

10:52 take into account yeah but that's what

10:53 that's waterfall security that we've

10:55 been practicing in the past which is

10:56 like we get the app ready then we get

10:58 the network ready then we get the

10:59 firewalls ready and then we launch the

11:00 app and that works as long as you can

11:02 execute the whole pipeline every time

11:04 but we don't need to run that as a

11:06 pipeline anymore right you can bundle

11:08 together the notion of connectivity and

11:10 security and put that control into the

11:12 hands of the users if you had a set of

11:14 tools that that gives users

11:17 the the control over firewall policies

11:19 applications at a level that they

11:21 understand

11:22 correct yeah

11:23 now we have so much activity going on

11:25 with data centers sustainability is

11:28 another business requirement so what can

11:31 modern data centers do

11:33 in that respect you know when we think

11:35 about sustainability we also want to

11:37 think about like the power consumption

11:39 of the device that's going in and that's

11:41 all very very important you know we're

11:43 not hardware guys to have that

11:45 conversation but what we can do is say

11:47 listen what we can do is allow you to

11:49 design a more efficient net

11:50 infrastructure choose the right hardware

11:53 for the job if it's a server that needs

11:55 to process that packet use a server to

11:58 process that packet if it's a

11:59 purpose-built piece of network hardware

12:01 that has 400 gig interfaces use that

12:04 device

12:05 but the experience is going to be

12:07 consistent and and the

12:09 process to deploy things is going to be

12:11 consistent across all of those

12:12 absolutely and that way teams can lead

12:14 with innovation instead of being

12:16 constrained by this you know the

12:19 logistical stuff that's right yeah and

12:21 maybe what uh what i'll add is you know

12:22 to me i i was always surprised with how

12:25 many organizations

12:27 avoid upgrading their infrastructures

12:30 because they're worried of causing an

12:32 outage or causing a disruption or they

12:34 don't have a person that knows the cli

12:36 or they don't know the resistance

12:39 they're left with these 10 year old

12:41 devices you know big honking switches

12:44 which today can be replaced with like

12:46 literally one new switches every three

12:48 years you you have the devices that

12:50 deliver the same performance for three

12:52 times less power and three times less

12:54 footprint right so if these same

12:58 softwares that help you automate will

12:59 help you migrate

13:01 right from the old to the new and it's

13:04 that much

13:05 it's that much that's that is very

13:07 helpful to sustainability you know to

13:09 have the ability to upgrade to the new

13:12 so i think that that's that's one area

13:14 the other area that comes to mind is

13:16 that as data centers are getting more

13:17 and more distributed

13:19 flying folks everywhere to go and manage

13:22 their networks you know can have a big

13:24 impact on sustainability rather than

13:27 having the ability to remotely

13:29 operate these networks through

13:31 automation so you have one single point

13:34 of control that from which you can

13:37 manage you know the networks that are

13:41 distributed across many many different

13:43 geographical areas you know you can see

13:45 that having an impact on sustainability

13:48 as well yeah and

13:51 that distribution of compute and network

13:53 is not possible without higher level

13:55 tools that let you kind of stretch your

13:58 operational control because if i'm

14:00 moving my data center into 100 small

14:02 facilities i can't manage that like 100

14:04 small data centers i have to be able to

14:06 have a high level interface that

14:08 reflects the entire state of my

14:10 footprint and then manage it from that

14:12 central point and then like scott was

14:14 saying let the apps teams decide the

14:17 best spot to run the workload

14:19 yeah exactly right so with all these

14:22 technologies uh evolving inside the data

14:25 center realm

14:26 what's

14:27 the next hot thing what excites you

14:30 about the future and what what new

14:32 innovations you think will come or

14:34 you may be working on

14:36 what i find exciting is

14:38 uh the combination of these various

14:40 technologies that we can leverage to to

14:43 improve the life of the operator right

14:44 you know when we talk about

14:46 uh experience first networking it's

14:48 about

14:49 improving the life of the operator and

14:51 then ultimately giving a better

14:52 experience to the user right

14:54 and there are

14:56 so many really cool technologies that we

14:58 can bring together

15:00 beyond just our own our areas and you

15:02 know there is a lot of work going on at

15:04 juniper to kind of bring in these

15:06 these solutions together the best of ai

15:08 the best of machine learning the best of

15:10 distributed systems the best of cloud

15:12 technologies in order to really deliver

15:14 on on that mission

15:16 yeah and i mean from my perspective uh

15:19 it's really exciting watching uh the

15:21 cloud native networking space evolve and

15:24 mature and it's

15:26 really exciting as well to bring some of

15:28 the the tools and technologies we've had

15:30 in traditional networking into the cloud

15:32 native space to show folks how they can

15:34 run their brand new applications with

15:36 the same sets of tools topologies and

15:39 and security that they've uh that

15:41 they're used to in private cloud

15:44 and with that i think this brings to the

15:47 end to today's episode time always feels

15:49 short when we're having a good time but

15:51 we can continue the conversation by

15:54 going to juniper.net slash data center

15:56 or follow us on twitter thank you so

15:58 much for being here today thank you

16:01 [Music]

16:09 you