A Serious Discussion About Data Center Management in the Cloud Era

0:00 [Music] good morning folks and thanks for

0:06 joining us today uh this is the first lighting talk uh this is the first lightning talk in our experience first

0:12 data center networking series today we'll talk about modern data centers architecture uh with abstract my

0:19 name is majid ansari and i'm an architect for cloud vertical with us we have jeff doyle we all know

0:25 jeff and thank him for educating us through the years with his books trainings videos and blogs uh thanks for

0:32 joining us uh jeff um so today uh we're gonna talk about data

0:37 centers data centers are rapidly growing in recent years and uh this growth is expected to continue it means existing

0:45 network operators and architects will have to do more with less our goal today is to offer a solution

0:51 for your modern data center challenges so that you don't have to make trade-offs between agility and

0:57 reliability uh deal efficiently with operational complexity and reduce constraints

1:03 so that you can make your technology choices without without constraints we have two more upcoming sessions

1:10 one focused on day two plus operations and another one on apis and automation

1:16 um uh we would like it to be more interactive uh so feel free to put comments and

1:23 questions in in the chat uh we'll try to get through as many questions as possible towards the end

1:29 um and and just to make it easier for us to deal with the questions just maybe prefix it with a queue or something uh

1:36 so that we can we can easily find it through the chat but uh definitely comment uh as you as we go along let's

1:42 jump right in quickly on what we plan to cover today uh we will go over the needs

1:48 for a modern data center and then how evpn addresses them uh we'll talk about uh essential

1:55 characteristics of a fabric management solution like what you need to look into a solution when when you're deciding to

2:02 go for a fabric management solution um and then we'll have jeff go over

2:08 you know how to operate an evpn fabric uh using our intent based approach so

2:13 first let's talk about you know modern data centers obviously hyperscalers uh have set a goal standard for modern data

2:19 centers and looking at that looking at the success they have

2:25 everyone wants to build a reliable and efficient and scalable data center like amazon or google right

2:31 however uh we don't operate on the same ground most of the organizations building data centers face unique challenges right uh

2:39 they need to build a modern data center but at the same time they need to support legacy applications uh that may

2:44 require layer to stretch uh while their modern application is talking on ip right so you need a a reliable and

2:52 efficient infrastructure that can support both layer 2 and layer 3. now when you're supporting these layer 2

2:59 stretches right you want to be able to have any vm or any host in any rack you don't

3:05 want to be reserving uh those rack spaces for you know to be on the same subnet uh that means that you wanted to

3:12 be on any rack or anywhere you wanted to right there's also a need uh for having

3:19 a physical infrastructure shared among like you know ports or or applications

3:25 that would sit on the same hardware but at the same time they cannot talk to each other uh they need to coexist uh

3:31 but you may want to isolate it using like vrf like construct or you

3:36 uh want to have some of those flows directed through the firewall um and and that way you can have like that secure

3:42 communication going through if that's not enough uh we actually have

3:47 modern infrastructure going uh using microservices and virtual machines uh

3:53 they add a unique challenge to the network uh there is a constant change in ips and mac addresses that is going on

4:00 in your backend infrastructure that is outside of your networking element and that means that your network has to

4:06 now react to those changes and adapt to those changes without making any configuration changes

4:12 on the network and all these things need to happen on you know a simple ip forwarding

4:18 infrastructure uh that is affordable um and uh you know you want to be able to use

4:24 like you know simple ip forwarding hardware uh wherever possible like you know spine super spine or core of the

4:30 network while your edge of your network is doing intelligence uh things like you know the the the things are described

4:35 above uh this is not a complete list uh but i just try to highlight you know some common scenarios uh if you have

4:42 seen any other challenges or if you want to share something that you know you you are running into uh please feel free to

4:48 put it in the chat and then you know we'll we'll try to get to it uh towards the end

4:54 so while we look at the challenges right um i i think we are in luck because

4:59 uh evpn offers an a standard-based and interoperable solution for some of the

5:04 challenges that i highlighted uh we can just talk along about evpn uh but let's

5:10 quickly see how uh it solves the problem we discussed earlier right so first

5:15 it uses a bgp protocol for signaling bgp as we all know um has you know proven

5:22 over time uh it can catch carry large scale like internet scale or even bigger uh in some service provider network um

5:29 so it is scalable and it will meet your scalability needs uh while bgp is doing it uh we also

5:36 talked about you know being able to do layer two and layer three extensions uh evpn is able to do it uh evpn is able to

5:43 offer you both little l2 and l3 extensions using the same evp and vxlan technology

5:49 uh with bgp now because signaling is happening uh in in bgp

5:54 you don't have worry about flooding like you know you have to worry about in a spanning tree network uh you get your

5:59 traffic low providence built in um and with all that you know characteristics now you can scale better right you can

6:06 you can predict your scale as well and not worry about loops and things like that in production

6:11 uh and beauty of evpn is it comes with a lot of things built in that in past

6:17 needed their own protocol so for example right you get all active multipath now you don't have to block

6:23 ports and and then worry about losing you know ports in your steady state you can use

6:28 all your links and do multi-path and then forward traffics and all links um esi lag uh which basically offers you

6:35 multi-homing not only two two routers or two switches it can do more uh it comes built-in with evpn

6:42 and uh in past you would need you know it eliminates the need for proprietary

6:48 mechanisms like nc lag or mlag now it just happen in an interoperable way with evpn um and i just want to also

6:56 add that it eliminates the need for a chatier vrrp protocol now you can do simply uh

7:02 uni using unicast gateway and then you now don't need vrp across the network uh so evpn just had like you know some of

7:09 these some of these things is just built in uh and it also meets your requirement of

7:15 traffic isolation now you can use l3 vpn like isolation using you know same bgp route target communities

7:22 and it's very easy to you know forward uh traffic through firewalls uh using

7:27 the same mechanism so you get your secured communication and uh using a vxlan overlay which is

7:34 you know simple ip udp private i ip udp packet you can now actually use it

7:41 on on simple ip forwarding infrastructure we can where you can like spines or cores um and then your

7:46 intelligence lies on your edge switches that you know take care of all the evpn functionality uh so

7:53 overall i mean evpn i think is a pretty good your go go-to solution for for the requirements that

7:59 we talked about uh in in a modern data center again if you have any other challenges

8:05 that you know you you think that evpn can solve for you or you have or not able to solve

8:11 please feel free to add it in the chat so we talked about evpn how it elegantly

8:17 addresses your complex requirements but you still need to manage your resources and logic for those complex

8:22 requirements right you still need to ensure the network is operating uh with all those complex requirements it's

8:28 doing intended things and then uh you know your maintenances are smooth and the network if something fails you you

8:34 kind of know things uh the way they are so i i think you all will agree that it's

8:40 uh very difficult if not impossible to do it manually uh but i would also like to say that uh traditional automation um

8:48 will also not be able to cope up with it you actually and we strongly re we strongly believe

8:54 that you need a sophisticated fabric management solution to manage your uh modern evpn uh data center uh let's look

9:01 at some of the essential things that you know you need to look for right whether you're building at home on your own or you're buying off the

9:08 shelf uh there are certain things you need to look at uh in an evp and fabric management solutions so first and

9:14 foremost right it needs to be easy to use uh i i think there's be no disagreement on that um and it needs

9:20 to be multi-vendor you shouldn't be logged into a given vendor uh when you are you know building uh building a

9:26 management solution um you should be able to simply you know declare or show your intent that hey

9:33 this is what i intend to do with a vpn should go from here to here all these hosts should be

9:39 allowed to talk to that host and the tool itself needs to be able to figure out uh you know what

9:44 configuration goes there depending on the vendor uh what best practices needs to be included in

9:51 that the tool should be able to do all that and you shouldn't be burdened with uh figuring out uh you know the

9:56 configuration and things like that um and uh i think the main thing is you shouldn't

10:02 be like looking at something that just slaps cli configurations that would manually you would do it um and then

10:08 expect it to you know reduce complexity right uh it's one of those examples like you know you

10:13 probably heard when people were asked if people were asked for a faster mode of communication they'll probably ask

10:19 for faster horses but we know that you need to solve it elegantly like a car is we know as we know today is a

10:27 much more elegant solution but it require out of the box thinking so

10:32 that's where you need a tool that takes care of uh you know intricacies of what you need to do it needs to allow you to do more uh

10:40 while you're investing less in worrying about the complexities you just focus on your business needs so that's what's

10:45 like i think the first and foremost thing you need to look at um the next is right while we're operating

10:51 the network i think it's very important that uh we have all aspects or all aspects of your network life cycle

10:57 covered right uh you you need to be able to like you know decide like what you

11:03 want to do then deploy it um and and then and then see it in operational network and then provide that feedback

11:09 loop back if something goes wrong right you're on and it's it's actually doable

11:14 and it's easily and and most effectively done if your architecture and operation teams get the same pane of glass so that

11:22 way they do what they want to do and then they can actually see what is deployed and operated so a unified

11:28 solution uh you need um you need a single pane of glass to do your you know network like life cycle across uh across

11:34 the life of your network uh the next um you want to be able to know like what's going on in the network

11:41 right i mean that's the only way you will be able to figure out what's going on um

11:48 so what what you need is you you need like you know you need to collect

11:53 the right set of information analyze it and then present in a way that's useful for the end user right

12:00 and the thing is like it needs to be on time you don't want to be like you know knowing about something went wrong two

12:05 hours later uh you wanted to be immediate and uh it also needs to give you context

12:11 because the fabric is so big it will be hard for you if you just had sea of errors uh coming at you you need in

12:17 context so that you can take you know precise actions um and all these needs to be done obviously in a modern way um

12:24 so that way you can you know you can query and then get the information you need uh

12:29 the next item is uh dealing with change reliability i i think this is probably the most important aspect of you know

12:37 operating a modern network smoothly um you need to able to have all this information that's going on in the

12:43 network but in order to offer it in context and and and and provide you the root cause in a timely manner uh you

12:50 need to store it in a in a way that's easily you know easy to query uh even though they're like you know these

12:56 relationships that are very complex uh and maybe not even known ahead of time right uh so something like a graph

13:03 database right that allows you to have this complex relationship and then easily query them

13:08 to figure out what's going on um i i think it's crucial uh for for success uh

13:15 that way you get information on time you can you know look at things and then prevent uh risk

13:21 of uh widespread outages and breaches right if you need uh if you get the the problem um right in and in a precise

13:28 manner uh so that's you know that's that's one of the critical aspect but

13:34 last but not least right it needs to be automation friendly we all know that you know most of the

13:40 organization they will operate a large infrastructure on top of the network so they basically preference on uh what

13:47 programming language they use how they you know their workflows are uh and then what's what oss bss systems are there

13:54 right so you want to be able to have that flexibility so apis and simple and usable apis is a must um that way you

14:03 can easily integrate with your existing systems and back against um and then you can do what you want to do and focus on

14:10 your usual business while the two take cares takes care of it so in in summary i would say you know with a

14:16 sophisticated tool uh you actually will get uh to operate in a smooth operation

14:22 or using a sophisticated fabric management tool uh you would get a good experience and then you can grow without

14:28 growing pains right so i think keep in mind all these things that uh when you're looking for a tool

14:35 um i i would say you know put put things in comments if you think there's something else that is more important or

14:41 you know if you think some of these are not applicable to you uh but after establishing that i think i'll

14:46 hand it over to jeff uh so that he can you know walk through a solution that that we can offer that would not only

14:53 meet but exceed you know some of these expectations we we have for for the tool so jeff i'm going to pass it on to you

15:00 great thanks marcia i'm going to skip the slides that i was going to present uh because i think majid did a great job

15:08 of of of really summarizing uh both the benefits of uh

15:16 vxlan evpn and also the challenges and and the major challenge is that it's

15:22 very complex to operate um uh uh manually uh or even with some homegrown

15:31 uh uh automation applications and so what i'm showing here hopefully it's uh uh my

15:38 screen is is showing is um and uh a juniper abstra uh

15:45 blueprint of uh working um well in this case not a working network but a network

15:51 and i've got a couple of different blueprints running in a couple of different labs

15:56 that i want to talk about and because we're really talking about operations and day two

16:03 [Music] complexities i won't spend more than just a couple of minutes

16:09 talking about really how this particular

16:16 blueprint is assembled but but i do want to say a little bit about it you know because i

16:21 think it's important to say okay how did i get to this point and what you can see here is is a

16:28 fairly simple uh data center topology with two spines

16:33 and six racks uh there's five server racks and in this you can see

16:39 we say there's 24 generics basically these are 24 servers

16:45 and i created this pretty quickly and the other rack here is a border leaf rack and by the way there are uh

16:52 redundant top of rack switches in each one uh this border leaf

16:58 provides outside access the two generics here are firewalls and and there's also three access

17:05 switches which just extend extends access out what's important to note

17:12 about this is that when i built this i built it by specifying just logical devices uh an

17:20 abstraction of the switches i want um so you know i specify for my spine

17:26 switches i want a certain number of ports and a certain number of port speeds for my top of rack leaf switches

17:35 i also specify i need a certain number of ports and a certain number of port

17:41 speeds and then i create a template that specifies how all of

17:48 this ties together at no point in there at that stage do i actually make a

17:54 selection of what specific vendors i'm using

18:01 i only do that after i know that that the blueprint that i want to use is actually working

18:07 and you can see in this diagram with everything put together in a blueprint

18:13 i also specify what resources i'm going to use i specify

18:19 pools of asn numbers loopback ip addresses link ip addresses

18:25 i do that in a separate menu excuse me

18:32 that then tells appstra this is the pool of uh resources i want to use for this

18:39 specific data center and i don't have to make any choices beyond that appstore will take care of

18:46 managing that if you uh have an outside resource management

18:54 platform um like ipam obviously uh we can connect to your ipin

19:02 ipam system through apis to further manage those resources

19:07 only when i have all of that put together would i actually

19:14 decide what vendors i want to use in this particular case

19:19 for my spines you can see here that and i'm doing this specifically to show

19:25 our uh vendor independence it's only after um i specify you know what my topology

19:34 looks like to answer my specific data center needs

19:39 do i look to see what vendors will meet that in this

19:45 particular case on spine1 i have a cisco nx os switch on spine 2 i'm running

19:51 arista for my leaves i have a mixture

19:56 of cisco and arista switches running here and for these three access switches i'm

20:03 actually using juniper exes but uh but uh so those are already selected but

20:11 you can see that you know for my spine switch you know i specified

20:17 uh for my spines i want 32 40 gig links

20:22 and i get a list of the various switches that meet that requirement

20:31 which includes the cisco that i chose and the arista that i chose so

20:36 so i'm selecting my vendors according to um to my specifications there

20:43 um so that's just i there's a lot more that i could oh maybe one thing that i should

20:49 mention is um once i have my vendors selected um i can actually uh get

20:57 a cabling map and appsera will take care of actually

21:02 telling me um what ports to cable to what ports and i can

21:10 not only get that within app store i can export that um actually i intended to

21:18 let's try that again i can export that as uh csv uh which will fit into any uh excel

21:26 uh i'm sorry uh any spreadsheet uh so for example here i'm using uh google

21:33 um google chrome and um i can

21:41 it should there it is uh it took a minute for for that to process i can i

21:46 can print out an actual cabling map uh that my data

21:51 center uh uh technicians can take and and when they

21:57 rack up the specific devices it tells them what interface

22:03 to connect to what interface and this is a verified cabling map

22:09 we can also do things like affinity checks so they're so no two

22:14 connections are in a common failure domain or you know sharing a transceiver

22:20 or uh anything like that so um so all of that is verified ahead of

22:26 time but what's really important is

22:31 let me get this so there we go is the because we're emphasizing operations

22:38 here what's important for operations are three parts it's the day-to-day

22:44 uh configuration changes that you make in your in your network it's

22:50 the day-to-day information collection that you need to make about your network

22:56 and when there are anomalies in your network you need to be able to quickly resolve those anomalies and all three of

23:03 those things are our day two operations so just to show a

23:09 little bit of that uh going back to i've got my physical view here

23:17 and this topology that i created but then i want to

23:22 just as an example of of day two operations

23:27 i want to start creating some virtual objects within this network so

23:34 the first thing i might want to do is create a couple of routing zones

23:40 and one of the things maybe i should have

23:45 kept a couple of slides one of the things i do want to show you

23:50 just from a slide here is this is from an another lab that i

23:55 put together and i've got a sonic configuration or a snippet of a sonic configuration a snippet of index os

24:02 and a snippet of eos and the question that i posed up here is well do all of

24:07 these things do the same thing uh the answer is yes and um and i'm not sure

24:13 how well that shows on your screens but what you can tell in here is that these are all

24:21 configurations for a vrf named vrf blue

24:26 and but what is significant is when you look across these they are very

24:32 different configurations uh in appearance and the

24:37 the uh the thing is they do actually all do the same thing these are all abstract

24:44 generated configurations and that's the important point to make here uh as i as

24:49 i go into this piece um so let's say on routing zones

24:54 and you saw that configuration for a routing zone let's create one or two and maybe i'll

25:01 call uh i'll create a routing zone called blue

25:06 i could if i want uh specify my vlan ids to be used here my vni's

25:12 i set my routing policy and i could choose my import and out uh

25:18 export route targets uh very specific uh and very important for

25:23 uh vxlan evpn but i'm not going to do any of that all

25:28 i'm going to do is give it a name and i'm going to let abster take care of all the rest of it so i'll hit create there

25:34 um and i'm going to create one more just for the uh fun of it let's call this one

25:41 green and the same thing i'll let abstra take care of all the other choices in there

25:47 and here you can see that blue is created and green will pop up in a second

25:52 and it will also figure out the route targets uh not quite there yet

25:58 and you'll specify uh the resources you want to use so in this

26:04 case i'm going to say for my loopback ips i've got a resource that i've called data center one loopbacks

26:11 i'll use that for green i also specify my lootbacks

26:18 and and for my l3 vnis i have a pool of vni's that i've

26:26 specified for data center one and i select all of those

26:32 and uh everything should go green and uh and we're ready to go there and you can

26:38 see that actually took a little longer than i expected but uh but you can see that appstra

26:43 has uh specified my route targets my vlan ids and my vnis for those three

26:51 vrfs what we call routing zones are just vrf's so then i can specify

26:58 uh some virtual networks and i'll just do a couple here but the point here is showing how fast the

27:03 operations go so well i have a via let's call a vxlan and um just to be a

27:10 little whimsical there i'm going to call it stan and i'm going to put that in that blue routing zone that i just created that

27:17 blue vrf i'll let um [Music] again abstra take care of making all the

27:24 choices about vnis i can specify that i want the same vlan

27:29 id on all leaves i want dhcp enabled

27:35 and i want to tag that and then i can specify uh what racks i want that

27:43 vtep to appear on and i'll create that and i'm kind of doing this randomly so

27:50 um it'll be interesting to see if i managed to do this with no mistakes one more vxlan just to show and we'll

27:56 call this one ollie um again let's put that in the green routing zone that we just created

28:04 uh specify again that it's tagged and let's say that's going to appear on um

28:12 on this on the border leaf and on

28:17 these first three server racks and again i'm just quickly making choices there

28:24 and uh then i have to specify

28:29 my uh svis i've got an svi pool here so i'm just telling i'm not actually

28:34 specifying any of this i'm just telling uh abstra the pools that i've uh created ahead of

28:42 time select your svis out of that pool

28:48 and so vni's and we'll get a little spinny wheel here for a second as things

28:53 process i specify that i specify my vtep

29:00 ips and at this point everything should go

29:06 green and let's see it will quickly go there

29:14 it goes um what's happening in the background when you see that is that

29:19 uh appser is actually going through the it's creating configurations and then

29:25 it's going through and verifying all of those

29:30 configurations so if there's anything that the validators that we have built

29:35 in uh don't like they'll inform you of that and um and

29:41 usually when i do these very random things i manage to make some kind of a mistake that a validator will come back

29:47 and say hey there's a problem there in this particular case it didn't uh but

29:52 one of the things uh i do want to show you here is that now with this uh with

29:58 this configuration i can for example uh go to a specific

30:06 node in this case i'm going to go to leaf one of server rack one and what i can see

30:14 here is that this is an incremental configuration this is what

30:20 in those very few clicks that i made uh abstract created and i can see just by a

30:26 quick glance if you're familiar with arista eos you'll know that this is an eos

30:32 configuration and you can see here's my vrf blue that i created here's green

30:38 and all of the other information i need under there is created for me so my

30:43 vlans are there uh all of my bgp configuration is there

30:49 and just to show you very quickly if i for example

30:54 uh switch to borderleaf one here if i look at my incremental config

31:01 there you can see very quickly that's an nx os configuration abstra knows that uh that

31:07 that border leaf is in xos so i have that all configured all of my details

31:14 are there they're all uh validated and so i know that's going to work and i don't have to go through any of the

31:23 minutia of actually physically creating this

31:28 configuration and then labbing up the configuration and then validating it manually it's all done for me and um and

31:37 you know and once i have verified everything i would go to the committed tab hit commit and that commits all of

31:45 my changes uh if you're familiar with juniper junos you certainly are familiar with

31:51 the idea of a candidate config doing a commit and then having an active

31:56 config and so that is a the uh day two operations as

32:02 far as just the day-to-day kinds of things you would do of creating new vrf's creating uh

32:09 virtual networks and uh and so forth and i'm going to switch this is a different

32:16 actually a different blueprint i one thing i should say about this blueprint um that i'm

32:22 running actually in in um a server in my home office here is that at this

32:29 point i don't actually have any of these devices i specified uh that i want you

32:35 know certain devices to be arista certain devices to be cisco um on here and you know and some juniper

32:43 devices i don't have any of those installed yet i've created all of that

32:48 um specification in uh preparation for actually

32:54 installing the devices i have a cable map that i can now uh connect all my devices together i've got all of my uh

33:02 logical configurations done and when i bring the devices up under appstore control i then just push all of

33:09 that out to the devices this particular uh

33:15 blueprint here it's a little bit simpler blueprint with two spines and two racks

33:20 one with two copper rack leaves one with a single top of rack leaf and in this case these

33:27 are all juniper devices and there is uh junos running underneath this and

33:33 as a result you can see that there's some differences uh at the top here um

33:39 you know where on my other blueprint dashboard analytics and active actually said in a

33:45 they're green here what i wanted to show you here very quickly is uh if i go to analytics

33:54 and interestingly enough there are three anomalies it says that i have uh some systems with high cpu

34:01 utilization um that actually surprises me i'm not sure why that's going on in in my lab but uh

34:07 um but nevertheless uh it's it's a good thing to go uh to to show that that uh

34:13 appstra quickly shows you those kinds of anomalies uh but another part of what

34:21 i would want to do as part of my day-to-day operations is

34:26 gather information about my network so for example if i go

34:31 to device traffic i can very quickly uh display and there's

34:37 not really much going on in this network so so pretty much everything shows zero packets per second

34:44 but very quickly i can i can see what's going on in my network in terms

34:50 of of traffic bandwidth utilization

34:56 another thing that i could do and by the way you do all this by creating probes

35:01 maybe i need to see some information about critical services and so i can very quickly select a probe

35:09 on critical services and then this is something i did ahead of time on

35:14 all of my servers i specified uh what kinds of services are running on

35:23 those services so in this particular case i might want to say that i want to know what my critical services are doing

35:29 for my app development and maybe for i have a test

35:36 network and maybe for finance

35:41 on those particular applications i want to create a probe and it takes a minute but but

35:49 what i have now again not really anything going on in the network is very quickly i can look

35:56 and and see for those critical services that are running in my network what my uh

36:03 bandwidth utilization is uh uh you know where they're running what interfaces they're running on and

36:10 so forth and there are a lot of uh variables that i can set to query

36:18 uh what's going on in that network there's also a lot of variables about how to

36:24 uh query uh the network the last thing i want to show because i know we're quickly running out of time

36:30 here is uh is uh within analytics and also on the

36:36 dashboard um i could very quickly see what's going on

36:41 in my network so when there are anomalies and interestingly enough let's just click on that anomaly uh because

36:47 i'm not sure what what's going on there oh you can see there was uh in this particular case for device system health

36:55 it's saying that uh that i have some cpu utilization that's running beyond

37:02 uh and it's on these specific devices actually that's all my devices

37:07 i'm running at 27 and i had a threshold saying let me know

37:12 um when my cpu utilization gets beyond a certain point

37:18 but with a longer demonstration what i normally like to do here

37:25 is sort of go in the background and create a problem on the network and then show

37:30 how quickly i can go through uh and troubleshoot that problem and literally resolve the problem in

37:38 in just a minute or two um but uh given our time limitations

37:44 here uh i'm gonna stop at this point um maybe one thing else to show because

37:50 this is something that uh maj had uh pointed out is uh

37:56 our theme here is managing complexity and if i actually again you can see this

38:02 very simple network that we have here uh in the background abstra is tracking this

38:08 network uh in a graph model and i could actually look at that graph model and

38:14 again you know just showing that this is a extremely simple

38:21 uh network and yet uh here's the the graph model

38:27 for uh for actually tracking that network and you can see that this is something that's that's tremendously

38:33 complex here and so this is something that's all being worked in the background and would be almost impossible uh to

38:41 completely manage if you're trying to do all this manually uh so i did want to point that

38:47 out because i feel like you know there's a lot that i could say about this graph model

38:52 but you know there's there's a lot i could say here in terms of uh managing complexity but i wanted to at least

38:59 show that to you and at that point uh much it i'm going to throw it back to you

39:05 and maybe at this point we want to open up for questions

39:12 yep no sure jeff i think uh we we can certainly open up for a question real quick i think in summary i want to say

39:18 like uh you know for our quick discussion today uh we we kind of highlighted that you know evpn vxlan uh

39:26 is is a technology that offers you an efficient and scalable l2 and l3 connectivity uh

39:33 but in order to manage your fabric right you need a sophisticated solution um and

39:38 then just not any solution but you would need a unified solution uh for seamless

39:45 automation from day zero to date day two so that you know your architects and

39:50 operations guys can see through the same pane of class right and based on you know what jeff showed you

39:56 and then the things that we talked about uh i think you uh we'll all agree that you know juniper abstra uh actually does

40:04 uh meet all the requirements of a of an effective fabric management solution and then does it in a way uh that you

40:12 know it helps you do uh more with less and then grow without growing pains

40:18 and then in addition to that if you want to know more or or learn more you actually actually have free resources on

40:24 juniper website uh we have a training portal uh we also have virtual labs um

40:29 and then there's a fantastic youtube list where you can actually see uh videos like this on you know how how

40:36 abstract can help you with your fabric management uh so i'll i'll conclude

40:41 with that and last thing is that we're actually going to have more lightning sessions like this about how abstract

40:47 can help your day 2 operations or how it can help integrate with your existing fabric

40:52 uh so please uh you know tune in to those those as well uh so jeff before we close any any closing comments from you

41:00 no i uh uh other than just saying you know if you'd if you'd like to see a more extensive demonstration of abstract

41:07 i'm always available to do that definitely thank you thank you jeff and

41:13 thanks for joining uh folks we'll look forward to talk to you more

41:19 [Music] you