Automating Data Center Operations — Roundtable and Q&A

0:04 hello you're watching telecom tv i'm guy daniels director of content and welcome

0:09 to our live q a program on automating data center operations

0:16 juniper networks and analysis mason have collaborated on a new survey with 30

0:22 communication services providers to find out why and how the introduction of

0:27 automated processes for data center operations can be accelerated and managed now we unveiled the results of

0:35 the survey on tuesday and if you haven't yet seen that discussion program don't

0:40 worry you can watch it on demand later just follow the links below this video

0:45 player window today though we're going to be sharing more thoughts about the survey and what

0:51 it means for csps as well as answering your questions during this live show now

0:57 there's a form here on the website for you to submit them we have already received quite a few following our

1:04 earlier program but we would still like more so if you haven't already done so

1:10 then please send us your questions now we have about 45 minutes remaining so

1:17 don't delay well let's now meet our guests and taking part in today's live q a show are

1:26 gurkham yeet principal analyst at analysis mason

1:31 michael bushong vp cloud ready data center at juniper networks

1:38 and ernest oldbart i.t architect at riveison informatic

1:44 welcome all of you thanks so much for taking part today now before we get to

1:50 our first viewer question mike let me first of all ask you um what is

1:56 juniper's philosophy when it comes to data center networking and nice hair by

2:02 the way thank you yeah green is my natural color so uh um the the green just to kind of dispel

2:10 that um we had an internal bet inside our company depending on whether how you look at it

2:15 i either won or i lost and so i am now green in terms of like just data center

2:21 philosophy i think the the philosophy is fairly straightforward in the physical switching layer we let

2:27 the economics and the physics dictate the choice the industry is largely settled on how we build physical data

2:34 centers that's typically merchant silicon lead building out fabrics with evpn or bgp

2:41 and once we standardize that it means the action moves from on the box to above the box

2:48 philosophically we believe that it's all about operations and within operations the thing to opt

2:53 to optimize for it's not necessarily going only for speed it's about reliability it's about

2:59 making sure people can do things the right way every time without fail we've embraced this whole concept of

3:06 intent-based networking which is the idea that we should be declarative we should say what we want

3:12 and allow the tools to translate what we want into how we get it because the future is decidedly

3:17 multi-vendor the future is uh it's decidedly automated it's software defined and so

3:24 that's really where the action is where most of our customers are engaging and frankly where i think most of the

3:30 decisions will be made over the next decade i think that's that's basically where we're headed

3:35 great thanks for that mike uh and why did you engage announcers mason to conduct this survey for you

3:42 so if we're convinced that the action is in the operation space if we believe that

3:49 automation is where things go if we believe that um things like uh intent-based models are

3:56 the way to solve pain and we're absolutely convinced of that but if you look at like just industry traction

4:03 people aren't moving as fast as you would expect and this is a actually a conversation like operations is a

4:09 conversation that's more than you know a decade old at this point and what we wanted to know was

4:14 frankly why you know what are people seeing what are the the barriers what are the challenges

4:19 what are the things that are driving their behavior and frankly we can have all the opinions we want but what we need is is actual

4:26 you know data to complete the analysis so that we can help meet people where they are because frankly as an industry

4:33 we've sunk you know i would argue millions of people hours into trying to solve the problem

4:39 and yet the problem still persists so how do we get enough information to take action how do we become more helpful

4:45 understanding that this isn't a pure technology problem there's more that sits behind it but what's the nature of

4:51 that what is what's holding people up and then frankly what what opportunities does that create how does that change

4:58 engagement and then finally like how do we help so the the hope was that you know a survey like this would would really open

5:04 up some more specifics allowing us to plug in where people are and then get them to where frankly we

5:10 all know we want to be great absolutely thanks mike so goku moving across to you uh we need to

5:16 understand more as mike says so tell us about the the server you conducted and and the key results you discovered

5:23 sure yeah so we did a quite detailed online survey with 30 large tier 1

5:28 communication service providers around the world and it included both strategy people and

5:34 technical staff and in addition to that um we had in-depth conversations with five other

5:41 csps to drill into the uh their data center automation um approach

5:46 so we asked them about their key drivers and transfer their data center strategies the current state of their

5:53 data center network automation their future plans and the main challenges and the barriers they are facing today

5:59 and we have several very interesting findings surfaced from this research first of all

6:06 in terms of the key drivers and the trends we see that the csp data center strategies are shaped around three main

6:13 areas one is the new opportunities in the enterprise many services and

6:18 connectivity markets and the second is network modification for 5g including the the cloud native

6:25 transformation of 5g mobile core and radio access network and the third is the deployment of

6:31 near-edge computing locations in order to support support both 5g and

6:36 enterprise and the consumers services so all these new services and networks they

6:42 need highly automated and programmable data centers to run on however when we look at the current state of

6:49 most csp data centers they are not really ready for the requirements of 5g edge and the new

6:56 new service opportunities i just mentioned csps have made very limited progress in

7:01 terms of automating their data center network operations across day 0 day 1 and day 2 processes

7:08 and also they don't really show a strong ambition or urgency to address that in the near term and which i believe poses

7:14 a significant risk to their aspirations and they identified

7:20 three main barriers and inhibitors causing this low level of automation and the first one the top barrier we

7:26 found was operational complexity and it stems from the highly fragmented um data center

7:32 network technologies and operations because csps they have a lot of vendor-specific and domain-specific

7:38 silos which makes it very hard to rationalize and automate their data center networks and as mike said it's

7:45 not just about the technology we also found some operational and organizational challenges for example

7:51 the most csps are finding it difficult to justify the business case for the automation investments and this is not

7:58 because there is not a business case but it is that most csps they struggle to understand the business case and gather

8:04 the data internally to build that and finally we found that csps have an

8:09 entirely different mindset towards data center networking and automation compared to enterprise and hyperscalers

8:17 most csps don't have an automation first mindset and they are more accustomed to throwing more people as human glue to

8:24 solve their problems or doing very expensive custom integrations or developing add-on tools so overall

8:30 they're generally behind the hyper skaters and even most enterprises when it comes to data center automation

8:37 fascinating thanks very much for that gurkham and a reminder that there is a link below this video and you can access

8:44 the research yourself now earnest uh let's let's uh come across to you because we invited you to this panel

8:50 because rifason is a service provider that is at the forefront of network automation so your perspective is really

8:56 going to be useful here tell us more about your business and do these survey

9:02 findings that gurkham has been talking about do they resonate with you well first for our business uh we are 90

9:08 service providers in the business for 50 plus years mainly targeting uh the reif eisen group here in austria and in

9:16 eastern europe uh there's banks and insurances in there so there's a lot of different use cases uh for over the

9:22 decades uh we we deployed it infrastructure for so we have a lot of experience and one of the findings that

9:29 that resonates is one thing said just before that

9:34 that's the mindset i think it takes a very long time to develop a mindset that

9:40 really drives you towards automation you have to feel the pain obviously um which we have felt over the decades

9:47 in doing so and also to develop the need uh for

9:52 standardization to really develop a standard for certain infrastructures that you deploy and run

9:58 um i think that comes over time because in the first places you always think i can manage i can do that by hand and you

10:05 of course do it by hand in the first place because you usually drive pocs and stuff like that

10:10 and that's not where you deploy full automation but at a certain point when you put it in the production it should be there and that's a hard transition to

10:16 do so that's that that's one thing that resonates quite well of of course i would maybe like to add two more angles

10:23 that were key drivers for us that's a complexity of the infrastructure that can hardly be

10:29 managed any other way than by automation because the the technologies we use become increasingly complex and

10:36 increasingly complicated actually to to deploy and and to run uh from just the

10:41 number of parameters think ebp and btb vxlan fabrics for instance there's a lot

10:47 of neighborships going on a lot of parameters and ip addresses to fill in doing that by hand in a medium to large

10:54 scale is next to impossible to do error-free that's one angle the other relates to that a little bit it's it's

11:00 uh targeted towards people because when uh when you look at the networking industry and that's something i learned

11:05 from from my engagement with juniper in in the course of the last months uh was

11:11 that it's actually a global problem it's something we experienced here in austria that there is not many people wanting to

11:17 join the networking industries there is not young guys coming up you you have a problem in in the in the hr area where

11:24 to get people in the in the correct skill set if you want and therefore you need to find a way to

11:31 to deal with additional features with additional scale with additional data centers networks targets

11:38 etc and one way of course to deal with that is to increase your level of automation because then you can actually

11:45 use uh the same amount of people um skillful people uh for for those things

11:51 and and scale uh much more easily than than you could if you'd deployed a

11:57 manual or even semi-manual method of doing that so those two things are are

12:02 something that were key drivers for us as well so ernest it does sound like there's a a

12:08 lot of factors there's many factors that you need to balance if you're going to successfully implement this

12:15 absolutely yeah short anecdote here of course there's the purchasing

12:20 angle as well i think mike mentioned it before uh that the vendor independency that's also

12:26 something that uh any purchasing department strives for and would like to achieve uh and uh i think i don't need

12:34 to tell the audience how hard that has been over the last decades with a with an automation layer that sort of

12:40 abstracts the hardware to a certain extent i have to say it's not a free-for-all where you can deploy

12:45 whatever box you wish just bought around the corner but when you do an abstraction of of your

12:51 hardware it gives you a certain kind of freedom and here's a short example in the latest rfp we did that was right at

12:58 the beginning of that of that that's called a global chip crisis where where uh the

13:04 supply chains getting increasingly under stress and delivery times uh prolong

13:09 themselves more or less um we were actually free uh to to name three

13:14 hardware vendors um and and to do that uh and tell them whoever can deliver

13:20 first um we will take your equipment if uh because we could do that from uh from

13:25 the tools that we we had deployed of course purchasing loved that um and that that was an angle that is important or

13:33 could be important when when you say on the time stress um luckily for us um it

13:39 all fit well together and we got the uh our number one of the rfp so we were

13:44 kind of lucky in that regard but it just gives us gave us the opportunity to do that if necessary so that's also uh

13:52 quite an interesting angle i thought in in the process wouldn't have anticipated that yeah absolutely thanks ernest for that

13:59 now we've been calling for your questions viewer questions for this program

14:04 and here's one we received in um mike i may put this one to you first possibly

14:10 um let me read the question out the lack of standardization was identified as the

14:16 number one barrier to data center automation in the white paper so who therefore should be leading the

14:23 standards push mike is that one you're able to take a stab at first certainly

14:28 have opinions um you know standardization tends to be a

14:33 team sport um i think the the what most people believe is that you

14:38 know people have to to show up you know companies vendors have to agree fundamentally there's standards bodies

14:44 that emerge you know obviously we've got you know long-time standards bodies like ietf we've got emergent

14:52 groups that come together like onug that all kind of aspire to provide

14:57 i guess guidance that i will tell you this though having run standards bodies before

15:03 it's not it's not only incumbent upon the providers of technology to drive things

15:10 i'll tell you what moves needles is adoption frankly you know dollars will will drive

15:17 or euros or whatever currency right will drive behavior and so what what needs to happen frankly

15:24 is that you know people need to to move away from what i would call legacy architectural practices

15:30 ernest commented on you know there's a mindset change the mindset is also what i call the derivative rfp when most

15:36 people build out infrastructure what they do is they take an excel spreadsheet that had all the requirements of the old infrastructure

15:42 and then they add two or three lines at the bottom we are reasonably good as an industry at

15:48 picking up new new things we are horrific as an industry at retiring old things

15:54 and it's the this sprawling technology bits that actually are a greater inhibitor

16:00 i think standard apis are fine but but common and and sort of openly defined

16:06 api layers can get past a lot of that even if they're not standard as long as they're they're unchanging they're well

16:12 documented and they're well used and then we can handle things like versioning so they're not changing all

16:17 the time like if that happens we can make progress but the thing that's very difficult is

16:22 managing diversity ernest also mentioned the whole complexity angle the more things we have

16:29 the more difficult it is and i think you know fundamentally it's this and

16:34 in the csp market in particular by the way we have a tendency to to sweat our

16:39 assets we want to hold on to the technology as long as we can because it drives a better return right

16:45 you don't want to decommission a service the the side effect of that though is that these things linger and they don't

16:52 linger for a few quarters or a few years they linger for decades and when that happens it means that anytime we bring

16:59 in like an automation or an abstraction layer we're not abstracting something clean you're abstracting you know

17:05 decades of technical debt and that frankly is is near impossible

17:10 and so we think about who drives the standards you know i would say flip that conversation a bit and let's

17:17 develop a practice of removing things one of the things i tell customers all the time

17:22 is make a list of everything you have right the number of devices the versions of operating system

17:28 the different protocols and technologies you have provisioned the firewall rules that are

17:34 you know exist across the the whole environment make that list and then record what you add to it and record

17:40 what you take away and if that list isn't meaningfully smaller year on year you're moving in the wrong direction

17:47 you can't standardize when it's a million things what you can you can standardize when things are relatively

17:52 small and i think that's what we've learned from like the hyperscalers instead of bending the infrastructure to

17:57 the applications they bent the applications to the infrastructure like that's and that that's a bigger more

18:03 bang for your buck kind of change that i think we make so i would say encourage customers i

18:09 guess i know long answer but i would encourage you know users in the space to to really

18:15 take control their destiny and and think less about how do you you know incrementally change and more about you

18:22 know how should you design what are the architectural principle principles you'll base it on and then automation

18:27 becomes a much easier task great point there mike thank you very much for that answer

18:34 um well let's therefore move on because we've got another audience question in here um

18:40 gurkham let me put this one to you based on the survey findings what do you

18:46 recommend that csps do in the near future yeah

18:52 well i think our research makes it very clear that you know the csps should take action

18:58 urgently and they really need to increase their focus and the investments in the data

19:03 center automation given that the new 5g and edge services you know they're just around the corner

19:09 and automation is a long journey and csps cannot really afford the risk of waiting for

19:15 you know end-of-life replacement or technology refresh cycles so as a first step i would recommend

19:21 csps to establish their priorities and timings for which parts of the data center

19:27 networks and processes you know they need auto they need to automate and of course they should be in line with their

19:33 um the business strategies and objectives and another key step um in this journey

19:39 i would say will be the tackling the technology fragmentation and the complexity i mentioned earlier

19:45 and one way to do that is adopting your horizontal multi-vendor abstraction and automation platform

19:52 and because operators they really need to collapse these um existing silos and

19:57 and they need to move to more intent driven and automation-first approach for their data center operations

20:03 and going forwards um csps they need to think about introducing new policies and

20:09 mechanisms in order to prevent um you know this vendor dependence bottoms up

20:14 approach on to data center designs as mike mentioned and there are lots of legacy technologies technical debt so

20:20 this cannot repeat again so for example they can institute a board or a central group

20:26 that oversees the introduction of any new data center technologies and they would insist that any new vendor

20:33 conforms to the data center automation plan and the regime the operator is already in place and we are seeing um

20:40 similar approaches um to this by advanced operators um in their 5g cloud platforms already

20:46 and lastly i would say um all this automation and the transformation is not something that csps should take on a

20:53 loan because um they need suitable technology partners to update this journey so they need to look for

21:00 external partners that have extensive automation experience and not only with you know other csps but also with hyper

21:07 skills and enterprises so that they can get the solutions and the professional services um to help these the

21:13 transformation costs disruption and the risk great thanks kirkum and mike you'd like

21:20 to add some comments as to what csp should be doing in the near term yeah i wanted to go back to the the opening

21:26 comments that gorka made about you know business case the other thing i would add is you've got to instrument your

21:31 environment you know we're decent as an industry at instrumenting our services you know

21:36 what's performance look like in the operation space we don't always have information you know what if you're

21:42 a leader in the space it's things like decision support how do you know how much you know time you saved how do

21:48 you know things like opportunity cost that requires a little bit of thought and it's not just about you know how

21:54 many heads can i remove it's about how more like how much more effective can i make people how do i improve

22:00 productivity i think you have to measure that and i think that has to become you know part of the operating environment

22:05 and it's more than deploying a technology at some point you got to understand what your baseline is

22:10 and that gives you some idea by the way of competitive benchmarking which allows people to compete in this space

22:17 i just i think paying attention to to how you justify decisions is really important and often overlooked

22:23 because it doesn't necessarily drive the specific service in the moment but but gorkham talked about a couple of things

22:30 that are important right the proliferation of you know edge clouds and 5g i mean it's basically a trade-off

22:36 we're saying we're going to make lots of small data centers that are going to be independently operated the way you pay for that

22:43 effectively is through operations if you don't instrument that environment you're never going to know if the economics are

22:49 going to play out as you think they should thanks mike and gurkham just to be absolutely clear here

22:56 unless they've already started of course csps should start this journey as soon as possible it's it's not just simply

23:02 part of a technology replacement life cycle you're absolutely right yes um

23:08 so as part of this research um we had you know conversations with operators and some of them um their thinking was

23:15 more around yes we are gonna think about um network automation sdn and you know

23:21 programmability when you know the the next technology um life cycle investment

23:27 um starts so they're more like okay you know let's let's wait and see if that we didn't really see you know the urgency

23:34 and there's clearly there's the disconnect between you know um all that 5g and the age aspirations and you know

23:41 the the the attention and then the level of interest that is um going into the automation um at the moment so

23:48 and our research i think it's it's very clear that um they're already late because um first

23:54 of all automation takes time you know there's not like like a magical solution so there are going to be you know

24:00 significant migration and modernization activities they need to take place to get there

24:06 right and so given the you know the low levels of its automation we see in the csp data centers today waiting for the

24:13 next replacement cycle would mean japan jeopardizing your you know 5g and age opportunities because the data center

24:20 network fabric and its automation is becoming an essential underpinning of their business

24:26 great thanks for those comments to to that question and a reminder to our viewers please keep sending in your

24:33 questions we are about halfway through our program now um ernest let me come back to you as a

24:41 company that's at the forefront of network automation and is one of those companies that's well on the way in this

24:47 journey tell us what have been your keys to success well first of all we know we're not

24:53 finished yet it's a it's a never-ending journey it goes on and on and on and on um the key is to to to success

25:00 is i think uh again a mindset approach we discovered fairly early uh and and

25:06 didn't even know that there were fitting solutions out there at the time that uh the the level of complexity i

25:12 mentioned earlier can only be managed with an automation layer of some kind so we didn't know exactly what we

25:19 were looking for we just knew um well this this has to be automated it's way too big and to complicate it to do that

25:26 by hand i mentioned the people of angle earlier as well so i guess a

25:33 key factor in our success if you want this is being aware of where some key

25:39 challenges were i'm not saying we discovered them all because there will be surprises along the way you you will

25:44 have problems this is it this is this is a given actually this is something uh there's something in it uh you you

25:51 didn't think of all the time and didn't anticipate but i think an awareness uh much like garcon

25:57 managed you need to do it now this this is this was actually one of the thoughts we had uh during uh the beginning of our

26:04 of our process here i have to admit though it it coincided with the life cycle so this makes it of course a lot

26:10 easier to do that um when when you know that that you have to do something

26:15 and it's actually uh driven by some end of life or end of sale announcements by your former hardware

26:22 vendor so that helps but i think the awareness of of many

26:28 angles is is a key factor in that in that journey

26:33 might be also that with 50 years of experience that you get a little more angles in the back of your head

26:40 than companies that are not that long in the market so that could also have helped in that regard but i'd stick with

26:48 with the general awareness of that you need to do it now as curcums aptly put

26:53 it um now is actually too late the best time to start was two years ago

27:00 and ernest you you mentioned the issue complexity there is it a case that you usually can't get rid of complexity

27:07 completely but you can you can either hide it or you can manage it and and this is the whole the whole point of

27:13 automation it absolutely is you never get rid of it um our company exists for

27:18 50 plus years there's a legacy environment that's there that doesn't go away just because you're doing something

27:24 new you have to deal with it and and therefore the complexity of course stays to to a

27:30 larger amount but what you can do is i'm going back to the beginning now standardization is is

27:36 something that we have been striving for over a couple of decades and is something we of course never fully

27:41 achieved but but it's something that we're always targeting for but having the tools to do that is is of

27:48 course lightening the situation a lot and by that intent based networking when you

27:55 set an intent and you have the tool set to keep that intent that is a very strong support factor in

28:00 in any standardization drive you you have and deploy and that's something that enables you

28:07 again to to run in in a very smooth way in fast and stable and and that's the main

28:15 characteristics that our customers think banking and insurance again

28:20 want from their i.t infrastructure they're not so much concerned about knobs and whistles they're really really

28:26 concerned about does it run 24 7 because this is where payment and insurance contracts run over and that's the that's

28:33 the the main business driver of for them um what it boils down to

28:38 technologically that's our job to do and of course once you standardize things and they are all done in the same

28:45 way uh it's very likely that they run a lot more smoothly than when you're dealing with an anthill of

28:51 of complexity in your data center sure thank you ernest for for those

28:56 views now we've had a question in it's of a similar type of question to the last one here so uh perhaps uh mike and

29:03 kirkham i could put this to you and mike perhaps i could start with you and it's about complexity and the question is you

29:08 know how do you resolve the operational complexity problem because this was a

29:14 big big problem that was identified in the research and as ernest has also said but you know is there an easy solution

29:20 here is there an easy solution i don't know that there's easy solutions for most of what we're doing and we have

29:26 to there's an inherent complexity and challenge in what csps have to do broadly

29:32 i think in terms of how do you address the complexity you know i do think we've talked a bit about how do you standardize around designs i think

29:38 snowflake designs are the enemy they everything that you add that's different and diverse is challenging

29:45 i think that you have to elevate so ernest talks a lot about abstraction you have to eliminate elevate operations

29:52 not just above the box but above the network ideally if you move to declarative models the whole premise of intent-based

29:58 networking just as a movement is this idea that you ought to say what you want not necessarily how you want it

30:05 when you do that the software controllers act as effectively api brokers translating intent into

30:12 underlying behavior that's what provides the multi-vendor elements by the way so when ernest talks about that procurement

30:18 advantage or the lead time challenge the only way you can get that is if your operations aren't

30:23 twiddling in vendor-specific cli but are rather looking at kind of declarative intent allowing the system to to adjust

30:31 to what the underlying devices are i think standardized designs that are you know templated or blueprinted so

30:37 that you can essentially stamp the same thing out over and over and over again that's critical as we talk about the the

30:44 proliferation of edge cloud the rollout of 5g which moves from these large national data centers to highly

30:50 distributed what that means is that you're going to replace you know very large data centers

30:55 with lots of very small data centers and that means that we've got to create those as as similarly as possible so we can

31:02 reduce the kind of the technology and design sprawl that would otherwise lead to even additional complexity

31:08 and then on top of all of that you can't leave things like you know choice and flexibility out so ideally

31:15 it's about how do you do all that with with a uniform operations layer i want my operations to be consistent allow the

31:22 economics the lead times the the physics to dictate my underlying choices

31:27 so you've got to then you know build out explicitly abstraction bits um it means

31:33 that it's you know a bit of an ecosystem play instead of like all in on you know here's my one vendor

31:39 and that's important because that's how you diversify risk by the way and the current supply chain crunch should teach

31:44 us all that that's important so it's how do you design abstraction at the top and and i want to go back to the

31:50 people side of that right the it's not only a technology problem right the reason we customize is because we want

31:56 control there's a bit of a control trade-off which you have to agree to is that you don't want everything to be you know

32:03 finely tuned highly tailored to a specific environment you've got to be willing to say look i'm going to go in

32:09 with with well tested blueprints or are kind of archetypes for

32:15 for different data centers that i want and then you have to have the discipline to maintain that the tooling can help

32:21 but there's a policy angle you know a corporate policy angle that comes in uh behind that to really support it if you

32:27 can provide both of those the tooling and sort of the cultural support then i think you have not just a chance you

32:34 know i think i think you can be successful and what's interesting is is because it's such a difficult thing to do that success will create outsized

32:41 business benefits for for the overall business and it will drive a much better experience for the customers that attach

32:48 great thanks very much mike um well look we do have quite a few more questions to get through so perhaps we should just uh

32:54 we should move on to our next question a lot of these questions are connected um

33:00 gurkham here's another question here um how do service provider data center

33:05 strategies vary based on geography is there any variance uh dependent upon

33:11 where they're situated yes yes absolutely and i would recommend anyone to go back and watch

33:18 the first part of this series where i explain all the experiences in detail but i can still give you

33:24 an overview so when we looked at the overall data center strategies by geography

33:32 we found stark differences between the western world and the east especially asia

33:37 pacific so north american and western european csps they've been reducing and

33:43 consolidating their to their traditional data centers for a while now because

33:48 they are transforming their i.t and the networks with public cloud and the sas providers and they're also looking to

33:54 cut their data center capex but of course there are still plenty of traditional data centers um these csps

34:02 still own and operates however the csps in asia pacific are

34:07 actually adding new data center sites and capacity because a lot of them are actually in the

34:12 business of providing data center services and historically public cloud providers presence in those

34:19 regions weren't as strong as as in the western world so

34:24 when we looked at the data center automation specifically

34:29 we actually found some quite interesting results some csps in asia pacific middle

34:35 east and latin america they reported higher levels of data center automation than their peers in north america and

34:42 western europe and we believe one of the reasons for this is that the csps in these regions

34:47 they have a newer technology data centers as they continue to invest in you know building new data centers and

34:54 new technologies come it's more automation in north america and western europe on

34:59 the other hand there's still a lot of legacy in csp data centers and that needs you know transformation and

35:05 modernization so this might be the impact of the data center footprints

35:10 and the capex reduction strategies that we see in these regions great thank you and i'll just echo what

35:17 gurkham said there if you haven't seen the first part in this series yet then then you ought to if you can there's a

35:24 link below this video player window now mike question uh to you based on on your

35:31 experience you know what are you hearing from your customers about how they're building the business case to automate

35:38 data center operations right so there's two ways to fundamentally build the business case

35:44 one is kind of cost avoidance the other is preparing for growth and looking at how do you accelerate

35:50 um on the cost avoidance side you know it's really how do you become more efficient and that's benchmarking the things that

35:56 you have to do and then you know building out the tools to make that easier to do i want to caution people here that that

36:03 if the goal of automation is that you're simply going to remove people i think that's a a bit of a narrow goal right i think we

36:10 have to look at how do we best deploy our folks i look at it as removing the toil like how do you get people out of

36:16 the things that are you know time consuming but not necessarily difference making the things that ought to be automated and allow

36:23 them to apply their specialist skills uh in support of things that are more

36:28 difference-making ernest talked earlier about the difficulty of just maintaining staffing this whole pandemic and work

36:35 from home thing is opening up hiring worldwide and that is very challenging

36:40 because it means companies that we're traditionally hiring in major tech hubs can hire from anywhere and that puts people under under pressure and as those

36:46 people leave it's like how do you get the remaining workforce to be as productive like that's

36:52 i i think you got to think about that you know from just a business case perspective how quickly do you onboard

36:57 what does your recruitment and talent look like um on the other side there's a kind of a growth

37:03 perspective which is you know for every unit of growth you know what happens to operations spend

37:08 ideally your operations don't grow linearly with the the with the the size or footprint of your overall services

37:14 and so there's a an element there that says how productively how profitably can we we grow services

37:20 i think that's the other kind of more obvious angle you put those two together though and you get a kind of a mix of opportunity

37:26 cost along with you know additional online capacity going forward that's that's how i would build the business

37:32 case i think it allows you to make a case to management that says you know allow us to be more effective and really look at

37:39 the you know the the bottom line while simultaneously supporting you know the growth initiatives required to drive top

37:45 line growth great thank you very much mike and ernest we have another question for you

37:52 how do you see your data center operations evolving in the next five

37:58 years well that's a that's a good one actually

38:03 much like in the in the data center in a fabric environment if you want where you

38:09 have obviously uh an automation solution deployed um there are similar environments uh all around those those

38:17 infrastructures like the sd van area the software defined data center

38:22 approaches that usually pop up cloudified everything that that pops up

38:28 all of those infrastructure areas have a certain networking layer aspect to it

38:34 and incorporate it actually so we would see ourselves developing in the

38:39 direction of integrating those those infrastructures together

38:45 one way or the other it's kind of hard to see and say how this is going to to

38:50 be implemented and of course drive the abstraction over all

38:56 these infrastructures so so you can actually deploy features in a similar manner or

39:01 in a connected manner because it might be that any advanced circuit and then the

39:06 software defined data center tenant might want to talk to some data center parts that you deploy for

39:14 x customer or and that needs to be implemented in in a certain way and if

39:20 you do it in three different areas um that might get complicated so we will

39:25 probably head in look in that direction of course again

39:30 when you when you do it now you have a chance of of having it once you need it

39:36 the immediate need might not be here right now because consciously we decided just to target data center automation um

39:44 at this point in time but i think uh things like that will will pop up in in

39:50 in the next couple of years say five years or something like that and then this is where we where we think

39:55 uh towards to fascinating thank you for that ernest uh

40:02 we've still got a few more questions we want to get through even though we are approaching the end of our show let's try and get through a few more if we can

40:08 uh gurkham let me put this one towards you the question is do csps have different

40:15 strategies for their edge data centers it's not a real difference but

40:22 our research showed that there is strong interest in building new edge data centers from the csvs across all regions

40:28 so despite the know the differences we see in the traditional data centers it is

40:34 quite contrary for the edge you know all operators um they have that they have strong interest um to

40:41 build out um new edge locations for example um they're typically um in the

40:46 regional or metro data centers and in some cases on the cell sites and and we

40:52 are also seeing there are two main drivers behind this um this edge data center investments

40:58 first is the new enterprise edge clouds of course in order to provide new iot

41:03 operational technology and enhance ict services and second is the network edge

41:09 clouds to host you know the 5g functions such as vran and openran

41:14 and interestingly and the csps are focusing on both of these areas at the same time but what is going to be really

41:21 important for csph is that we are looking at a highly distributed cloud

41:27 environment and the network fabric so starting from 20 to 50 edge data centers

41:32 which what csps indicated is first in terms of the size of their age

41:37 footprints but in the long term we expect that this number will go up to hundreds and even thousands of edge

41:44 locations so considering the sheer number and the complex of these age footprints you know

41:50 as well as the nature of the new and advanced services that csps are planning to deliver network automation in these

41:56 edge data centers will be a must-have not a nice to have and this is this is really crucial if csps want to play in

42:03 the edge complete and better chains yeah it's going to be essential thanks kirkum uh ernest question to you are

42:12 there any specific tasks in the data center life cycle that you are

42:17 prioritizing and automating first well there's a couple of obvious ones of course like the onboarding process where

42:24 you where you start uh equipping your data center uh to in certain uh

42:30 tranches of of equipment be it racks or rows or whatever that's an obvious one um our

42:37 first and and one of the more important things is anything targeted towards um i'd call it

42:43 migratory work as i mentioned earlier we have a legacy environment that's still there when when you want to move that over

42:50 or connect to to your current uh modern data center fabric

42:56 those aspects would of course be first uh to to be automated because this is what you will need most over over the

43:04 run of the year or two it takes to to complete your migration cycle from

43:10 current to to new data center fabric so those would be a priority for us

43:16 great thank you ernest for those answers and i think we can get at

43:22 least one more question in um here's a good one uh mike

43:27 is there a role for cloud hyperscalers in all this if so you know how do these cloud public cloud hyperscalers how do

43:33 they fit into this so you leave the controversial ones for herd um so that's great uh hyperscalers

43:39 and csps it's an interesting relationship i think it varies geographically um i

43:44 think uh hyperscalers look like partners uh hyperscalers look like competitors um

43:51 gurkham talked a bit about the sort of the proliferation of edge cloud uh you know talking about pushing it all the

43:56 way out to the cell sites um in those cases the the csp footprint starts to look a little

44:03 bit like a kind of a cloud hotel where we had carrier hotels in the colos to

44:08 provide connectivity if customers end customers like the enterprises that purchase

44:14 sp services if they've got applications that reside in different clouds the question is how do they connect those

44:20 so there is a role with the hyperscalers for sure there's a lot of questions that come from that what's the point of

44:25 demarcation between the service provider and the csp uh who maintains account control

44:31 because it's you know who is a partner today might look like a competitor tomorrow so

44:36 those are really big strategic questions and frankly you know we would need like another hour to talk about that

44:43 even a little bit obviously some of the hyperscalers have built out their own services

44:49 around some of the vnfs that drive you know telco cloud type deployments and so

44:54 there's opportunity to leverage that in some environments with the same kinds of considerations you know where's the point of dmarc how do you monetize where

45:02 do the profit pools get divided i think what people need to do from a hyperscaler perspective is you know go

45:08 in pretty pretty broadly and have really deep conversations about not how to roll

45:13 out a service but how is that going to evolve over the next few years because i think things like the you know what

45:19 what's what's the edge of the managed boundary is that really the the transit gateway or the vpc gateway

45:25 on the public cloud side or does it go into the the application like the cni um in an application instance

45:32 understanding where that point of dmarc is has huge operational implications on what tools have to be used how policy

45:39 and control is centralized how do you scale operations across these hundreds or thousands of of smaller data centers

45:45 that were or distributed edge clouds that we've talked about um so that's what i would look at and then on the

45:51 hyperscaler side the other thing to think about is you know what are the routes to market as you think about carrying those services out how do you

45:57 scale them quickly and rapidly so i think there's just a very rich conversation to have in the space so so

46:04 absolutely in short the hyperscalers have a role absolutely they do just be very thoughtful about what that

46:11 role is because you want to have a deliberate business outcome you don't want to sort of incrementally find

46:16 yourself into a a business position you know two three five years down the road that you hadn't planned for

46:23 great thank you mike we did keep the the trickiest question there till the end but terrific answer thank you very much

46:29 for those comments well unfortunately we are out of time so i would like to thank all of you for

46:36 sharing your views and expertise and helping to answer our audience questions

46:42 and don't forget if you haven't yet watched tuesday's discussion program with gurkham and ben baker who's senior

46:49 director of product marketing over at juniper networks where we unveiled the survey and discussed the results then

46:55 you can still do so just follow the links below this video player window and we've got links to other related

47:02 information it's all there in the video description plenty for you to watch and read

47:08 for now though on behalf of juniper networks the team here at telecom tv and to our guests today thank you very much

47:15 for watching and goodbye