SD-WAN Webinar - Beyond the Driver: Achieve Ultimate Performance with AI
Hear from the experts: How AI-driven SD-WAN delivers a premiere experience for IT teams.
Interested in learning more about the Juniper AI-driven SD-WAN solution and how it can accelerate your IT team’s responsiveness, deliver a premiere experience for your IT team and offer the economic value your organization expects? Then don’t miss this hour-long webinar on Achieving Ultimate Performance with AI.
Here’s your chance to hear from leading SD-WAN industry experts and have your questions answered. Ray Mota of ACG Research will dive deep into the SD-WAN market and numbers on a macro level. He will also discuss AIOps, cloud security, and DIY SD-WAN deployments. Peter Fetterolf, also of ACG Research, will talk about the total cost of ownership of SD-WAN and all the factors that play into it, as well as how the Juniper solution with Mist AI stacks up against the alternatives.
From anomaly detection to automated troubleshooting, you’ll see how the Juniper AIOps solution will create a better user experience across your entire enterprise. Yes, there is a cost to AIOps, but the cost of doing nothing is far greater. Register for a full demo.
You’ll learn
How to tackle network complexity with AIOps
Improve your IT team’s experience
Because it’s all about the business case –– learn how to estimate your economic savings with ACG’s new ROI tool
Who is this for?
Host
Guest speakers
Transcript
0:03 hi everyone please get settled we'll get started in just a minute
0:20 hello to you in virginia thanks for joining we'll get started in just a minute
0:27 kentucky all right excellent feel free to use the chat let us know where you're calling in from
0:34 north carolina omaha colorado great we've got great coverage across the states
0:39 looks like the canada's here excellent quebec city
0:44 lebanon fantastic got a great obvious audience
0:50 we'll get started in just a few minutes earth is here excellent glad
0:57 we have some visitors from earth turkey
1:04 mumbai excellent
1:11 all right good afternoon everyone my name is trish and i'll be your host for today as we discuss beyond the driver
1:17 ultimate achieve ultimate performance with ai this is brought to you by juniper and today's webinar is hosted by
1:23 vib our session for today will be approximately 40 to 55 minutes and we'll have about 10 to 15 minutes of q a
1:30 please be sure to use a key the q a button as we progress through today's presentation and for those questions
1:36 that we're not able to get to we will do our best to follow up with you after without further ado let's go ahead with
1:42 our first speaker for today ken delvo delevo he's the product manager marketer sorry
1:49 product marketing manager at um juniper networks thanks ken
1:54 yep no problem trish thank you so much for uh for joining us she said my name is ken delobo product marketing manager
2:00 with juniper's ai driven enterprise you know portfolio which includes our ai driven sd-wan solution
2:06 as well as gartner leading solutions for wired and wireless today we're going to talk about ai driven sd-wan and i'm so
2:13 lucky to be joined by uh ceo and principal analyst of acg
2:19 research ray moda ray is going to talk about uh the uh macro level sd-wan market uh and
2:26 numbers um as well as aiops uh cloud security manage diy
2:32 sd-wan deployments and and really dive deep um into the sd-wan market
2:38 we also have uh with us today peter federal who's the cto also at acg
2:43 research to talk about sd-wan tco all right the total cost of
2:49 ownership and all the uh different factors that play into you know making sure that it teams have the best tco
2:55 option when deploying you know their sd-wan you know solution um stay tuned at the end of today's
3:02 program uh we do live demos of the ai driven sd-wan solution show
3:07 we're going to share with you how you can get access to that as well as potentially win a oculus vr2 headset
3:14 by registering for our our upcoming live demo so stay tuned for to uh for the end
3:20 of today's presentation as trish said uh we will actively answer your
3:25 questions today about sd-wan the sd-wan markets total cost of ownership uh and
3:30 uh uh use the q a function that you have available um to uh to answer your
3:36 question and we'll do our best as trish said you know to get to that so um
3:42 let me step aside here and uh give the stage to uh to ray
3:48 ray uh we've got a great audience today i know we have a jam-packed agenda so i'm going to turn it over to you to get
3:53 things started uh thank you ken and and welcome everyone i didn't know about the oculus
3:59 offering so i hope i can qualify for that as well so hopefully there's an opportunity there but
4:04 i'm honored to be part of this managed services and sd-wan deployment considerations honestly has been
4:11 accelerating over the last few years what we're seeing is that enterprises are are really in a
4:18 situation where they're accelerating their digital transformation and in the past the discussion about
4:24 digital transformation was more slidework but we're seeing a lot of enterprises get to the point where
4:30 they're looking to figure out ways to accelerate and take these digital assets
4:35 and map them to economic or operational process whether it's enhancing the customer experience we're enhancing the
4:42 capabilities of that and when we look at whether it's an enterprise or an msp
4:48 this need to go kind of like beyond sd-wan and be more efficient as a
4:53 requirement where uh we're looking at and getting a lot of calls from a lot of enterprises saying we need more than
5:00 just a traditional se1 how do we implement what we call the next level of
5:05 ai or opportunities with automation so so again i'm honored to be part of this
5:10 and peter federoff uh joining us and and really when we look at uh 2020 we can't
5:16 have a discussion without looking at the macro climate from a point of view but there's opportunities where if we look
5:22 at 2020 2020 was a year that many of us will never forget right but one of the
5:29 interesting things i tell people that in my family when i used to tell them what i did they didn't quite understand but
5:35 now telecom and technology has become the lifeblood of our communications and if we look at
5:42 what's happened because of connectivity we've been able to continue business continuity from that area despite some
5:49 of the economic people wanting to stay connected from that point of view and and what we're seeing is for
5:56 an overall perspective is that it is looking promising so far we see based on
6:01 the imf the gdp is projected to be up the next three years this year about 5.5
6:07 and in 2023 uh about 3.2 so what does that mean from an enterprise point of
6:13 view it means discretionary income more positive spending and more opportunities where we see from a wire line and a
6:20 wireless perspective we're seeing catbacks also projected to be up we're
6:26 also seeing that even the cloud providers and what i mean web scalers those are the cloud providers whether it's amazon
6:32 google you know all those tied to calling the cloud titans are projected to increase their spending as well as i
6:39 t spending about 5.5 so so spending going on across the board wire line
6:45 wireless now there is some uncertainty regarding where the existing war uh and some of
6:51 the supply chain issues and that's something that we're monitoring close but from a gdp perspective and spending
6:57 is pretty positive right now when we look at enterprise right one of the things that we look at is what's
7:04 driving the enterprise mindset today and what's driving the enterprise mindset when it gets to sd-wan and if we look we
7:10 talked about over the last two years that because of what's happened regarding the pandemic
7:16 you we see a lot of companies that traditionally had let's say 50 branches because of this new hybrid
7:23 work at home go to 500 branches and they needed to maintain
7:28 compliancy with corporate from a security connectivity and honestly there's a lot of enterprise cios
7:36 and ctos that i speak to that says this is going to be the new norm for us right so so we need to make sure we maintain
7:43 customer experience as a top priority because we spend so much money to acquire these customers we need to make
7:50 sure that we're able to deliver and maintain that but at the same time we can't compromise we can't compromise the
7:57 threat on security and we need to make sure that we're able to meet compliancy from a corporate point of view
8:04 and now what they're seeing is is that for the first time in a long time i used to have a lot of enterprise and even
8:10 some msps coleman says i wish we had more automation i wish we had more visibility i wish we had more
8:16 control so now is an opportunity because of the spending that
8:21 a lot of these enterprise customers want to have a level of flexibility and control as well as implement what we
8:28 call cross-domain uh intelligence from automation and what they need to
8:33 understand is that acg did a study that we can share is that the traffic's changing is
8:39 becoming unpredictable used to be that traffic was a one-to-one we could predict it but now with uh iot devices
8:47 and these other areas where they're communicating to different parts of the network we're seeing a shift that is not
8:53 going to a centralized area because of some of these applications it's moving closer and closer to the customer so
9:01 we're seeing traffic shift really close to the metro or even at the
9:06 customer premise but even from a wireless point of view uh acg did an
9:12 analysis where currently uh 2020 mobile transport users was about 7.2 gig and
9:19 projected to be about 25 gig so so as an enterprise you need to be prepared and understand that this is a global
9:26 situation and how do you address i always you know when i was a former cto and cio
9:32 in the enterprise people said what's the one thing that uh you could tell me when i joined the cio board is that there's
9:38 constant change so people say cio role is changing no that's part of the job
9:43 it's constant change and you need to address what's going on from that perspective moving forward there
9:49 so i mentioned security as a top priority i looked up some numbers from our analyst
9:55 team and this is interesting we see that you know ransomware and the whole thing associated is becoming more and more of
10:02 a priority we see that in one second one identity was stolen that's just one second one identity that within one
10:09 second two websites have been hacked and within one of those seconds five malware
10:15 have been deployed now that doesn't sound like much right but when we look at it from a day perspective that's over
10:22 86 000 identities stolen in one day 172 000 websites uh uh hacked uh in one
10:31 day and over 432 malware deploy so the magnitudes of scale
10:37 for security is becoming a big priority and what we're seeing is that
10:42 these attacks are becoming extremely sophisticated we're seeing things like black clouds
10:48 where you know the way we see the value of traditional cloud cloud compute all that
10:54 there's clouds being developed for security and a lot of these enterprises are in a situation where for what
11:00 they're doing is they're building right and developing a group where you see the ciso becoming a more important
11:08 part where they used to report to the cio now is such a priority that they're reporting to
11:15 to the ceo right and as you add more tools whether it's wearable
11:20 where there's open source or implement things like virtualization which brings some economic value but it potentially
11:27 if you don't bring in the right orchestration system and management system you create um
11:33 a situation where you make the network potentially more complex so as enterprises address these pain points
11:40 they need to be able to make sure they they look at visualization from the standpoint to say i need to look at
11:47 what value brings to my customer but i need to look at it from a situation where it doesn't compromise the
11:53 application performance still maintains security and is able to scale so that
11:59 any remote user is able to connect agnostic of device agnostic of location
12:06 agnostic of the application they need to be able to feel like just they're sitting right there in the office so so
12:12 these are these are some of the challenge and what we see is that sd wan is becoming an important play
12:18 in that area but what i wanted to do is the term sd-wan in many cases for a lot
12:24 of people it's overused and they don't know exactly what it means so the way we
12:30 define it in a good simple way is that you look at a smartphone right is that
12:37 if you have a smartphone and you want let's say facebook application or you want a phone
12:43 application or twitter application um you don't want a situation where you're
12:48 buying a different phone for these applications so sd-wan becomes an opportunity that
12:54 allows you to create a platform that brings the level of visibility and control
13:00 that makes it flexible for people transitioning from sd-wan to to an opportunity where they want to
13:07 either do a mix but it gives them the flexibility that when they're in this platform just like a smartphone they
13:13 have the compliancy security and all the connectivity that where they log in right so you want to look at sd-wan
13:20 overlay what we call is as a platform for innovation right and this overlay is
13:27 able to connect into that area right and give you the capabilities to
13:33 make the connection whatever the connection is you have control of visibility from a central location so
13:39 you can do zero type deployment you can create automation have centralized policies and if you're an msp increase
13:47 margin if you're if you're um an enterprise customer you're able to
13:53 address as moves and changes more quickly and be able to bring up branches on the fly based on let's say a
14:00 wireless connection and be to a connect agnostic of the endpoint so so it's a
14:06 proven area it addresses the multi-cloud piece and the multi-application
14:12 piece but there's a lot of misconceptions on what sd-wan is right but then what sd-wan allows you to do
14:18 is to bring in additional value added application whether it's security load
14:23 balancing session border control and really take true advantage where there's an ip set connection to
14:30 interconnect but create some type of supply chain to be able to connect applications right into the lan area
14:37 more effectively and if we look at from an industry point of view it's a global impact right we're
14:44 seeing i mean the us or in north america is the biggest play we see that it represents
14:50 about 50 percent of the market but acg is projecting about a 45 percent growth
14:56 in 2022 continuing in this area in about a 3.6 billion dollar market and the and
15:03 and the biggest use case that we see is is hybrid land right where uh what
15:08 we're seeing in the particular piece where we have a customer that has let's say
15:14 two mpls links right one is a primary one as a secondary link it's not
15:20 initially that the primary link is going away it's on the backup link where they
15:25 like this is where we could save money this is where we have flexibility and in this area we create this hybrid line
15:32 that they connect either a layer 2 connection or a 5g connection or whatever
15:38 the type connection and be able to connect like cloud port more effectively the other use cases associated with
15:45 application visibility and intelligent routing that brings a software component
15:50 where before when you look at a lot of the as moves and changes the enterprise was
15:56 at the mercy of the msp to make those changes right
16:02 and then another use case which is hot the third is this internet breakout to sas where instead of having a bunch of
16:08 what we call lollipops on a stick to connect to these cloud applications
16:13 you're able to connect to them and you don't have to backhaul all the traffic
16:19 to the corporate location you could interconnect within the sd-wan and put a layer of security to do that and then
16:26 the last piece depending on the vertical is what we call fast connectivity for remote users so we see that for retails
16:33 that want to pop up a shop for all refineries so so it's a large market as a market that's going uh very quickly
16:41 and if we break it down into the the areas is that the business of
16:47 cloud compute uh in in applications being developed whether the cloud native is over 95 of the applications being
16:55 developed today are developed cloud friendly or developed in a cloud right so it's an area that you need to make
17:02 sure that you're able to address the clout and the mobility aspects of connecting to these
17:08 applications but at the same time have the security and have the simplified
17:13 architecture so that if you're doing a merger and acquisition
17:19 all that it's easy to on board that acquisition but you need to have some cost management and improvement on the
17:26 performance to that piece so so those are key drivers but i think the most
17:32 successful deployments that we see whether it's an msp or whether
17:37 let's say it's an enterprise is don't let just look at sd-wan from what we call a common offering
17:44 if you're a health care provider look at the initial additional features that
17:51 that let's say a nasty wine solution gives you right so uh from that point of view
17:56 retail and and if you're an msp you want to make sure you're having the discussion as well properly right so
18:02 that when you're talking to a retail shop you know sd-wan gives you the operational simplicity application
18:08 visibility offloading mpls traffic cost reduction but take the discussion a
18:14 little bit further with your enterprise because they'll appreciate it and you need to make sure that you understand
18:21 when you use the word compliancy right is you really can't go
18:27 to let's say a retail uh or let's say a health care and say hey we're hipaa compliant what you
18:34 should do is change the discussion to say we offer a platform that enables you
18:40 to be hipaa compliant right and then you can work with them to help them get along that path so your
18:47 solution should be based on if whether you're in enterprises each enterprise has a
18:52 diversity of what the customer personas is and when you're having the discussions with whether it's a vendor
18:59 or an msp challenge them on these additional specific for your verticals and the ones
19:06 that do that from an enterprise are having more successful deployments the ones that don't are finding it more and
19:13 more difficult to deploy uh sd-wan
19:18 and and what you should do right is look at twofolds whether if you're an msp you
19:25 should create an architecture that builds on what your service profitability and time to market is
19:32 if you're an enterprise it should be focused on what are the right tcos in our eyes to develop this so this is an
19:39 architectural model that we've used that you can edit and feel free to use any way you want is that we have the
19:45 existing underlay network on whatever the connectivity type that you have but
19:50 then decide is what are the endpoints what are the cpe devices what architectures what
19:57 work at home work from home strategies what are the edge and map it to the
20:02 underlay but don't stop there right look at st when to be able to connect cross domain
20:10 to those platforms and take it a step further to the applications whether they're virtualized services for for
20:17 edge computing and when you do this right you're able to map the vendors on the left side and then on the right side
20:25 be able to map what the right tco offering what the right roi what is the
20:30 customer experience and what is the architecture so this is a an architectural map that you can use to
20:35 leverage that and and in a lot of the discussions that i have with some of the larger enterprises like ray i have
20:43 automation why you talk and when we look at it the problem is is that what we see
20:48 is they're not taking automation all the way through they have silos of automation and in order to maximize
20:57 automation right you need to be able to look at it from across platform cross-domain perspective and
21:04 cross-organizational to be able to actually maximize the op-acts that's being delivered so our
21:09 priority should be is how do i remove these silos of automation and this is where ai comes to the rescue and the
21:16 opportunity of ai uh in that piece now we we've talked about ai ops in the past
21:23 right and automation in the past and the barriers have been with automation with certain people is like well i feel that
21:30 it lacks security i feel that it lacks some real time intelligence and real-time reporting to
21:37 for me to be able to address that uh and this is the problem where it's still
21:42 a situation where you need to bring some type of ai capability to be able to maximize the delivery and peter will get
21:50 into some of the uh economic models of that case to maximize ai we've been talking about
21:57 aips for a long time but now that you can run a ai in the cloud where there's machine
22:03 learning it gives you the flexibility to decide what input parameters you need
22:08 and be able to optimize your your delivery and and if we look at ai based automation the tm form kind of broke it
22:16 down into four uh components from customer experience to reducing the outbacks on closed loop
22:23 automation you know preventing outage right with the predictive is more like i
22:28 used to manually when i used to install frame relay i used to track lmi links and that was my automated way that
22:35 doesn't scale properly right so really we need to be able to have situations that we can track things and
22:42 be alert before the problem happens that by the time you're the enterprise customer you're calling
22:48 the msp already knows there's a problem going on they've opened up a ticket and they're working on it so that level of
22:55 efficiency impacts the customer experience and this is the key role that ai plays in this
23:02 particular area and one of the things in a uh juniper i'm sure i could share the papers we did a
23:09 detailed uh paper from a technology point of view about the use of tunnels
23:15 uh from that area and the advantage of tunnels because what ends up happening it becomes in economies of scale when
23:22 you have too many tunnels for the applications and it might not apply for all verticals
23:27 there are some verticals that the way they solve the problem is just get a bigger
23:32 pipe you know just get a bigger pipe and they over provision or kill a fly with a sledgehammer but if you're an enterprise
23:39 that's not in that situation where you could just continue to get a bigger pipe you need to figure out is how to
23:45 optimize my land by not having unnecessary overhead with these
23:50 particular applications so this concept of having tunnel free technology becomes
23:56 an important piece and bringing ai in order to be able to optimize your wan offering is key right and probably the
24:04 number one question i get from an msp is what are the
24:10 main pain points that enterprise need to understand for me well you need to understand the propensity to buy and
24:16 willingness to pay for your service but then enterprise come to me is when
24:22 and where do i decide in a customer segment depending where i'm at right is when should i go to an msp right so
24:30 an msp you need to drill them on the concept are they able to give you
24:36 visibility and a level of control if they're able to give you visibility and control that's an important part but
24:43 then on top of that you need to make the decision do you build it yourself right or do you go to an msp to
24:50 outsource that a lot of that sometimes depends on the skill sets you have and the problem that you want to solve but
24:57 regardless of that what we've seen is enterprise that have been successful is that if that msp is able
25:03 to give you 15 to 20 savings then do you doing diy
25:09 that's where it makes sense it depends on the customer characteristic you have so
25:14 i don't care it doesn't matter whether you go diy or msp is more important depending on what type of customer do
25:21 you do you have the skill sets and if you don't do the analysis whether you decide whether to work with the msp to
25:28 build what we call a co-managed service where you co-manage it together or they
25:33 fully manage the service for you because you don't have the skill sets and you want to focus on your core competency so
25:39 understand where you are in the pyramid understand where you are in the skill set and make sure that you're able to
25:45 get savings versus diy and what i'll do is that i'll close with
25:50 the uh well actually there's a few parts i want to get in because i want to do spend some time for a quick poll
25:56 is that if we look at present mode of operation today even sd-wan
26:01 currently what we call 1.0 there's still a lot of capital at intensity in order to maintain
26:07 we looked at some of those gdp numbers if you stay in prison mode of operation capital intensity which means how much
26:14 capital it is to be able to deliver my service is projected to increase 17
26:19 so there is a cost of doing nothing and the idea is to say yes you took the
26:26 first road to automation now i need to look at how do i bring in a level of ai
26:31 ops to be more efficient to create operational efficiency to maximize my roi
26:38 and more importantly create the visibility uh that meets my customers needs and our customers customers needs
26:45 from that perspective so maybe uh uh trisha if you could do the poll question now
26:53 if you can take the time the question is what is your preference for for sd-wan manage code manage
27:01 diy and what we could do is take a look at this at the end if you could answer uh
27:08 the the question and we'll take a look at paul at the end and and share the outcome so maybe for now what i'd like
27:14 to do is is kick it over to dr peter fedorov to get into economics
27:20 yeah thanks a lot ray and uh thanks everybody for being here so uh yeah we've uh uh been doing some work uh
27:27 we've actually been working with juniper for many years and then most recently we've been doing some work modeling uh
27:35 you know the solutions with the ai ops and then comparing that to similar
27:41 solutions without so if you look at the left hand side what we've looked at is a juniper solution which is uh the ssr
27:48 routers which are the session smart you know tunnel free sd when routers with
27:54 win assurance and win a insurance is basically the product name for the ai
28:00 ops applied to the sd-wan and then there's also some other components so there's sd-wan but then
28:06 there's wired lan and wi-fi as well and those are using the missed ai products
28:13 and and missed ai is really also the basis of what's used within win assurance so the basic idea is we've got
28:20 sd-wan we've got uh you know the wired land wi-fi infrastructure it's all cloud
28:26 managed and using ai ops to basically do automation and problem reduction then we
28:31 compare that to an alternative solution again with sd-wan wireless and wi-fi but
28:37 no ai ops and also no cloud management so basically what we found um
28:44 you know using assumptions and data we've gotten from the field is that you know there's some significant opex
28:50 savings uh so ray can you move on to the next uh slide so just as a little bit of an
28:56 introduction here uh because of the terms so the win assurance really applies to the
29:03 juniper ssr routers and then missed ai is is a term that applies to the the
29:10 wi-fi access points as well as the whole set of juniper uh ex switches so if we
29:17 look at well what is it right it's really um it's really ai ml machine
29:23 learning um there's you know the data science data is basically ingested from multiple
29:29 sources so so the you know the sources that exist out there and network including wi-fi access points switches
29:35 routers etc is where the data comes in the whole thing is cloud managed and and then the ai engine of
29:43 course is in the cloud um and then additionally there's a feature called
29:48 marvis which is basically an ai driven virtual network assistant you can communicate with kind of via texting and
29:55 it's like a natural language uh kind of communication um you know to to present
30:01 problems so what does it do for you so the key thing is you know simplifying the operations dramatically
30:08 you know reducing also the level of the skill levels uh required on automating a
30:15 lot of troubleshooting as well as remediation and there's options for automatic remediation or for you know
30:22 the uh the text to be involved in remediation and clearly this is going to be not only reducing opex but improving
30:30 uh the customer experience and moving towards say the self-driving network so that's just a little background so ray
30:36 you can move on uh to the next slide so acg what we've been doing economic
30:43 models for many many years and about five years ago we decided to make a big investment in a
30:49 software platform uh to do economic modeling why we do this well first of
30:54 all you know when every we're building just like everybody else in the industry building these models in excel and
31:00 they're basically these silos so you build a model for 5g or for open ram for sd-wan or whatever and they're all
31:07 customized logic data all different models not anything
31:12 reusable they can also have bugs in them and difficult for people to understand and use so what we've done is we've
31:18 developed a platform we call the business analytics engine or the bae
31:24 which is a common platform for all the use cases and then we have libraries of
31:29 reusable components and libraries can include services they can include
31:34 applications it can include routers etc all kinds of devices and then we can
31:41 in a visual way we can build these models very quickly and easily and then they have a lot of credibility with the
31:46 customers so juniper has been using now the bae for uh quite a few years um and
31:52 we've used it for white papers in fact we have some new white papers coming out and then using juniper can use this bae
32:00 tool to engage directly with customers so if you if you do would like to have some engagement on that kind you should
32:06 follow up with your juniper sales rep and ask them about it okay ray can you move on uh to the next slide so okay so
32:14 what does the model look like so again we've got our bae platform there um what's going into it so we've got some
32:20 basic network architecture input assumptions and we've looked and and we've done a couple models we've looked
32:26 at msp model as well as enterprise based models the one i'm going to discuss today is the enterprise centric model so
32:32 we looked at took a typical example where we have some small branches large
32:37 branches headquarters campus et cetera and a set of network elements then we have expense
32:44 assumptions in terms of hardware software etc and then probably most importantly is our opex assumptions
32:51 again acg has a very rich set of data experience on opex side as well as we've
32:58 been able to pull data from the juniper uh customers in both day really day zero
33:04 day one and day two uh opex and then all that goes into the bae platform we get
33:09 some very rich output so ray can you move uh now to the next slide so this is
33:14 just a screenshot of some of our input so this is for the current network we're looking at right now and you can see
33:20 we've got the network growing so we start with the initial value go grow to a terminal value we grow that be an s
33:27 curve and this is over a five-year period so we're looking at network growing
33:33 branch offices growing from 100 to 2000 a single corporate headquarters and then large branch offices growing
33:40 from one to 50 and then we also consider a campus so that way we look at both indoor and outdoor and a fairly uh
33:48 diverse uh environment so ray could move now again to the next slide and now this is
33:54 what uh you know we're we're assuming in terms of you know devices so we have you know ex
34:01 switches we have indoor aps outdoor aps you can see you know where they are on
34:06 the campus or in the corporate headquarters and the the the large branch office or the small uh branch
34:13 office now one thing that we have assumed here is uh for the campus the corporate headquarters and the large
34:19 branch office we're assuming an 8j or high availability architecture for the
34:26 ssr routers that means we have two of them um so physically you have two devices but
34:32 then the the software licensing fees aren't multiplied by two so they're they're
34:38 they're sli they're different for the 8j than they are if you just say multiply that by two but you do actually have two
34:44 physical hardware devices for the small branch office says we did not assume that there's h a we figure we just need
34:51 need one router there and you can see kind of the numbers that we're we've assumed in our model so ray you can move
34:57 now uh to the next slide now okay again said opex was a big thing so there's day
35:04 zero which is basically the planning and design day one which is basically the
35:09 system deployment and installation right and then day two is the ongoing you know
35:14 essentially managing fault management performance management and remediation
35:19 and we're looking at you know with ai ops without ai up so some of the very
35:25 specific categories of labor we have there on the right hand side so you can see those categories of labor and we've
35:32 used some of our acg data as well as with data we've gotten through juniper
35:38 to look at that and and looked at the reductions we could get and and what that amounts to so uh ray please move uh
35:46 to the next slide so basically these are the top level results and like i said we're having a couple detailed white
35:52 papers that are going to be published shortly within the next month or so so i'd encourage you if you if you want
35:59 like a lot more detail to be able to go there you can go to either the acg site or juniper site or ask you know your
36:05 juniper contact but you can see the kind of savings we're looking at is pretty significant opex savings again this is
36:14 due to essentially reduction in labor also reduction and skill sets required now
36:20 that doesn't mean that those people are necessarily going to get laid off or anything because there's plenty of work
36:26 to do inside of the enterprise there's a lot of strategic projects there's all
36:31 kinds of stuff in digitization so instead of having people busy you know chasing down
36:37 uh problems and and trying to you know fix uh bugs and performance issues it's better
36:44 to have them working on on really more strategic uh initiatives and you know
36:49 there's mergers and acquisitions and all kinds of problems that need to be solved so why work on stuff that can be automated um
36:57 and you can see you know the table below shows the actual numbers there for capex
37:03 opex and tco and and for the specific example so ray you can move now to the
37:08 next slide and this just shows you again in this particular network we were assuming that it was growing over time
37:15 in some cases the network might be static but it's nice to look at a growth kind of scenario and you can see year by
37:21 year this is basically tco where we look at opex and depreciation again this is a
37:26 screenshot you know from our bae um and and so uh you know you can see how that
37:32 as as the network grows uh the key thing from this slide is basically as the
37:38 network grows uh the benefits become become better over time so uh ray you
37:45 can move on i think that's the end of my slides there so um yeah anyway thanks
37:51 everybody for joining and i would encourage people to check out the white papers they will be available um in in
37:57 the next few weeks thanks peter and you ray uh for going through this i think let's um let's take
38:03 some time here to to answer some questions you're from the audience we do have a lot of people on and a lot of uh
38:09 you know questions regarding the material that that you just covered um if you're comfortable uh with taking
38:14 some questions okay great um i did
38:19 actually before i you know we even get into that um uh ray you mentioned some research um
38:24 regarding tunnel free sd-wan the benefits and advantages of tunnel-free sd-wan versus a tunnel-based sd-wan so i
38:32 did put a link to that in the chat oh good oh good yeah to the paper all good we had access to that so that's
38:37 available for those of you that want to have access to that that's free um so you don't even have to register for that
38:42 so you feel free to access that and there was a question regarding sd-wan architecture i did answer it in the chat
38:48 but yeah i'm not sure if there's anything that you want to add the question was
38:53 does a enterprise need to have appliances
38:59 at all locations for an sd-wan deployment um i you know my
39:05 response to that is generally with sd-wan architecture just not a juniper architecture that you does require you
39:11 to have an appliance at each branch location as well as a head end in the data center
39:17 as well as a single pane of glass for managing all of those policies and appliances just as a general rule of
39:23 thumb is there anything that you would add to that ray or yeah no a lot of it depends on sometimes
39:29 the type of applications because in in some cases they might need a ucpe device
39:34 because the application requires some level of low latency that can't run on
39:40 the cloud right so it could be a mix uh depending on the the application
39:47 architecture for for that key but you do have the flexibility where you could put a knit device and run the intelligence
39:54 in the cloud or vice versa or put more intelligence on the edge so a lot of it varies on the particular customer knee
40:00 there can great peter anything to add there
40:06 no i mean i think i just expanded ucpe basically you know uh there's there are
40:12 a number of customers right now especially for the bigger sites where you're looking at say one gig or 10 gig
40:18 sites where they are starting to use the ucp devices so that they can host sd-wan
40:24 as well as other virtual applications so virtual firewall other kinds of virtual
40:29 security uh analytics uh various types of edge computing um
40:36 and and so those tend to be again for the bigger sites if you're looking at like the smaller
40:41 uh sites like say 100 meg or or even lower they tend to be uh smaller you
40:48 know sd-wan appliances that are just implementing sd-wan then some of those functions like ray said can be can be
40:55 done in the cloud so those are just architectural projects you do need something on the edge though you
41:01 definitely need something there but it can be a pretty small cheap device or it can be a bigger more complicated one
41:09 great um great question here i think there's something that's on everyone's minds from a macro you know standpoint
41:17 what is the impact of the global supply chain on sd-wan today uh ray start with you
41:24 sure yeah i mean it's a it's a valid uh question because things could change depending on you know if the war expands
41:31 outside of its existing area but i think if you look at
41:36 today it becomes more and more important to look at some type of um
41:42 moving more and more of your assets into this software concept because of this
41:48 uncertainty with um you know a supply chain so that you can
41:53 have the capability to use multiple
41:59 uh cpe type device and not be stuck to to one specific uh vendor one device so
42:07 so you want some type of sd1 that's able to address some of those
42:12 particular concerns because it could impact the roll out of their cycles
42:17 for your branch offices and that keys but it's a it's something that we're monitoring closely it continues to to
42:25 change dynamically but i think from an sd-wan you want to look at some type of agnostic software
42:31 that could connect to multiple type of devices not just one yeah i'll just add on that a little bit
42:37 so if we think of you know physical appliance architecture where let's say i have a firewall
42:42 you know an mpls router uh you know a wind optimization device a session board
42:49 they're all like separate physical appliances now my supply chain issues apply to
42:54 every single one of those right whereas if i just have like a say a ucpe
42:59 device and i'm virtualizing the software i'm still potentially have supply chain issues but also they tend to be white
43:06 box solutions you tend to have multiple options and then you only have one
43:11 physical device that you have to really source and and get delivered so it can can really reduce those problems that
43:18 being said you know whenever you have hardware you know and now that the supply chain can be an issue you just
43:24 want to try to minimize it agree
43:31 another question here from um a attendee today so how do future
43:37 connection models such as 5g effect architectural decisions for today so i
43:44 guess it could be 5g or other types of
43:49 satellite you know i would say outside of your typical um
43:54 you know link status uh what uh you know what architectural decisions you know should
43:59 should enterprises be making yeah i would say um i mean and peter
44:06 could add to this when you know that's the beauty of when we look at sd-wan right is that it gives you the capability to
44:13 be agnostic of the connection type of device right so uh in in many cases what we see we're
44:20 actually seeing some deployments for a lot of retail uh shops where
44:26 they're going out there saying well the the service fighter or the csp
44:32 is taking too long of a time to provision the circuit for me for these branches and i have to be able to deploy
44:39 it quickly so being able to actually provision you know a 5g
44:45 connection have that branch up and running but not just have it up and running through 5g be able to actually
44:51 still maintain all the corporate security and compliancy and as p dimension be able to offer all these
44:57 other applications i think that flexibility kind of validates the the
45:02 use for for an sd-wan where you're not thinking as much of the connection now
45:08 the level of qos you bring into it becomes an important factor this is where you could decide
45:14 what sd-wan how much goes capability you want to implement there yeah i'll just add a little to that so
45:20 so you know for years um people have been using lte uh with sd-wan as a backup especially
45:28 and that's very effective you know way to do backup i think now with 5g what's kind of changing is a couple things one
45:35 of course there's a lot more bandwidth and qos and capabilities in 5g than there is in the lte but the other thing
45:42 is there's also a whole fixed wireless component of 5g so a lot of service
45:47 providers now are rolling out fixed wireless so so in fact instead of just being a backup it's a very valid just
45:55 you know link and so why would i want to do that well you know it could come down to
46:01 economics right in certain places but the other thing is just speed of rollout so like for instance
46:08 um you know there could be use cases where i just need to get an sd-wan connection up like really fast like in
46:16 hours or something as opposed to weeks right well fixed wireless private 5g is great to do
46:23 that once you get it set up and get a relationship with a search bar you can turn up lots of links or bring them down
46:29 so you know either for things you want to get up fast or or for you know for situations that are dynamically
46:36 changing you know rolling out sort of a temporary you know environment and you bring it up for a month after two weeks
46:43 and then take it down it can be really good for that um and then also there's
46:48 you know there's uh the no new north orbit uh satellites and a lot of
46:54 other options so like race i think that's the beauty of sd-wan is basically it's kind of independent of the underlay
47:00 but it's is still the the the uh the architects need to think about what
47:06 their options are and what's the best option and and how should i do backup and and you know and
47:13 and take advantage of those things and let you know because they exist
47:18 yeah okay so curious to see ken what the poll results yeah i'm just gonna bring that up so it's actually pretty even
47:24 across the board but uh are you surprised by these results yeah i actually am a little bit right um
47:32 because while i shouldn't say that because i i don't know the profile of all the enterprise
47:37 uh customers but that's good to see uh if i'm an msp here i would be smiling
47:43 and say i need to get very aggressive at offering you know my service not just
47:48 from a traditional managed service i need to make sure i offer a co-manage so that the enterprise feels
47:54 that they're not they don't feel like they're losing control of visibility so whatever msps
48:01 are in this particular session should get very aggressive there i guess yeah
48:07 um you know in we we touched upon this you know quite a bit but i think you know it bears you
48:13 know worth repeating is you know the ai you know peace right so we have a number of questions uh regarding artificial
48:20 intelligence you know one from stanley chan you know does this ai require some
48:26 form of machine learning period and then uh jungnu kumari can strong ai become a
48:34 reality if yes when can we see it happening um ray what's your response yeah i'll start
48:41 off is that if you look at let's say ai machine learning right i mean i'm old enough to know that
48:48 about 30 years ago i was talking about ai machine learning a lot of those algorithms the problem is that we needed
48:54 like a phd in-house for every engagement uh to get some of these things required
49:01 the difference today is ken is that a lot of the complexity of that just like where peter talked about bae
49:08 on economic modeling simulation all of that is moving to the cloud whether it's
49:13 tensorflow or amazon machine learning so that you're able to have people
49:19 right focus on what questions do i need answer right create those input
49:24 parameters send it to the cloud right and it does the complexity so if you look at a a product like with misc right
49:33 this isn't rocket science in the past this is happening today
49:38 where it runs in the cloud and a lot of the complexities so that a customer
49:43 knows today what problems and where to identify and where to troubleshoot i mean those are
49:49 things that used to take us a lot of time to solve uh back then so that phd
49:55 capability of machine learning it's kind of like in the cloud and and it's accelerating at a much higher pace than
50:03 we've ever seen before uh from an application peter i don't know if you want to add to it yeah y'all add a little bit to it so i think you
50:09 know for one thing first of all what we were talking about today with missed ai with the win assurance those are turnkey
50:16 products um and they're developed in the basis of mist and of course you probably know
50:21 miss was a juniper acquisition pretty good good acquisition
50:26 um and so so that's a turnkey thing but you know one of the things just more generally you know ray was talking about
50:32 that um you know these these neural network algorithms you know these were around actually in the 1950s i mean it's
50:40 not there's actually nothing new not a lot new in terms of the the the
50:45 actual algorithms and the approaches but the problem was it was never impossible
50:50 to really implement them because of you know you need to have enormous amount of computing horsepower so what's really
50:56 happened is really two things one is the cloud right where you can put this stuff in the cloud two is horsepower now we
51:03 have a lot of computing horsepower so i mean you know everybody's used to using alexa
51:09 at home and and hey google and and stuff like that and that couldn't happen without the cloud it couldn't happen
51:16 without that that scalable horsepower even though people knew how to do it
51:21 years and years ago it wasn't practical to implement so basically the way to think about missed ai and when insurance
51:28 is just thinking about sort of like alexa it's like you don't have to you don't have to really worry about it it's
51:34 just there and it and it works and it's in the cloud and you know just makes life simpler
51:40 and i would just add um you know trying to be as vendor agnostic as possible ai you know is becoming a bit of a buzzword
51:47 um yeah in machine learning um and we have to be careful of whitewashing the term and um
51:53 be candid in uh with your vendors that you're evaluating sd-wan and ask them to take a
52:00 peek behind the scenes of what do you really mean by ai because i think if you start to compare
52:05 you know different solutions you can see quite a bit of variance across the spectrum and what
52:10 that ai and what that machine learning really means yeah that's a good point
52:16 a shock patel asks uh if sd-wan device already has a firewall function
52:24 then does the local branch traffic have to still go to another layer of the sassy cloud before
52:30 going to um a website or any cloud-based data center so that's a really
52:36 interesting question i'll let you think about that for a second so i'm not sure we talked we talked a little bit about security
52:42 today um sassy obviously is you know a really you know hot hot topic you know
52:47 in the market um ray uh i know we've been coming to you a lot you know i've heard these videos
52:53 yeah yeah yeah they're trying to understand the question a little bit but i think if i understand it correctly is
52:59 if they have a firewall already in-house do they need some other assets in the
53:05 cloud was that related to that is that a good way yeah that's how i would interpret it like if you look at your natively at the ssr right there are some
53:12 firewall functionality based into that technology and much of what the branch looks at for you
53:18 know for security can resolve about 85 percent of that but what what happens with the rest of the traffic that may
53:24 want some i guess next generation firewalling tech you know uh
53:29 capabilities yeah i mean i i think when it gets to security i always say
53:34 you know security is as strong as your weakest link right um so
53:39 uh and each customer based on their type of traffic and data really have to evaluate how much secure just like
53:46 disaster recovery right how much do you want to implement at a certain point but
53:52 uh having it from my opinion when i worked in wall street is we want it end to end so we have to have
53:59 enable security at the endpoint device and start from the end point and follow
54:06 the packet kind of like the life the life of the packet i would call it right and and as we follow the life of
54:13 the packet what points or area is that packet touching and do i need to secure it or not so i would always start to say
54:20 yes in the dmz zone in that area you want to have a level of security or some level of of encryption and then pass it
54:28 to whether it's to your corporate location or to your cloud location and then pack it from there but i don't know
54:35 if i'm answering the question correctly because i don't know if i understood it properly can yeah no i mean
54:41 peter good yeah i could add a little bit so i think that you know basically pretty much all sd-wan
54:48 solutions have some level of firewall built into them however that being said you know there's
54:55 a lot of additional security so some solutions have are more sassy kind of
55:01 integrated into the solution some of it's more cloud-based but also what we see a lot of especially for like i said
55:08 the bigger sites um is like a ucp device is being used where
55:14 you have the sd-wan software but then also maybe additionally on top of that
55:20 like a palo alto firewall or infoblox or some other kinds of devices and you know
55:26 this is a moving target you know so i mean i i would my for those bigger sites i think having
55:34 that kind of virtual architecture is good because then you can you can kind
55:39 of mix and match stuff for smaller sites they could be cloud-based right as long
55:44 as you have basic sd-wan connectivity to the site you can have those capabilities in the cloud so you don't have to have a
55:50 real expensive you know ucp device at a you know little tiny
55:55 uh branch office or retail store or something yeah no i appreciate that peter so um
56:02 yeah we're right we're running out of time here so i do want to just quickly plug um if you do want to see can you
56:09 guys see that by the way that this slide yeah i see it okay great so if you do want to
56:14 see a live demonstration of juniper's ai driven sd-wan solution
56:20 take out your phone scan the qr code in the upper right we do these bi-weekly here in north
56:26 america as well as in emea the next one actually is tomorrow morning in in
56:32 europe and then i think may 25th if i have my date correct we do give away um a oculus vr headset um so if you'd like
56:41 to register for that and like to see the demo you know scan the qr code in the upper right in the lower left juniper
56:47 has their ai um excuse me our global summit uh coming up in uh just a few days here
56:54 may 11th great opportunity to hear from execs across you know the juniper uh landscape about uh you know the
57:00 direction and uh the technology uh that we continue to innovate within in
57:05 driving optimal uh end-user experiences you know really that is the uh true
57:11 north uh for for for juniper products across the board um ray peter um thank
57:17 you so much this is but probably about the fastest hour um that i've ever been a part of we covered a lot here
57:23 today um let me just you know start with you ray any last takeaways uh from today's presentation yeah no i would say
57:30 first of all thank you and and more importantly thanks for all the questions i'm still gonna bid for the oculus i'm not gonna give up on that but uh
57:36 i think it was some very insightful uh questions from that point of view and i
57:42 would say for those enterprises that that join is to understand the concept that you
57:48 know you know there is a cost of doing nothing but there's opportunities to be able to address you know what we call
57:55 one 2.0 or the next version of sd-wan an opportunity to implement what those are
58:00 there so so thank you excellent peter any less takeaways from you yeah i mean i just want to thank
58:07 everybody for joining and i think it was a great session and and uh you know my area is economic modeling and and this
58:13 is a area we've put a lot of technology investment into and i would just encourage people you know when you're
58:21 looking at alternatives whether it's sd-wan or any other kinds of i.t technical solutions is look at
58:27 technology and look at benefits but also always consider business case
58:34 tco revenue etc roi um because the thing is that at the end of the day that
58:41 that's going to be uh that's going to be a very important part of those decisions
58:46 fantastic well on behalf of ray peter myself ken delobo product marketing manager at juniper networks thank you so
58:53 much for joining us today uh we look forward to helping you in the future drive those optimal and premium end user
58:58 experiences for both the end users on the network and for your it operators trish back to you thank you yeah thanks
59:06 guys uh this was great uh really quickly um everyone's asking how they can get in contact with you all um did we cover
59:13 that by chance i'll throw my email address up on on the screen here um if you have any
59:19 additional questions for both you know myself peter or ray i'm happy to field those and we can do everything we can to get
59:25 that information back out to you sure be great awesome i'm just going to give everyone a second here just to go ahead
59:30 and get a quick grab of your email there and then we'll go ahead and close out thank you everyone for attending we do
59:37 appreciate you all attending today's event and hope you found it as enjoyable as we did as ken mentioned today's webinar flew by
59:44 so we're definitely engaged with ray and peter today so thank you all so much for attending and have a great rest of your
59:50 day thank you thank you everyone thank you bye