Latest Data Center Trends, Best Practices, and AI

0:00 hello everyone and thanks for joining us

0:02 today on our data center Round Table

0:04 with so many new technologies new

0:07 requirements and AI development going on

0:10 today we'll be discussing the best

0:12 practices in data center management to

0:14 help you prepare for the future and

0:17 lastly we will get a sneak peek on the

0:19 role of AI for data center because we

0:22 can't have a tech discussion these days

0:24 without talking about AI can we and I

0:26 still follow along we encourage you to

0:28 leave your own questions in the comments

0:30 section and we promise you that Scott

0:32 will answer all of them

0:33 before we dive into the discussion let's

0:36 give our panelists a chance to introduce

0:37 themselves Scott let's start with you

0:39 since you're going to be answering all

0:40 of the questions sure so my name is

0:42 Scott sneden I'm a data center

0:44 specialist here at Juniper focusing on

0:46 mostly our Enterprise Market but um just

0:49 about anywhere where we work with

0:51 customers on Data Center yeah and I'm

0:53 Claire Del quartz I'm from Belgium and

0:56 I've been living in the states for seven

0:58 years I'm a product manager for Network

1:01 automation for that much time too

1:04 and I'm product manager for abstra Samir

1:07 Parikh I'm the vice president of product

1:09 management for Juniper's Data Center

1:11 business focusing on app store and

1:13 Automation and a lot of our software

1:15 solutions that sort of provide the

1:17 operational foundation for our data

1:18 center offering great it's an honor to

1:21 have you guys here so with new trends

1:23 new technologies comes with new

1:24 challenges as well so maybe you could

1:26 share you know some of the challenges

1:28 that you have observed in the last year

1:29 and maybe even some interesting customer

1:31 stories so maybe I can talk about what

1:34 I'm currently hearing from customers is

1:37 how the data center network has not

1:40 become more simple on the contrary the

1:44 network has become more complex because

1:45 we're bringing in new technology in the

1:48 network to accommodate things like

1:51 virtual workloads continuous VMS and so

1:55 forth we want more security

1:56 we want more Mobility for the workloads

1:59 so you need to bring in order to

2:02 accommodate those new demands you need

2:05 to bring new technology in the data

2:06 center Network and it makes it a bit

2:08 more complex uh you know thinking about

2:10 technology like if you can fix Lan new

2:13 security layers so you lose a lot of you

2:16 add complexity you lose observability in

2:18 your network and it becomes harder and

2:21 harder yeah I totally agree with you I

2:23 I'm seeing exactly that with one

2:25 customer I'm working with currently um

2:28 this is an advanced customer they're a

2:30 smart team they know how technology

2:32 Works they've got some older Data Center

2:35 architectures and made the wise decision

2:37 to move forward to a data center fabric

2:39 using UDP and be excellent you know all

2:41 of the recommended architectures that we

2:42 talked about

2:44 but they tried to do it manually they

2:46 are the kind of guys that they're really

2:48 smart so they want to figure it out for

2:50 themselves and they read the guides and

2:52 and they developed their own

2:54 configurations and then moved forward

2:55 and it didn't go perfectly and they

2:58 introduced some errors during their

2:59 maintenance window that were fairly

3:00 difficult to recover from they missed a

3:03 few key things that were in the manuals

3:05 but maybe not totally obvious to find

3:07 right so yeah we talk about moving

3:10 technology forward and these

3:11 architectures that we've introduced are

3:13 a lot simpler compared to when I was

3:16 operating data centers but they're still

3:18 not completely simple the skills don't

3:20 always align to where we are with

3:22 technology today so yeah the challenges

3:25 are are even though people have agreed

3:28 to move forward on architectures the

3:30 skills don't always match and even in

3:32 the smartest environments mistakes can

3:34 still happen

3:35 yeah I mean it's a great point on skill

3:37 Scott I mean the thing that we're seeing

3:39 is you know Engineers coming out of

3:42 college today they're not being trained

3:43 on classic networking Technologies right

3:45 they're they're learning the ways of the

3:47 cloud right because that's the dominant

3:49 pattern that's where you know that's the

3:50 exciting pattern and so as you see the

3:52 networking industry getting older you

3:55 have this big skills Gap and so that's

3:57 why tooling and what we can do on

3:59 automation to sort of bridge that Gap is

4:01 super important because you know

4:03 otherwise we're going to run into this

4:05 perennial problem of how do we sort of

4:06 train up people to handle these

4:08 technology transitions because the

4:10 transition itself from an old

4:12 architecture to new architecture however

4:13 simpler it might be presents complexity

4:16 in itself for sure

4:17 so you you did touch on you know some

4:20 recommended you did mention some

4:21 recommendations on how to manage the

4:23 data center more efficiently but you

4:25 know maybe you can spend more you know

4:28 considering all these companies that our

4:31 customers are doing these days our

4:32 audience are dealing these days give us

4:34 some advice on how to manage it more

4:37 efficiently both on operational and

4:40 economical standpoint I think that's

4:42 that's a great question I'll touch on

4:44 most of the economics of it right I

4:45 think you know if it's top of mind for

4:47 all of our customers we look at the

4:48 macro environment and you know the key

4:50 element to sort of the economics of the

4:53 data center is using standards-based

4:54 interoperable and open technology so you

4:56 know what we see with our customers is

4:58 this move towards EVP and vxline fabrics

5:01 because it's an industry standard like

5:02 it operates across multiple vendors

5:04 there's test results all the vendors

5:06 participate in these testing forms to

5:08 demonstrate Improvement or operability

5:09 so that's a key foundation for sort of

5:11 how you deal with complexity at least on

5:13 the economic standpoint and you know on

5:15 the operation side I mean automation is

5:17 key but you know automation is its own

5:19 can of worms which I'm sure Scott you've

5:21 lived through you see a lot of customers

5:23 that customer I mentioned before is a

5:25 really good example of you know they

5:27 know networks and they know network

5:29 engineering and they

5:31 um expect that that they'll be able to

5:33 write their own automation really

5:35 efficiently and and yeah they have cost

5:37 constraints and so they didn't want to

5:39 pay for a tool from a vendor to build

5:42 this automation because they know they

5:43 can write that code themselves and in

5:46 the end maybe it might have been a bit

5:48 more expensive for them to automate

5:50 their way into a problem and then have

5:52 to recover from that so yeah really

5:55 considering the long-term economic

5:57 impact of your decisions is needed and

6:00 and taking those considerations very

6:02 very seriously is important in fact

6:05 talking about DIY we have a report

6:08 showing that people doing DIY end up you

6:11 know costing more than just acquiring

6:14 third-party information and there are

6:16 places that have done it very

6:17 successfully we see this with with the

6:20 cloud scale companies right they've they

6:22 can hire the best engineers write code

6:24 all day long and do that you know

6:26 they've figured out their economic model

6:28 that works for them but that model

6:29 doesn't work for everybody yeah

6:32 yeah I think another advice I could

6:34 probably give in time of you know

6:38 building this dream solution with

6:40 software

6:41 uh is take your time to evaluate the

6:44 software that are out there I think it

6:46 can be very overwhelming for customers

6:48 that don't have a foot in automation yet

6:51 to understand where to go do I do it

6:54 with myself do I go with a vendor and do

6:56 I trust a vendor that they're gonna give

6:58 me the best software for the book so

7:01 take your time to evaluate software

7:02 solution find something that's going to

7:05 work well with the skill sets you have

7:06 because every team every Enterprise has

7:09 a different type of skill set in the

7:10 team and

7:13 make sure you you you go with a solution

7:17 that is going to

7:18 get you where you want to be longer term

7:21 for your automation goal dream big you

7:24 know dream reliability dream automation

7:26 you know dream multi-cloud don't be

7:29 afraid to dream big and go with

7:32 software and then you can justify the

7:35 investment to your management much

7:36 easier yeah I mean I'm the skills thing

7:39 I think that a great example that we see

7:40 from customers were they succeeded if

7:42 you look at the hyperscalers they've

7:44 done a very good job of pairing Network

7:46 engineers and software Engineers right

7:48 because ultimately we think about people

7:50 think Network automation it's not the

7:52 hard part is not necessarily the

7:53 networking piece it's the software and

7:55 systems component of building an

7:57 automation system and this is where like

7:59 we encourage people to look at vendor

8:01 Solutions because you're getting

8:02 somebody with software expertise that

8:04 builds that for you but if you're going

8:05 to approach it in sort of your own on

8:07 your own and build it from other

8:09 components really look at how you pair

8:12 Network engineers and software Engineers

8:13 together to sort of solve that problem

8:14 because that context is important

8:16 talking about doing things easier and

8:19 more efficiently

8:21 um I've read in a report by Enterprise

8:22 strategy group saying that 75 of

8:26 Enterprises will keep most of their apps

8:29 in the private Cloud so cloud has been

8:32 popular for many years but you know how

8:35 come people seem to repatriate to

8:37 on-prem why is that and you know what is

8:40 your advice to our audience when it

8:42 comes to deciding whether they should go

8:44 on-prem

8:45 public or hybrid it comes back to

8:48 economics I think that's what we're

8:50 seeing you know the promise of cloud was

8:52 that everything would be easy and it

8:53 largely is and the promise of cloud is

8:55 that everything would be fast and you

8:57 can turn things on right away and do

8:58 everything you need on demand and it is

9:00 I mean those those promises all came

9:03 true

9:04 problem is that the cost can be

9:06 unpredictable and it's really difficult

9:08 for a business line to justify such a

9:11 variable cost when their revenue isn't

9:13 the same variability and and so the

9:16 economics are really dictating that

9:18 people work towards models that are more

9:20 financially predictable financially

9:22 manageable and so I I think the the now

9:26 and the future is is decidedly hybrid I

9:29 think the decision is not Cloud first

9:32 but Cloud right and and it's all Cloud

9:35 but which cloud is it that private

9:38 environment that's inside the walls of

9:39 your data center or is it some blend of

9:41 public and private and I think

9:44 I see more and more successful

9:45 businesses considering application by

9:48 application including considerations of

9:51 the skills that they have available to

9:53 support those applications and making

9:55 wise decisions as to where workloads are

9:57 going to run

9:58 yeah I mean I think one of the things on

10:01 the when you think about economics of

10:02 the cloud data becomes a key anchor

10:05 there because you know the data transfer

10:06 costs is like the most common complaint

10:08 you see users on Reddit of whatever like

10:11 they complain about the data transfer

10:12 costs and and so data gravity becomes a

10:15 big decision point for what you put in

10:17 the cloud versus what you keep in-house

10:18 and and then you also see a lot new a

10:21 lot of new regulations around data

10:23 sovereignty and privacy and that's

10:24 another Factor so I think Enterprises

10:26 are now looking at saying okay Cloud's

10:28 not this Panacea it's a very important

10:30 and valuable tool in our tool chain but

10:32 it's not like everything is fit for the

10:34 cloud we're going to make a more

10:35 thoughtful Choice around what makes

10:37 sense and what doesn't and we cannot

10:39 point out the problems without giving an

10:42 answer right so what does Juniper has to

10:44 offer to help people manage private data

10:46 centers professionally I I mean I I

10:48 think

10:49 I I said it's all Cloud but in reality

10:52 most private data centers are not that

10:54 cloud-like this is why developers still

10:56 like to go to the public Cloud as it is

10:57 easy and it is fast and they can do what

11:00 they want to when they want it without

11:02 having to go through a bunch of process

11:04 to get it done and and so I think the

11:07 onus is on the private Cloud operators

11:09 the data center operators and Enterprise

11:10 to present their environment like a

11:13 cloud to make things on demand to make

11:16 things

11:17 easily consumable easily configurable if

11:20 you've got a weight

11:22 15 minutes

11:24 let alone two weeks to get something

11:26 provisioned well that developer is just

11:28 going to go to put cloud in unless you

11:29 take away their credit card they're not

11:31 going to stop doing that

11:32 and so you've got to consider ways to

11:35 build private infrastructure and that's

11:36 where our focus is our our what we're

11:39 doing around optimizing operations

11:41 leveraging AI to make operations simpler

11:44 really the end goal is to just make it

11:47 easy to use a data center right and you

11:50 know that's where it becomes super

11:51 important for your network team to not

11:53 be composed of just networking experts

11:56 anymore and to Bringing expertise from

11:59 people that are software developer that

12:01 has developed automation solution

12:03 whatever they are and bring those games

12:05 in your team and try to rebuild the

12:07 Cloudland experience for your on-prem

12:09 data center and in terms of what we have

12:10 to offer I mean I think a foundation for

12:13 us is abstra right because it gives us

12:15 it provides our customers with the

12:17 ability to give their customers a

12:18 cloud-like experience right with the

12:20 work we've done around intent-based

12:22 Automation and you look at what we've

12:23 done around infrastructure as code

12:25 Frameworks with terraform right you're

12:27 giving now with apps or and our

12:29 customers have the ability to give their

12:31 customers a cloud API you know if you're

12:34 if you're an AWS user more likely you're

12:36 consuming that through terraform and

12:37 you're provisioning and destroying

12:38 infrastructure that way well we can give

12:40 you that same interface on top of App

12:42 Store like that becomes the foundation

12:43 for how customers can now you know have

12:47 that seamless experience with this

12:48 on-prem private cloud or public Cloud

12:50 because you give them a common tool

12:52 layer yeah that's pretty amazing and you

12:54 know as we talk about in 10 days and

12:56 make things easier for our customers

12:59 um I'm sure we have plans in the AI side

13:02 of the house right so you know how can

13:04 we walk away today without mentioning AI

13:06 so what what are the plans in the

13:08 Juniper data center group you know in

13:11 terms of Integrations with AI yeah I

13:13 mean that's a great question you're

13:14 right we can't we can't end a session

13:16 without talking about AI I think when we

13:18 think of AI there's two amazing sort of

13:21 ways AI is going to change the network

13:23 in the future right one is AI Ops and

13:26 the other one is sort of how you build

13:28 infrastructure optimized for AI

13:29 workloads so I'll talk a little bit

13:31 about AI Ops I mean obviously at Juniper

13:33 you know we are the leading AI engine we

13:36 have the leading AI Ops provider in the

13:38 in the world today with Marvis which is

13:40 you know today sort of the AI foundation

13:42 for our Enterprise business for campus

13:44 branch and wireless and we are

13:47 absolutely going to infuse that Marvis

13:48 engine into the data center so you'll

13:50 see very soon from us you know

13:51 discussions around how Marvis you know

13:54 and its AI Ops engine combined with App

13:56 Store and its intent engine so to truly

13:57 transform the way you build and operate

13:59 data centers so that's going to be the

14:01 key foundation for it and then you'll

14:03 also see a lot of work from us around

14:04 how do we help our customers build

14:06 cost-effective highly reliable scalable

14:08 infrastructure to support these next

14:10 generation AI workloads yeah I think

14:12 that ties back to what we're talking

14:14 about in the first question is there is

14:17 Network complexity it's harder to

14:19 observe traffic you know you can't just

14:21 get away with doing a packet capture on

14:24 every thing that turns into your network

14:25 it's too much too much traffic it's too

14:28 encrypted it's too many layers so what

14:30 you need is you need to be able to bring

14:33 up solution or mechanism to collect the

14:35 data process it so that it's ingestible

14:38 by a human again and the answer to that

14:40 is right now it's AI obviously

14:43 and then there's there's also another

14:45 conversation about the infrastructure

14:47 for AI so there's Ai and how that's

14:49 leveraged in Ops there's also the

14:51 conversation of how do we build

14:52 infrastructure to support all of this

14:54 new AI infrastructure and and we think

14:57 that that's a really exciting

14:58 opportunity for us there are a lot of

15:00 things on the horizon in Juniper around

15:02 very very high speed switching platform

15:05 we do think that that industry which is

15:07 largely in finiband today is Shifting

15:10 towards ethernet and we see a lot of

15:12 evidence and data around that and so you

15:15 know we've been a leading ethernet

15:17 switching provider for more than a

15:19 decade and we'll continue down that path

15:21 and

15:22 we're very excited about that

15:25 opportunity well that's exciting and I

15:27 think that's a great note to add on I

15:29 want to thank our Pano again for joining

15:31 us today and we encourage you to drop

15:33 your own questions into the comments and

15:35 we will answer them as soon as we can to

15:38 learn more click on the links in the

15:39 description and follow us on Twitter see

15:42 you next time

16:01 possible