Terry Sweeney, Contributing Editor, LightReading

DIY Data Center Automation Survey: Challenges/Opportunities for CSPs, Enterprises, & Cloud Providers

Data CenterNetwork Automation
Terry Sweeney Headshot
Still image with image of a black screen that has a red box in the upper right-hand corner that says ‘questions’ at the top with three pictures of people and copy underneath that is not readable.  Next to the red box is a bigger box outlined in white that is a video conference screen with a slide that says ‘Lack of multi-vendor support, cost of automation and lack of skills are the main pain points for DC network automation.  Then there is a bar graph showing ‘Lack of multivendor support/integration’ at 52%, ‘Cost/time to create and maintain automation’ 48%,’ Inadequate tooling’ 44%.

Data centers have become ever more complex and critical to most businesses. Automation is crucial to securely and reliably manage them at speed and scale.

Many organizations employ a DIY approach to automate their DC networking infrastructure using substantial in-house development resources and a variety of available open-source tools and technologies. These organizations went down the DIY path with expectations of cost and time-savings, but this approach often proves to be more expensive and operationally complex than using off-the-shelf DC automation solutions.

Juniper Networks teamed up with Analysys Mason to conduct an industry survey of CSPs, Enterprises, and Cloud Providers to uncover the reality of taking a DIY approach.

Show more

You’ll learn

  • How to approach the ‘to build or to buy’ question

  • How much of your data center operations to automate

Who is this for?

Business Leaders Network Professionals

Host

Terry Sweeney Headshot
Terry Sweeney
Contributing Editor, LightReading

Guest speakers

Gorkem Yigit Headshot
Gorkem Yigit
Principal Analyst, Analysys Mason
Arun Gandhi Headshot
Arun Gandhi
Senior Product Marketing Manager, Juniper Networks

Transcript

0:01 foreign

0:07 and thank you for attending today's webinar DIY data center automation we

0:13 die challenges and opportunities for csps Enterprises and class providers sponsored by Juniper before we begin I

0:22 will cover a few housekeeping items on the left hand side of your screen is the Q a if you have any questions during the

0:29 webcast please type your question into the Q a box and submit your questions to our speakers

0:35 what questions will be saved so if you don't get to answer you we may follow up your email at the bottom of your

0:41 audience console I multiple application widgets you can use if you have any technical difficulties please click on

0:48 the yellow help widget here you can find answers to common questions a copy of today's slide that is available for

0:54 download in the green jeans sources budget near the end of today's presentation please take one we need to

1:00 complete the survey that's open on your screen your feedback is extremely helpful an on-demand version of the

1:06 webcast will be available about one day after the event and can be accessed using the same audience thing that was

1:12 sent to you earlier today I would now like to turn the event over to live feelings contributing at the

1:18 Harry Sweeney Terry thanks Barbara and let me add my own

1:23 welcome to this DIY data center automation Deep dive Terry Sweeney here

1:29 and contributing editor with light reading and I will be moderating today's discussion

1:34 um to build or to buy that's been a question in Telecom and technology management for decades now

1:41 um whether you have 10 servers or or 10 000 the the landscape has changed with

1:46 the Advent of cloud computing virtualization Ai and machine learning

1:52 all of it is transformed Network management and made the question of buy versus build a lot more complicated

2:01 thankfully uh today's speakers gorkam unit of analysis Mason and Arun Gandhi

2:08 with Juniper Networks will really dig into this issue especially as it pertains to how much of your data center

2:13 operations you automate um today's webinar is also a departure from the usual presenter with a

2:21 PowerPoint setup that you may be accustomed to um gorkam and Arun will have more of a

2:26 conversation really about the findings of a recent study that Juniper commissioned an analysis Mason conducted

2:34 gorkam and his team talked a communication service providers cloud service providers and Enterprise

2:39 customers and the goal there was to peel back the layers of these different customer segments with regard to their

2:46 their thinking their planning and their spending where Automation in the data center is concerned

2:53 um but as Barbara mentioned we want to hear from you our attendees as well while gorkam and Arun chat uh please

3:00 send us your question using that q a box at the bottom of your screen which will address later in the program

3:07 so that's probably enough logistics for now um Arun Gandhi of juniper let's let's

3:12 start with you um talk to us why Juniper commissioned the study in the first place what what insights were you

3:18 looking for here thanks Terry and it's free to be on the panel with both you and gorkham and uh

3:24 to answer a question Terry as you know data center networking has become more

3:29 complex in the last few years and the added complexity needs to be managed the tools we've used before to manage them

3:37 no longer apply analysis Mason conducted an extensive survey for us last year and

3:43 that was focused on the ASP data center automation strategy and one of the key findings was that in-house automation

3:50 tooling is prevalent among organizations and most of them face cost and

3:55 scalability challenges with their custom DIY automation Tools in addition

4:01 in-house created automation requires developers and network Engineers to spend large amounts of time and efforts

4:09 building testing integrating custom DIY automation tools the maintenance and the

4:15 continuous development and Improvement of such tools can become more expensive over the life cycle

4:21 and and also last last service showed that uh csps with large numbers of

4:28 In-House DIY automation tools on average have higher cost spaces than their peers

4:35 that rely mainly on vendor automation Solutions so we kicked off this survey

4:40 project with analysis Mission once again to develop an understanding of In-House DIY automation data centers activities

4:48 the use cases and the pain points and this time we did it across all segments

4:54 so it's like a communication service providers Enterprises and Cloud providers

4:59 and the service survey results are very insightful and provide meaningful context to the mind of the data center

5:06 leader Kirkham let's let's turn to you now if we could um help us understand a bit

5:11 about how you went about collecting data for the the dentists that are study

5:16 itself what was your methodology here sure so we carried out an online survey

5:24 with a total of 96 organizations and these are made from three segments

5:30 you know including Enterprises cost service providers and Telecom operators and his own mentions um our objective

5:36 was to gain an in-depth understanding of the in-house DIY automation activities and reveal the main motivations use

5:44 cases and then the challenges behind them so we surveyed a large number of

5:50 Enterprises from a wide range of sizes and verticals you can see these on the

5:55 left hand side and the majority of these Enterprises they own between 25 to 50

6:01 data centers but even the smallest ones they have at least one or two our focus on the cloud service providers

6:09 was not on the major players such as AWS Google and Microsoft but rather the smaller scale more Regional type of

6:16 players providing cloud and data center services for the Telco operators these were

6:21 typically the large tier one operators with extensive mobile and fixed Network and data center assets

6:29 in addition to this online survey we also conducted deep dive interviews with senior decision makers from each of

6:35 these segments and you can also see on the right hand side that this was a global study with a

6:41 good balance between the Western and Eastern sides of the world and finally in terms of the titles and

6:48 responsibilities of the respondents we have a very balanced mix of senior

6:53 Executives who are in charge of their data center strategy and transformation of their organizations and also the

7:00 technical stuff from the production floor such as Network engineers and operations people who perform the

7:06 day-to-day operation activities in the data center so tremendous diversity among your

7:12 sample um not to mention like there's on on every continent from the

7:18 data center realm um stiving and with some of the the key

7:23 findings from the research um uh also curious about any similarities you may have seen among the

7:30 key drivers or the trends for data center strategies across these segments of course so one of the key findings

7:38 from our research is that the level of Automation in data center networks is very low across the board

7:44 and we were quite surprised by this because many of the companies we talked to stated that their data sentiment work

7:51 automation strategies are driven by ensuring customer experience security and business continuity and often

7:59 automating these network operations is one of the key pillars of their digital transformation strategy so despite this

8:05 strategic importance of data center Network or automation companies are struggling to make progress in it

8:12 and the main problem they are facing is that they find difficult to achieve reliable and repeatable Automation in a

8:18 cost efficient manner as part of this research we had a deep dive into how companies are building

8:24 Automation in their DC networks and we found that the majority of the DC

8:29 network operations today rely on a fragmented set of In-House Fields DIY tools and solutions

8:36 this indicates that the current DIY base approach DC Network automation is not

8:41 delivering the results companies desire and there's a strong and urgent need for rethinking these DIY automation

8:48 strategies so if you zoom in on the level of

8:53 automation achieved today we see some differences between the operational areas and also the type of segments so

9:01 we asked companies uh to assess their automation levels across key operational processes including phase zero planning

9:08 and design they want deployment and configuration and data and plus zombie operations such as monitoring

9:15 troubleshooting and so on we see that the companies the more progress has been

9:21 made in day two operations which are often a priority for companies because they too has a direct impact on customer

9:28 experience and company's reputation so think about the outages and slas

9:34 so this is an importance area but if you look at the overall automation levels it is still low

9:40 and there are several challenges um causing this and and we'll Circle back to that we found that day Zero

9:46 automation is lowest and many companies in our research reported that they're finding the data center design processes

9:53 highly challenging and difficult to alternate when looking at the segments Cloud

9:58 providers are the most automated among all three and this is expected because they need automated data centers and

10:05 operations to stay competitive however compared to the levels of automation we

10:10 see in major public Cloud companies these providers seem to be lagging behind and this could be a major risk to

10:16 their business Enterprises on the other hand are a mixed bag but we identified mainly two

10:24 types some Enterprises mainly see their data centers as cost centers and they tend to

10:30 limit Investments and usually outsourced operations to third parties but there's also other type of

10:36 Enterprises who consider their data centers as strategic assets and key part

10:42 of their business and digital Journeys so this attitude and focus makes a big

10:47 difference in the automation Investments and those that place more strategic importance on their data centers also

10:53 make more progress in DC Network automation and finally the Telecom operators

10:59 achieve the lowest level of Automation in all areas last year as I mentioned we

11:04 collaborated um on another research that focused on state of operators DC Automation and the

11:10 key finding was that operators are struggling a lot with technology and operational silos and complexity in

11:17 their data centers and this is hindering their ability to automate and now we see

11:22 that there's considerable gap between operators and other digital companies in terms of Automation and this might have

11:29 major implications to their 5G and Edge service ambitions gorkam I think would

11:35 be helpful if you could talk about some of the main challenges that are contributing to the low levels of

11:40 Automation in data center networks absolutely so our survey showed it um

11:48 there are three common pain points I mean DC networks today uh first is the

11:53 multi-vendor support and integration especially in Telecom operators and also in a large size Enterprises and to some

12:01 extent Cloud providers as well um this is a major challenge these companies often have to use multiple

12:07 vendor proprietary management and automation Solutions and most often they use it in addition to their in-house

12:14 Chrome DIY tools and all these disparate systems result in high operational complexity and

12:21 presents a significant challenge for automation and this gets even worse for operators

12:27 and Cloud providers that provide managed hosting services in which they operate customer networks and each of these

12:34 networks come with a different set of vendor technology environments so overall this is the top barrier and

12:40 the pain points are for automation the other pain points are actually interrelated the many companies believe

12:47 that the automation tools that are available in the markets are not adequate for their automation needs so

12:54 they decide to build their own automation tools and solutions as I mentioned before

13:00 um there's huge amount of DIY activities going on in data centers but then companies also struggle with the costs

13:06 and resource limitations of creating and maintaining these diagrams so there's a serious concern about TCO

13:13 and Roi of In-House field Solutions and we're going to talk about it later and

13:18 there's another aspect to consider here and this regards to you know the inadequate tooling because we know that

13:25 many Enterprises adopt Technology Solutions but they can't make the most out of them because they don't have a

13:30 skilled personal and the right operational processes in place so even though vendor automation Solutions come

13:37 with many features and functionalities a lot of them are underutilized or not utilized at all so it's not always about

13:44 the technology but how these Technologies are used and the processes are designed around them thanks cork I'm

13:51 Arun turning back to you um based on Juniper's experience with

13:56 customers um what part of the data center life cycle are organizations prioritizing for

14:02 automation what are you hearing there uh that's a great question Teddy uh now

14:09 if you talk to Data Center buyers overwhelmingly overwhelmingly they care about two things one is the Automation

14:16 and the other is security and you see the industry has changed but the past was defined by architecture which boxes

14:23 do you deploy and where what protocols to use to connect them the future is defined by operations and we aren't

14:31 scaling bandwidth so much as we are scaling up operations so and let me point out that uh the

14:38 entire industry has made automation a focus for at least last 15 years but

14:43 despite the millions of hours collectively it poured into the this Pursuit most operations outside of the

14:51 large Cloud Pro properties are still manual and I think God can uh alluded to those if you are generous maybe you'll

14:58 acknowledge how hard it is and if you are more cynical maybe you will side with devops Borat when he says to make

15:04 error is human to propagate the error to all servers in an automatic way is devops

15:11 and it's not just Borat breaking things and networks is almost lower at this

15:16 point ask your customers how comfortable they are pushing to production on Friday

15:22 so whether it's a supply chain issues or the great resignation or any other half a dozen or so Dynamics playing across

15:29 the industry the fact of the matter is the old way of doing things is dying the

15:35 trends towards automation is important but it is now about developing familiarity with the new host of methods

15:42 more conducive to automated operations gorkam turning back to you for a moment

15:49 um let's let's dig in a little bit more around um the the extent of data center

15:55 automation particularly with regard to this whole build versus by question that

16:00 we began with can you talk about some of the key drivers for organizations here

16:07 sure um so yes I mean the DIY automation is is highly prevalence and you can't see

16:13 you know some numbers here and this you know it's pretty much applies to all segments we surveyed the DIY is the main

16:21 approach and they're using to their DC Network automation and um we are seeing that these DIY

16:27 activities they use a broad range of Technologies for instance um software

16:33 automation platforms created by you know in-house you know the devops teams is pretty common and these are not just you

16:40 know the few lines of codes but they are like proper software systems with um you know many functions and and the features

16:46 and these platforms are typically developed and maintained by sizable amount of developer resources you know

16:53 and also with the support from skilled engineers and operation stuff and there's also a lot of Open Source

17:00 tools based DIY automation for example scripting tools such as ansible and agent-based solutions and

17:07 these are widely adopters and used by Network engineers and operations teams to automate their day-to-day tasks

17:14 but these are typically used in a firefighting mode as opposed to a consistent automation approach so many

17:21 of these are developed in an ad hoc basis so you can consider them pretty much as Band-Aids after Band-Aids to

17:27 resolve specific problems and a traditional method of you know the CLI and the shaft scripts are still

17:34 quite common and as we've seen you know of the DC network operations are highly manual so many companies are relying on

17:40 this fat finger error from approaches to manage their DC Networks and if you look at what is driving these

17:47 companies um to DIY automation the top driver is the ability to create tailored

17:54 automation for their specific operational needs and this is particularly true for cloud service

17:59 providers because they believe that they keep their technological Edge and the competitive differentiation through DIY

18:06 Solutions also the telcros they allow customization but oftentimes they regret

18:11 it because it just leads to another snowflake environment among many others they already have

18:17 and another top driver is is cost savings but here what they actually mean

18:22 is capex sales the companies we interview during our research highlighted that they find out of the

18:28 box Solutions usually expensive and they need more cost efficient and flexible commercial models such as space super

18:34 pairs to go and as the service models however as and we are going to talk about it later but this short-term capex

18:41 savings are actually offset by the Optics so there's a big question mark whether this DIY can really save money

18:49 and finally um a major challenge that we are seeing in

18:54 the DC networks is the multi-lander complexity and a lot of the companies are are having difficulties with that

19:01 and they're trying to solve these multimanded issues through the ioi automation so

19:07 and when we look at the challenges associated with DIY automation lack of

19:12 resource and skill availability and the dependency is on a small set of people who build these automations are the top

19:18 two pain points because the initial development of DIY tools and solutions is just the tip of

19:25 the iceberg it also takes large amount of valuable resources to manage the life cycle of these tools this includes

19:31 testing integration maintenance improvements security so these are all highly time and resource consuming

19:38 activities and then they face a turnover problem when people who build this code and have

19:44 a deep knowledge of these ad hoc tools and systems when they leave the organization it becomes a trouble

19:50 and on top of that there's the skill shortage um the CTO of an insurance company we

19:56 talked as part of this research told us that they're having difficulty finding skilled individuals who possess both

20:02 knowledge of managing data center networks and the ability to code for automation so the issue is not solely

20:09 that they are expensive but also that they are difficult to find and hire and the final challenge is is the ROI

20:15 and the cost of ownership and this is no surprise so if you will not we just discussed about amount of resources it

20:22 takes to develop the eye by Automation and also keep it functioning so there's actually a large amount of Opex involved

20:27 in the DIY automation and most often companies Overlook these costs or simply they are just not fully aware of them

20:34 because a lot of these costs are hidden and fly under the radar of senior Executives and the decision makers

20:42 thanks corkum I um I think it's safe to say that most of our listeners if not all are are interested in the actual

20:49 costs of DIY automation um can you give us an estimate of the

20:55 effort and resources required if you were to in fact build this yourself

21:02 of course so we analyzed two distinct groups of companies in our survey sample

21:08 the first group that consists of those that heavily rely on DIY automation with

21:14 minimum use of vendor Solutions and we call them DIY automation Reliance and

21:19 there are like 11 of them and there's another set of companies that rely more on the third-party vendor

21:25 automation Solutions than the average but they still use DIY to a lesser extent and we call them third party

21:32 leaning and they're like 10 companies in school so the first thing that jumped out of this analysis was that there is no

21:39 significant difference between these groups in terms of their level of automation so their DC network operations are pretty much automated in

21:46 the same amount however when we look at the level of effort and resources providing DIY

21:52 Reliance companies we see the true cost of DIY automation as this slide shows

21:57 DIY Reliant companies are dedicating a good chunk of their data center budgets to these activities and it includes the

22:04 head count so there's bigger effects right there and their Network Engineers are also

22:10 um facing a heavy burden by these DIY activities because the majority of the diyers they said that level of effort by

22:18 the network Engineers is very high and we see the evidence of this here most diyers actually they more than half

22:25 of their Network engineers and operations staff they write their own automation scripts in addition to their

22:31 you know daily tasks and almost every day they are creating and updating these scripts so this actually you know begs

22:38 the question is it really automation if you need to manage it every day or every week so if you look at the left hand

22:44 side the companies that have a better balance between the third-party Solutions and the DIY can achieve a

22:50 similar level of automation but without putting all these you know the substantial amount of In-House resources

22:56 so this clearly shows that a DIY heavy approach to automation is inefficient

23:01 and puts a lot of burden on skilled resources and doesn't deliver a strong Roi

23:07 everyone Let's uh let's turn back to you then um what is uh Juniper's reaction to

23:13 hearing this information um any any surprises uh among the findings that gorkam has just reported

23:20 uh thanks Teddy first of all I want to thank gorkam and the team for the great work absolutely incredible uh second as

23:28 we talk to our customers worldwide it is very clear that operational complexity

23:33 is driving up the data center Network cost and automation is needed to reduce this complexity lower the cost and

23:40 improve customer experience and the staff productivity and uh DIY automations have significant

23:47 hidden internal costs and cannot support a rich set of functionalities and scale

23:52 I think God can alluded to it in his uh research data as he presented and as I remarked earlier and you've

23:58 seen the results are very insightful and reassuring that our abstract can solve

24:03 many of the challenges that these automation DIYs have and just to remind

24:09 our viewers abstra is a multi-vendor intent base platform with Rich features

24:14 in Telemetry it abstracts uh control so we can manage a flick of devices from a

24:20 single point of control in a nutshell abstract makes it really hard to make mistakes

24:26 with data centers being decidedly multi-vendor and oriented around software defined operations

24:32 we do understand the challenges in managing them efficiently and this research affirms it across all segments

24:40 because I'm turning then back to you what sorts of recommendations do you have then for accelerating data center

24:48 Network automation it's done your research and the data that emerged yeah

24:53 absolutely so I would say first of all what this research shows that the level of PC

25:00 Network automation is really low despite the fact that it is business critical and has an impact on services customer

25:07 experience and company's reputation so there's an urgent need to take action especially the ctOS and the cios they

25:15 need they need to step up and prioritize investments in this area a best practice I heard are from a fair

25:22 automated organization was that they developed a multi-year roadmap to

25:27 specifically address on the DC Network Automation and they tied their senior

25:32 managers kpis and the bonuses to their automation objectives so this is a really good way to drive an

25:38 organization-wide automation initiative and this can be adopted by others too

25:43 the second not all DIY automation is bad in fact having in-house software

25:49 development and automation capabilities is a key part of becoming a digital company however our research shows that the

25:56 companies rely too heavily on DIY for their DC Network Automation and this is

26:02 not an efficient way to achieve their automation goals I recommend that companies rethink their DIY strategies

26:08 to strike the life balance so they need to examine their current processes and automation activities to reveal their

26:15 true cost of DIY and think about switching to a third-party vendor out-of-the-box Solutions wherever it

26:22 makes financial and operational sense and finally this means that they need to

26:27 identify and adopt the right vendor Solutions and based on the key pain points we found on DC Network automation

26:35 today there are some essential capabilities including their vendor checks firstly companies should look for

26:42 a truly multivendor automation platform because many companies struggle with operational complexity of multilander

26:48 environments seconds they need reliable and repeatable automation which means

26:54 that they shouldn't have to allocate large amount of resources and worry about people leaving the organization

27:00 one way to achieve this is having industry standards best practice automation templates for design day one

27:06 and day two activities so that they're not building snowflake automations but instead something that can be easily

27:12 replicated and extended across data centers and operations and finally of course they have to

27:19 consider the ROI and this should be based on a holistic analysis that not only includes the short-term capex but

27:26 also the long-term effects associated with the DIY automation activities Arun tell us how viewers can access this

27:33 research uh Teddy the report is available on

27:38 Juniper's site www.junippa.net and reports for each segment are going

27:46 to be there so for csps the Enterprise and the cloud providers we will also be

27:51 sharing the links with all our viewers after the webinar along with the recording of the webinar and you can

27:57 also reach out to your account team Juniper accounting and Juniper spot uh for to get this report uh you know in

28:04 your hands foreign

28:14 terrific context setting for the the very timely topic of the DIY Data Center

28:20 and automation um your conversation has sparked lots of questions from viewers uh before we go

28:26 there uh I just want to underscore what Arun just mentioned with regard to the

28:31 follow-up email for this webinar event that will allow you to view this webinar

28:38 again pass it on to colleagues they'll also be an evaluation form in there we appreciate your feedback uh this email

28:46 will also contain links to the the study being discussed here in today's session

28:52 so uh as a reminder to uh our our viewers and listeners uh typed your

28:58 questions into that q a box we'll get to as many of your questions as the time

29:04 allows um Arun there's a great question here from the audience with regard to uh

29:11 definitions and and scope of automation uh you you got into this a bit uh during

29:17 your your comments but um Talk a bit more about whether this includes is it just the network only

29:23 does it also extend to servers storage apps basically everything from bare

29:29 metal to Virtual Network functions how are you defining that

29:36 but very great question um actually um so how we Define is uh managing

29:42 automation the automating the entire data center which includes your uh your your switches your servers and all the

29:49 boxes and uh and certainly we're looking at the basically automation from the

29:56 operational point of view as I described in the in the uh earlier and I think also uh uh gorkam has wanted to share

30:03 some thoughts on this as well based on the research you got them you want to add something to it please

30:10 I mean I think yeah you you nailed it it's um when we talked to and also for the online survey uh we Define you know

30:17 the data center networks you know all the switches routers um firewalls and you know there's all

30:23 sorts of you know other um F4 and seven applications to the data center and of course I mean data center

30:29 networks are not isolated right all the servers storage applications they drive

30:34 you know a lot of um you know Automation and requirements on the data centers so yes basically uh

30:41 for the whole Space um we talked to in our conversations yeah

30:48 okay thanks gentlemen um another question from the audience um Arun I think this one is for you can

30:53 appstress support single fabric multi-vendor say a fabric with switches from vendor a and vendor B

31:02 uh yes uh the short answer is yes apps for certainly support abstra is I know

31:08 the uh the only multi vendors uh Network management solution available today uh

31:14 to manage your data center networks and it is multi vendor so you can actually uh from a single control point you can

31:21 manage your Juniper devices Cisco Arista and your uh you know the Sonic boxes as

31:27 well so so again yes it is and it was built ground up you know multi vendor

31:33 was has been a strategy from uh day one and we still continue to believe in and

31:38 we still continue to go forward with uh having this uh having multi-wind as part

31:44 of our key uh you know proposition for for data center Network management

31:52 gorkam I I was curious if you might be willing to unpack a bit some of the key

31:57 use cases for DIY automation for cloud service providers uh Communication

32:03 service providers and and Enterprises you you alluded to this a bit in your your presentation but um Talk a bit more

32:10 about some of the key use cases if you would yeah of course that there are some major differences

32:17 between these segments in terms of their DIY automation use cases if you start

32:23 from the communication service providers csps so their DIY tooling is mainly

32:29 driven by a multi-vendor network operations because they have highly fragmented Network equipments management

32:36 tools and and the data models so they need a common layer on top of these to

32:42 abstract you know this underlying complexity and most often these DIY tools are built for um like monitoring

32:49 service intelligence and and the configuration of these multi-mender environments

32:55 and in addition csps do a lot of DIY developments for the integration of

33:00 these multi-vendor devices to the other parts of their Automation and management systems such as like Network

33:06 orchestrators in their Telco clouds but we also discovered that advanced csps

33:12 they start to look at Automation in a more multi-domain way meaning that they

33:17 are considering more converged automation for their data centers when

33:22 and and the rest of their 5G and fixed networks but so this is still in the early days but there's definitely a

33:28 growing interest in this area cloud service providers on the other hand they're highly DIY driven

33:35 organizations and they pretty much build everything in-house it's just their model software

33:40 and buy so they traditionally have a limited use of vendor tools and they are utilizing all sorts of In-House and open

33:47 sources across you know their operations from provisioning to monitoring so there's no specific use case for them

33:53 you know for the DIY automation it's just a DIY is their primary way to build automation

33:58 Enterprises so the ones we surveyed in this research showed many similarities

34:04 to Cloud providers in terms of their approach to you know the DIY Automation and the DIY tools are widespread across

34:12 you know all key operational processes but of course there are some differences based on the type of Enterprises and

34:20 also their level of you know technical skills and technology advancements for instance some Advanced Enterprises

34:26 they're now building automation to tightly integrate their DC networking resources to their devops and CI CD

34:33 processes so that the developers and the product owners they can consume these resources on demand

34:39 we also heard um some of the Enterprises are going through on the cloud repatriation they

34:46 are migrating some of their workloads back to their data centers so they are rebuilding automation for these

34:52 applications and services and finally I would just add um especially in the highly regulated

34:57 verticals you know such as financial services and Healthcare there's a lot of

35:03 security and and compliance through and automation use cases we resoldering our research

35:10 in the conversations that you and your team had with these these various

35:15 customer segments um what changes could they would would they

35:20 make in retrospect or if they had to do it all over again what changes might they make to their in-house tooling

35:28 um what were some of the the common things that that you heard from the the survey respondents yeah so it's a great

35:35 question um one of the most important findings from our research that's almost no one

35:42 is fully content uh with their DIY automations and they would all go back and do things differently so this is

35:48 actually quite interesting um and when we look at you know the most automated csps and Enterprises they

35:56 generally said that you know they would use uh third-party vendor Solutions you know instead of like rewriting their DIY

36:02 tools and mind you that you know these are the ones that's been longer on the automation journey and they experience

36:09 you know all sorts of challenges and limitations of DIY automation so now they're realizing that in a lot of areas

36:15 actually they would be better off with an out of the box solution um similarly some of the least automated

36:22 Enterprises and cloud service providers they're also saying that in retrospect and they would use the you know the

36:28 right vendor Solutions instead of DIY because they started to acknowledge that

36:34 you know it is very difficult for them to get the skill sets and also you know enough resources to do it by themselves

36:39 so there's definitely a growing interest and demand uh from these type of

36:44 organizations for out of the box automation however um one thing I need to mention also you

36:51 know when we look at the most automated cloud service providers um they're a little bit different most

36:57 of them they are thinking about changing their approach you know to building their tools for instance and they're

37:03 thinking about dedicating more developer resources or maybe getting help from you know external partners and such as

37:09 Outsourcing so they will still continue to do automation in-house as much as

37:15 possible but I would say that you know best way forward for them actually really you know focusing on what you do

37:22 best you know to achieve competitive differentiation so concentrate you know all the DIY on the Strategic areas but

37:29 you should offload the rest of vendor Solutions because there's no need to you know to reinvent the wheel and you know

37:35 there are some solid Solutions in the markets that could do the same at the same level if not better than you know

37:41 the DIY tools and it would be probably you know in a much more efficient way to do things

37:47 thanks for that um I I think would be remiss if we didn't address uh the uh the the key

37:53 market trends of cloud and and 5G cellular on data center automation

37:59 um gorkam what's your take I've got a few ideas but um I'm just curious what you see is the the impact of of cloud

38:06 and and 5G on automation okay I'll I'll start um so definitely I mean cloud and 5G

38:14 they are the biggest Trends I mean driving the data center and also the automation um strategies today

38:21 um let's start with clouds um because you know the main Enterprises and and also the csps they're moving their

38:27 workloads to public clouds and also to SAS and in most cases um they're using you know multiple Cloud

38:34 providers however of course this doesn't mean that you know if they're giving up on their data centers and and their

38:39 private clouds um especially you know many Mission critical services and applications

38:45 um they're still running on-prem so overall this really means that you know they're operating in a increasingly

38:51 hybrids and multi-cloud environments and their networking and and automation needs are getting more and more complex

38:58 and this means that um they are facing new requirements and also challenges and all these are

39:04 basically you know redefining the way um they are doing Network automation and also they're thinking about how they are

39:10 securing these hybrid environments and and how they're ensuring service reliability customer experience and slas

39:17 so all of these are now you know in retails uh with cloud and in csps

39:24 cloud and 5G you know it pretty much goes hand in hand because 5G is designed to be Cloud native so there are you know

39:31 the major Cloud Transformations going on and Cloud native 5G networks age and

39:37 advanced Enterprise Services they will be all deployed and run you know across many centralized and distributed data

39:44 centers so csps needs you know highly automated and agile data center network

39:49 operations otherwise they won't be able to realize they are you know 5G Edge and iot service Ambitions in the past you

39:56 know the csps they often overlooked you know their data centers but now they

40:02 started to realize that you know actually their data centers are very critical to their business and Automation in the data centers is is

40:08 quick moving up in their agenda by the way you know of course I mean 5G and Edge is not only a key driver for csps

40:15 but also for you know wide range of Enterprise verticals you know think about manufacturing automotive oil and

40:22 gas and even in retail right so they're looking to drive more Automation and better customer experience using the 5G

40:30 on private networks and Edge so again all this infrastructure it relies on

40:36 their data center networks so automation office DC networks is an essential element you know of their digital

40:42 transformation I'm not sure if you have anything more to add on this well this isn't going to be much of a

40:48 debate I'm in violent agreement with just about everything you said um it's it's clear that that

40:55 um Cloud obviously is a bit more mature than 5G um but it's it's also clear that both of

41:01 these Technologies are bringing Untold about some bandwidth in in places

41:07 we couldn't get them before not to mention performance and flexibility

41:13 um but um also the the complexity that it also begins to introduce uh this this

41:19 is going to completely outstrip um humans own uh individual ability to

41:25 to manage the the amount of of log data and just to to keep current much less uh

41:33 keep things secure and operational so um it seems as if uh that they will

41:40 hasten actually the the the adoption of of automation it it seems impossible that they can get to the next levels of

41:47 of networking and service and and the customer experience that they that they

41:53 all want to get to without something more sophisticated than the tools they they currently rely

41:59 on so um let's see um we have another question from the

42:06 audience uh gorkam I'd like you to start on this one if you could um

42:11 what's your position on the evolution to to net Ops um the the network operations

42:16 uh equivalent to to devops uh using an open ecosystem of tooling

42:24 um the the questioner says that this doesn't exclude vendor-specific

42:29 Solutions but there is a scope boundary what the vendor platforms are supposed to do I EG not CI CD in a box

42:38 um what's what's your take on on netops is is is that something

42:43 um we can expect or should expect uh definitely this is an excellent

42:49 question and we are definitely seeing its run towards on the netsops and

42:54 and I I agree with that so um it's it has to be you know the open

42:59 source DIY along with the multimander tools right so you can't really argue

43:04 that you know you can just go with one way and don't think about other because as I mentioned earlier when I was

43:11 talking about recommendations it's really their key capability for any

43:16 digital um organization to have those in-house software skills and these are

43:22 often enriched with you know the open source and and the devops kind of tools so and also the best practices that we

43:29 are seeing for example from you know some of the most automated cloud service providers and Enterprises they basically

43:36 complement you know the vendor tools and solutions with open source and also you

43:41 know from their own developments and their own in which route of these open source tools so the ideal way to do it

43:47 is really you know create your um devops cicd pipelines and the chains your

43:53 workflows only integrated you know with this you know um complementary tools and solutions

44:00 because that's really you need to get the best out of you know the each all right so there's no need to just exclude

44:06 one and go with others so definitely um that's kind of thinking that you know we are hearing from the most advanced and

44:13 organizations okay Arun turning back to you

44:19 um it was curious uh Once Upon a Time abstra was only supporting uh EVP and

44:26 vxlan reference designs is this still the case uh good question Teddy this is a

44:34 question we often hear from folks who've seen abstra before so uh earlier let me

44:40 say you know maybe a year back or before but with uh 4.1.1 release in September

44:48 of last year abstra now supports free from reference design and with free form

44:55 customers can build Hub and Spock Ring full mesh and other topologies

45:00 leveraging any future protocol or architecture that fits the deployment scenario and now the network designer is

45:09 responsible for creating and validating all device configurations and as with other reference designs

45:15 freeform still leverages the context graph intent based analytics

45:21 configuration validation Loss Management time Voyager and other

45:26 numerous apps for software features

45:33 that's great thanks thank you for that um corkum turning back to you um just wondering with the research that you

45:41 conducted which segment do you think is most motivated about automating their data centers and I'm just curious why

45:48 you think that is I I I don't think there's one particular

45:55 segment that is you know the most motivated because for all segments you know data science

46:00 Network automation is is critical you know I just um about you know all the 5G

46:06 and Edge you know if I'm all the industrial now 4.0 use cases this is all you know very much reliant on on data

46:13 center automation maybe maybe cloud service providers and they're slightly more motivated because

46:19 at the end of the day you know data centers are their main business but we've seen you know some Enterprises and

46:26 csps that are as as motivated as this class service providers so it really

46:32 comes down you know to the individual organizations uh Within These segments how they think about Automation and to

46:39 what extent it is you know Incorporated in their strategy because I mean they're all facing a range of technical

46:45 challenges you know we discuss them today but a common characteristic you know in the ones that we could say more

46:52 motivated or or committed are those the one that you know putting also you know the organizational and operational

46:58 changes because you know it's not just about technology it's it's a it's a full

47:03 transformation so those more motivated ones um they are already putting they're

47:09 already devising you know long-term plans and road maps um they're getting support and

47:14 sponsorships from their cto's and cios um they're thinking about you know how

47:19 automation is going to redefine um the staff responsibilities and job descriptions and how they are going to

47:26 you know to allocate these resources you know to more strategic areas so it is you know I guess

47:32 um it is it really calms down uh you know to The Specific Enterprise and they're you know unique thinking but you

47:39 know there are some common characteristics you know in when we look at this you know bunch of um organizations that are more motivated

47:46 and more committed so this this is not a you know coincidence it really shows maturity in their thinking

47:52 sure um this this feels like a great place to to

47:57 also address um The Internet of Things um we were speaking about uh cloud and

48:04 and 5G just a few minutes ago um the the predicted explosion of of in

48:11 devices that will be uh made network accessible everything from thermostats

48:16 to light switches to every every operational technology piece

48:23 out there seems also poised to add to both the the the complexity and and the required

48:30 bandwidth for particularly Enterprise needs and um

48:35 gorkel I'm curious if you see this uh also uh becoming an issue for a

48:41 communication service providers but it's it's abundantly clear that um automation

48:47 will be absolutely essential for managing um what's exponential growth and and the

48:54 the whole iot sector of of the industry what are your thoughts there absolutely

49:01 I mean the the sheer complexity of all the networks on the services it's just

49:06 becoming impossible to do it in a manual way I mean you know if you want to also you know as an operator or as a service

49:12 provider if you want to you know deliver these Services um in a cost efficient way in a

49:18 competitive way and also you know a lot of people are you know asking for on-demand and Digital customer experience so automation it's not um you

49:27 know something nice to have but it's it's an imperative without that you can't really deliver these services in

49:33 the way that you know they're supposed to be definitely we had an audience member who's uh

49:39 following up on the the netapps uh threat that we've been on uh during the

49:44 Q a session um this uh this questioner is asking what you see as some of the main tools

49:50 that would fall under the the net Ops Banner um they're citing existing tools like

49:56 ansible puppet python toolkits or or something else um yeah

50:02 and how maybe netops May differ from devops with regard to tooling do you

50:08 have any thoughts there yeah yeah um yeah definitely so um for I mean when we look at what's

50:15 going on in data centers and the networks team they typically use um so

50:20 scripting using for like for example on ansible playbooks are are very common so

50:26 that's definitely a lot of network Engineers there using that and also the python um you know the how they base the

50:34 codes in the networks they seems like you know the de facto on the programming language the agent based tools

50:41 um server and storage people they tend to rely more on Chef and puppets but one

50:46 of the interesting things that we found um one of the data center operations

50:52 manager we thought to be we interviewed he said that initially you know they all started with this you know different

50:58 tool sets and everyone of us specialize on one thing but some something that they start to notice is that they

51:04 started to exchange you know these these tools and also their their knowledge you know their know-how so you know now that

51:11 networks people they are starting to use you know a potent Chef even though maybe it was more server and storage people

51:17 kind of thing but it is you know uh spreading into other areas and also uh I

51:22 I alluded this earlier um a lot of you know the networking resources and you know the devops people

51:29 and um and the product um owners they want to consume this networking resources as a service um you

51:36 know on demand so they have to you know integrate with the you know the broader devops and the CI CD pipeline

51:44 environment so it is converging um definitely so maybe you know the networks people use more ansible and

51:50 fights a little bit more but um other tools are increasingly becoming adopted than animals too

51:56 thanks for that um Arun in the in the just a few minutes

52:01 that we have left um you're in regular contact with with customers across all of these segments

52:07 that we've been talking about in the last hour what types of data center operators and and data center networks

52:13 do you think are are the best targets for the abstra solution

52:19 Great Commission there so um abstra it's just a general you know thumb rule is

52:25 abstract can be used across all data centers Enterprise Telco Edge data

52:31 centers a multi-tenant or co-location and hyperscale so scalability is really

52:37 an important consideration when you talk about the the platform or the tool and uh from the ground up it apps was

52:45 designed for scale I mean I've our first customer for example Yahoo Japan uh with

52:52 its initial release they actually deployed abstra across the data centers with hundreds of switches now of course

52:58 thousands and in some cases tens of thousands we build the architecture from the beginning to fit the largest data

53:05 centers so that we can scale with the customer as they grow the deployments and uh since when we talk about scale it

53:12 has two directions right so on the other side the smallest deployment that we can support with abstra is what we call the

53:20 collapse spine and which is essentially uh one spine connected to two leaves so

53:26 um in the nutshell the operational model scales with the size of the data center which is one of the key design goals of

53:33 abstract well unfortunately that is all the time

53:40 we have left for questions today um my sincerest thanks to both gorkam

53:45 yet with analysis Mason and Arun Gandhi of juniper Network's Light reading

53:50 appreciates your time and expertise on data center Automation and the smartest ways to handle that transition yourself

54:00 uh thanks to our sponsor as well Juniper Networks as well as to everyone in today's audience we appreciate you

54:06 joining us for this light reading webinar DIY data center automation Deep

54:11 dive thanks for your attention and participation we'll see you next time

Show more