CSP Data Center Network Automation Strategies for 5G, Edge, and Advanced Enterprise Services

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

0:09 to take a detailed look at csp data center network automation strategies for

0:15 5g edge and advanced enterprise services we'll be identifying the main barriers

0:21 and challenges and finding out how csps can accelerate the progress of their

0:26 data center network automation and joining me to explain more and to

0:31 discuss the results of a new global study and white paper are

0:36 ben baker senior director of product marketing at juniper networks and gurkham eat principal analyst at

0:44 analysis mason well welcome both of you gurkham in a moment will go through the findings of

0:51 the research you conducted with tier 1 csps worldwide but first ben

0:57 why did juniper commission this study well guy we have literally hundreds of

1:04 communication service provider customers around the world and our account teams are working with these companies every

1:10 day and then when you think about cloud and cloud native infrastructure again on the

1:15 juniper side some of our largest customers that are the cloud hyperscalers and we've been working with

1:20 them closely since their inception so we do think we have a good understanding of the sp data center

1:26 environment however we want to get some third-party validation of what we think we already know and

1:32 of course when you conduct a survey like this you do learn a bunch of new things about the situation at sps and

1:38 when you aggregate this information across a broad diverse range of of service providers like analysis mason

1:45 has done it helps you get a fully coherent view and we certainly take these inputs and make sure they match

1:51 what we're doing in terms of product roadmaps and finally we want to specifically dig into the state of data center automation

1:59 what are the problems that sps are having in their in operating their data centers and ultimately we think that

2:05 juniper has the portfolio to help sps manage their data centers their telco cloud infrastructure in hyper efficient

2:12 ways similar to how the big public cloud providers do it so we want to understand why are sps not

2:17 adopting some of these tools as fast as we think they should be sure okay thanks ben

2:23 so gurkham tell us about how you conducted this research

2:28 yeah sure um so over the last six seven months we collected a great amount of

2:34 inputs and data first through an online survey with 30 tier one communication

2:40 service providers csps worldwide and we also did five additional in-depth phone

2:46 interviews with them so our aim was to identify the key drivers and the trends for csb data

2:52 center strategies and particularly to understand csp's approach to data center

2:57 network automation their main drivers and challenges target use cases and the

3:02 tools they are using to automate their networks so you can see on the slide the

3:07 demographics of the csps we surveyed and most of the csp participants were from

3:13 north america western europe and southeast asia which also includes japan and india but we also had csps from

3:21 other regions as well so overall this is a global study and when we look at the types of csps

3:28 the large majority was integrated service providers with mobile and fixed networks and enterprise managed ict

3:34 services so these are typically large csps with extensive network and data center assets

3:40 and service offerings and in terms of data center size most participants say clustered around

3:47 10 to 24 data centers but we also had some csps with more than 50 or hundreds

3:53 and we actually did in-depth interviews with some of these large csps that provide global data

3:59 center services with locations spread all around the world and finally the titles and

4:05 responsibilities we have a very balanced mix of respondents here which helped us

4:10 get a very holistic view of the state of the csb data center network automation today and their future roadmaps

4:17 not only we had senior executives that are in charge of their data center strategy and transformation but also the

4:23 technical stuff such as network architects engineers operations managers

4:29 so these are the people who get their hands dirty you know with actually implementing and using data center automation solutions

4:36 well let's now look at what you discovered what did you find were the key drivers and trends for csp data

4:44 center strategies sure so there are three main drivers that are

4:49 influencing csbs data center strategies and investment focus today

4:55 so first is the new opportunities in the enterprise many services and connectivity markets

5:00 there is a major transformation underway in the enterprise market today and this is started with sd-wan sassy and more

5:08 recently csp started to add multi-cloud networking um industrial applications

5:13 and use cases at the edge and also private networks to their enterprise service portfolios

5:19 and the second is network clarification or toggle cloud especially for 5g

5:24 networks it is it's a major driver and the third are the deployment of new

5:29 edge computing locations in order to support both these enterprise and 5g services i just mentioned

5:36 so all these new technologies and advanced services are very much interrelated and complementary to each

5:42 other and they will be deployed and operated in csp data centers so csps needs really state of the art

5:48 operational efficient and programmable data centers in order for them to be successful

5:54 so these are the high-level top three drivers but uh when we asked csp is their number

6:00 one priority the majority of the csps rank 5g telco clouds is the most

6:05 important driver as you can see on the rights csps are in the process of building

6:11 their 5g networks using cloud native the segregated technologies and they're starting from their mobile core networks

6:18 for example running on centralized data centers and gradually expanding into the radio access network where you have the

6:25 vdus and the vcus and deployed in network age data centers

6:30 so this finding suggests that the csbs acknowledge the need to modernize and automate their data centers to

6:36 successfully manage this infrastructure transformation we also split the responses by region

6:42 here and one of the key regional variances that jumps out of it is the data center agenda of the majority of

6:48 the csps from asia-pacific is driven by cyber clouds but the main drivers for

6:54 csps from other regions are more evenly spread out especially for north american and

6:59 western european csps the main focus on data center investments include things like adoption of public cloud and

7:06 multi-cloud capex reduction and to lesser extent edge computing

7:11 so what is driving these regional differences in the data center strategies

7:17 so the chart i'm showing on the left uh shows the percentage of workloads that run in different environments so from

7:24 traditional on-prem data centers to public clouds and sas and this is both today and in two years

7:30 time overall the shift from on-prem to public cloud and sas is definitely underway but

7:36 it is not happening very rapidly however when we zoom into the data we see that the csp is in north america and

7:43 western europe particularly are the fastest in terms of moving their workloads to public cloud and sas

7:50 and we know that the number of partnerships between these csps and the public cloud providers have grown significantly over the past couple of

7:57 years most often for multiple aspects for their it transformations and more recently to support the joint launch of

8:04 their public edge computing services to enterprises and when we look at the chart on the right we see that

8:10 most csps in north america and western europe are reducing and consolidating the number of their traditional data

8:16 centers indeed many csps in north america such as atnt verizon canada or in western

8:23 europe bt telecom italia telefonica they all divested a large portion of their

8:29 data center assets over the last decade instead they've partnered with carrier neutral co-location providers such as

8:35 equinix digital realty and others for their data center needs so overall for

8:40 the western world while the 5g and telco cloud is absolutely important their main

8:45 data center focus is more skilled towards the adoption of public clouds multi-cloud and capex reduction

8:52 the story in the pacific region on the other hand is is a little bit different um these csps they show a stark

8:58 difference in their data center footprints footprint strategies and this is mainly due to um the historical lack

9:05 of strong presence of public cloud providers and high ip transit costs and some data localization laws

9:12 and many csps are making large investments in in growing and the number

9:17 and the size of their data centers in this region both for their itn network needs and there are some large csps such

9:24 as china telecom kddi or ntt they are competing in data center infrastructure

9:29 markets at a global scale and they're expanding their data center presence across the world however we also see that

9:36 the public cloud and the sas adoption is growing in this region and public cloud providers are rapidly expanding

9:43 their presence and service offerings in these new asian countries and we also saw that in some cases they are taking

9:49 advantage of csp's strong positions in the local enterprise markets to introduce joint offers in partnerships

9:56 with csps so in some i think we can conclude that the csp on-prem data centers and private

10:02 clouds will remain important and this is for supporting their both digital and the

10:07 network transformation ambitions but csps will need to operate increasingly hybrid infrastructures

10:14 where it network and enterprise applications will be deployed across multiple different data centers and

10:19 cloud environments well that's very interesting and gurkham what about the edge though do csps have

10:26 different strategies when it comes to edge data centers yeah it's a great question um so we

10:32 looked at edge data center investments um separately the operator's plans in

10:37 this research and by edge data center by the way we mean uh the smaller and more distributed data

10:43 centers located in csp networks such as south sides metro data centers or central offices

10:50 and in contrast to the regional variances we saw in the traditional data centers our

10:57 research shows that the many csps across all regions have very strong plans to invest in these new distributed data

11:04 centers at the edge so the majority of the csps are simultaneously focusing on both

11:10 enterprise edge clouds and this is to provide new iot operational technology or enhanced

11:16 enterprise services and also the network edge clouds to host their 5g functions such as vrand and openram

11:24 however our research and interviews also indicate that in most csp organizations

11:29 these network age and enterprise aid strategies are executed separately by different business units

11:36 today the enterprise hclouds are often based on public cloud provider stacks

11:41 and they're being deployed in regional and metro data centers and this is mainly to process certain types of data

11:48 locally and this is for security and regulatory reasons and when we look at the network hclouds

11:54 these are mainly located in cell sites or near south sides and this is to cater the needs of real-time processing

12:00 requirements of cloud-based ramp functions so there are some advanced csps are joining up or they are planning

12:07 to join up these different age strategies and deployments because there are going to be some new

12:13 enterprise use cases that require near real-time response and processing in the run but today these areas are fairly

12:22 separate activities and when we look at the number of edge data centers

12:28 the csps are planning to roll outs over the next couple of years

12:34 we see that the most csps that has an aids strategy are expecting to have between 10 to 24 h data center locations

12:42 and you may recall that that is kind of similar to the footprint of their existing traditional

12:48 data centers but we also see that there are several csps from north america

12:53 western europe and asia pacific they have more aggressive edge rollout plans with more than 50 or 100 edge locations

13:01 in the near term but of course this is just the beginning based on our general edge competing

13:06 research we expect that the age footprints feel likely to be much bigger in the long term and because enterprises

13:13 will require a more distributed edge native compute fabric that is available pretty much at every conceivable

13:19 location so this means that potentially hundreds or thousands of edge nodes in a country or region

13:25 so all this means that this sheer number and the scale of centralized and distributed age data centers will make

13:32 the data center network automation capabilities uh for managing this infrastructure even more important

13:39 okay so these key drivers and trends are for overall data center strategies

13:45 ben can you explain why data center networks and their automation are so

13:50 important yeah clearly as we've seen here data centers and cloud infrastructure are becoming increasingly important to sps

13:57 and we've all seen that as the application world is is moving to microservices and

14:03 kubernetes orchestration that the network is more important than ever to connect all these dispersed

14:08 workloads so data center networking and operations are critical uh sps are too

14:14 big and their businesses are too complex to be successful with these data center strategies that gorkham just went

14:20 through without a heavy dose of automation you know without the right tools for dc automation and

14:26 orchestration what we're talking about today just becomes a big operational mess and and sps have to learn to do

14:33 this themselves it would be a mistake to cede your strategy entirely to the public cloud providers

14:39 so given this importance gurkham what is the current state of csp data center and

14:45 network automation yes um so despite the importance of the data center network automation as band

14:52 just explained we see that the csps have actually made little progress in automating their data

14:58 center network operations so far so let me explain the results i'm sharing

15:04 now so we asked csps to rate their current level of automation for each of these

15:10 operational areas and we gave them ranges such as less than 10

15:15 10 to 25 percent 2050 so and so forth and we took the median points of the

15:21 aggregated inputs in order to assess the overall level for example um 38 percent

15:26 here means that most of the csps automated that particular area between the range of 25 to 50

15:33 and similarly 18 here means that it is in the range of 10 to 25

15:39 so when we look at these results we see that csps have made some strides in automating

15:45 relatively straightforward areas such as self-service portal integration and software upgrades and patching however

15:52 many key operational processes such as configuration root cause identification

15:58 and change validation all these remain highly manual today and csps do not really demonstrate

16:04 strong ambition or urgency to ramp up the level of automation in the near term either

16:09 we ask csps about their expectations for automation in two years time by the way which is

16:15 quite a long time for cloud and data center technologies that are likely to evolve and change rapidly in that

16:21 periods and csps expect to introduce automation to just about half of their

16:26 operational processes within this time frame and while their expectations for other key areas such as

16:32 root cause identification and telemetry are still low so all these results suggest that there

16:38 is a major gap between csps and hyperscalers large-scale enterprises and

16:45 the emerging network as a service competitors you know in terms of their current data center automation abilities

16:50 and then the future plans so we then zoomed into the day two plus operations which we found that many csbs

16:58 are facing significant challenges today and csps are generally struggling to automate fundamental day two processes

17:05 such as capacity management monitoring and assurance and service provisioning and change

17:10 so this will pose major obstacles for csbs when they are expected to meet their key service metrics such as time

17:16 to markets on-demand provisioning and slas including mean time to repair and uptime and this is especially true for

17:23 the more advanced services they're preparing to launch you know once their 5g networks and edge data centers are in

17:29 place and in our in-depth interviews uh most csps highlighted monitoring and

17:35 troubleshooting as their main pain points for day two operations and they especially highlighted the lack

17:40 of service intelligence and visibility across the multi-vendor network elements and this is due to the underlying

17:46 fragmentation of vendor proprietary data center automation tools and the data models

17:52 so what are the main barriers and challenges then to data center network automation

17:58 yes so according to our survey csvs are facing two major barriers that

18:03 are hindering their data center network automation investments and then the progress first

18:09 it's the lack of technology standardization and fragmentation in their data centers and the second is the

18:15 difficulty in justifying an opex driven business case for automation and in addition to these two barriers

18:22 our interview surfaced another barrier that is related to the data center ownership structure and organizational

18:29 mindsets so let me go into detail on each of these barriers and let's start with technology

18:35 fragmentation so many csps we talked to pointed out that their data center

18:41 networking environments are highly fragmented and very complex operates and automates and this is because their data

18:48 center networks consist of several vendors specific hardware as well as the management software tools

18:54 and there's already the existing heterogeneity of the traditional its switches and tools that are procured

19:01 from juniper cisco and arista but in addition to that ascsps cloudify their mobile networks

19:08 network function providers for taco clouds such as ericsson nokia and huawei they're also bringing their own or

19:14 partner networking components to csp data center and this further increases the technology and operational

19:20 fragmentation and in the cases where csps operates there are enterprise many service

19:26 customer environments they have additional complexity because each of these customer networks come with

19:32 different mix of network equipments so overall serving a wide range of services and use cases so i.t taco cloud

19:40 and enterprise and all these use a diverse set of underlay silos and overlay networks

19:46 makes it very difficult for csps to rationalize and automate their data center networks

19:52 as you can see on this pie chart csps often have to use multiple management and automation software

19:58 solutions and they are provided by a range of hardware and i.t vendors as well as we see the use of some scripting

20:05 tools like ansible or python and also the in-house ground solutions so all these disparate systems result in

20:13 high operational complexity and which is the main challenge um the csps associated with their data center

20:18 network automation technologies as the charts on the right shows and this is especially the case for csps

20:25 that mainly rely on single vendor automation solutions because they are typically focusing only on the

20:31 operations of their own network elements and the domains and moving to the second biggest barrier

20:38 which is the difficulty in justifying the business case for automation but before that i should explain what this

20:44 chart shows so based on the responses on the vendor technology choices the pie chart i

20:50 showed in the previous slides we categorized csps into three groups the first one is the csps that only use

20:57 third-party vendor solutions like juniper cisco and nokia and others for their automation needs and the second

21:04 group is the csps that mainly rely on in-house tools and the third group is

21:09 the the hybrid meaning that they have a mix of in-house and vendor solutions

21:15 and the reason why we did this is to isolate and understand the top technology challenges associated with

21:21 each group of csps so we see that operational complexity is the main problem for the csps that only

21:28 use the vendor solutions or hybrid environments with vendor plus in-house solutions which is what we discussed in

21:34 the previous slide but when we look at the csps that mainly rely on their own tools in-house grown tools and platforms

21:41 for automation cost and scalability are their main challenges and this may not come as a surprise

21:47 because in-house created automation requires developers and network engineers and operations people to spend

21:55 a lot of time and effort in building testing and integrating these custom diy

22:00 automation tools and the maintenance and the continuous development and the improvement of such tools can become highly expensive over

22:07 their life cycle so given the high cost of manual processes and in-house developments why

22:14 csps are struggling to build a strong business case for automation and this is because

22:20 the on-prem operations and in-house development costs are actually hidden so csp data centers serve multiple lines

22:27 of businesses and the costs are shared across these different teams and units within them

22:33 and also data center staff they often straddle multiple engineering operations and development responsibilities at the

22:39 same time so all these factors make it very difficult to have a clear-cut granular

22:45 view of the existing costs and as such csps cannot really carry out a healthy evaluation of the costs and the benefits

22:52 of new automation solutions and initiatives against the status quo

22:58 and the final and maybe the most important barrier is that csps have an entirely different mindset

23:04 to data center networking and automation compared to enterprises and hyperscalers

23:10 for example hyperscalers have a very much devops driven constant optimization

23:15 and automation approach to solve their data center problems but many csps are used to throwing more

23:21 people as human glues or they do expansive custom integrations or add-on tools to solve their operational

23:28 challenges i shared some of the quotes here that we got from the interviews it

23:33 basically reflects their their their thinking especially the first one on the left that highlights you know the

23:39 culture and the mindsets it's really fostering um these practices

23:44 and also csps have a very telecoms network-centric view most csps we talked

23:50 to about data center automation first thing they think about was network or service orchestration and

23:57 also they were thinking about things like nms ems or other tools they use in the wide area networks

24:03 so overall their automation thinking and focus is is primarily on

24:09 access networks or wide area networks and they tend to overlook their data centers

24:14 but given that the data centers are an integral part of their network transformation there is an evident

24:20 disconnect we are seeing in csps thinking between their network and service ambitions and the actual focus

24:26 they are putting in their data center network automation to support them it's fascinating thanks for that gurkham

24:32 ben what's juniper's reaction to hearing all this information

24:38 yeah there's so many important insights to unpack here in the last few slides um but but just a few things briefly the

24:44 juniper portfolio can certainly help sps here tremendously and what we've done is we've changed the

24:50 economics for an sp to build out and to operate their own cloud infrastructure their data centers and

24:57 it's really interesting how this survey found that sps fundamentally think about network automation in different ways

25:03 than an enterprise or hyperscaler does so juniper for years now you know we've

25:09 had our paragon portfolio of tools to automate the wan and we have a great solution for the data

25:15 center overlay and that's our contrail product that orchestrates cloud workloads and virtual networks and more

25:21 recently now with appstore we have a fantastic product to manage the underlay the data center fabric that we've really

25:27 focused on for this survey and philosophically appstore fits well with how juniper has

25:34 always approached the marketplace you know for example in terms of being open appstr is multi-vendor and pretty much

25:41 all sp environments today are multi-vendor and as gorkham said a lot of that fragmentation in the data center

25:48 networks is is a problem and abstract can help solve that problem and also if you think about reliability

25:54 you know historically this could be you know the one single word that you would

25:59 use to describe what drives a telco business and as it turns out reliability is is a key

26:05 founding principle of of apps and the idea that the blueprints that appstr is based upon essentially give

26:12 you reliability and consistency and this is what ultimately gives you speed of operations

26:17 and you know one of the interesting things is that that connection between reliability and consistency and speed

26:24 this is one of the key insights that we can all learn from the cloud hyperscalers who are masters at managing

26:29 their infrastructure so we bring that knowledge to the sp world

26:35 a final point here about understanding the business case for data center automation as gorkham

26:42 said the inertia of sps doing it themselves with tools they've developed in-house is tough to overcome

26:49 but but stay tuned in the coming months we're going to start putting out a lot of details around the savings that our

26:54 customers have realized after switching over to appstr thanks ben and gurkham what are the

27:01 solutions and recommendations for accelerating the progress of csp data

27:07 center network automation yeah well i think our research makes it very clear

27:12 that most csb data centers aren't really ready for the automation and agility needs of

27:18 5g edge and all the new enterprise or support changes we talked about and this

27:23 really needs to change fast so there are several things csps need to do both from

27:29 technology and organizational point of view and let's start with the easier one which is which is technology as i

27:35 mentioned before um and the benefit to it as well the high level of fragmentation and operational silos is

27:42 the really the biggest technology barrier to automation so csps need to change the way they

27:48 design and operate their data center networks by moving away from the traditional vendor-drone bottom-up

27:54 approach to a top-down paradigm and this means that csps should start from business intent and application

28:01 requirements first and the design and the choice of lower level hardware and software components

28:06 should be driven by these requirements not the other way around this is the case today and a key step towards making this

28:13 paradigm shift will be a creation of a horizontal end-to-end automation layer that collapses these

28:19 vendor-specific silos and of course every csp has a different starting point for their automation

28:25 journey and building this horizontal multi-vendor automation layer will not be an easy task especially over

28:32 this silo than the complex brownfield networks however having a truly multivendor

28:38 network automation platform will be essential for all csps in order to tackle this operational complexity and

28:43 fragmentation and as we can see here in this chart csps acknowledge that they strongly need

28:50 such a multi-vendor automation platform for their data center networks

28:55 and for the hardest parts which is organizational change there are several things to do

29:01 one of the most important findings from our research was that the cto is often the main budget holder for data center

29:08 networks not the cio unlike in many enterprises and in some csps data center

29:13 responsibility is dedicated to heads of data center network operations

29:18 so we know that the ctos typically have much greater focus on mobile and fixed networks than on data centers so the

29:26 fact that the ctos do not give strong backing to lower level budget holders for data center automation may be a key

29:33 factor behind the low priority and the low level of automation we just seen

29:38 and we dig deeper to understand these organizational and the mindset issues further

29:44 and one csp said that the complex and the continuously changing data center

29:50 ownership structure in their organization is one of the major inhibitors to their automation efforts

29:56 and another interview we mentioned that um despite having the budget to implement automation tools there is no

30:03 group dedicated to actually getting it done and the current stuff are just too busy with their daily firefighting

30:09 routines so all these issues should be addressed by c-level executives urgently as csps

30:16 are getting ready to launch their 5g and edge networks and ctos in particular should step up their efforts to

30:23 prioritize and accelerate data center automation initiatives and they need to think about how to restructure the

30:29 responsibility for data center operations so that they can be pretty much equal to other network

30:34 operations and they have as much visibility so they really need a long-term transformational automation plan that is

30:41 inclusive of all lines of businesses um it's going to be a critical part of the efforts and just going back to um the

30:50 business case issue one of the key broad block sets you know to to implement this

30:55 plan is going to be building this business case so csps really need to look at it um holistically

31:02 and this should be driven by their strategic objectives and and then the long long term plan

31:07 so and the first step they need to do is actually decomposing and mapping their data center network processes across

31:13 operations engineering and development in detail to uncover the hidden costs i mentioned before

31:20 and finally i would also add that csps might benefit from working with external partners in order to develop a framework

31:26 for this business case so that they can identify the realistic benefits and the costs of automation based on market

31:33 evidence and experience which can be really helpful to strengthen their business case all right thanks gurkham so ben can you

31:41 summarize the findings here and explain how viewers can access the white paper and also tell us what's coming next

31:48 uh yeah well you know first thank you telecom tv for hosting this session and thank you gorkham for the excellent work

31:54 we hope everyone watching has found this to be informative one of the things we're going to do

32:00 further down the road later this year is we're going to do another survey that digs down even further into some of

32:06 these key areas um but as a short-term follow-up for the audience please read

32:11 the reports created by analysis mason that delves even deeper into the the topics that we went through today and

32:17 then be sure to tune in on may 5th where we'll have a panel with analysis mason and special guests from juniper networks

32:24 to get their perspective on on all of this and finally thank you all for the audience for for listening in

32:30 well thank you ben and also gokum thank you both very much indeed and a reminder that you can download the white paper

32:36 now just follow the links below this video and as ben says don't forget to join us for our live q a panel

32:43 discussion on thursday the 5th of may just check the website for local times

32:49 and that's in just two days and we look forward to receiving all your questions

32:55 for now though thank you for watching and goodbye