Cisco Data Center Network Manager Overview

my name is Lukas karate gir principal engineer working out of the Cisco data center network business unit so us gerra already said we're owning pretty much everything which is labeled with a nexus or with an ACI or with the DCM or everything around that area the the story simply around data center network manager we went through numbers iteration of data center network manager any one of you guys saw it in the past it was called San fabric manager or something like that that was grandma's one so we're now at version 11 11 is approximately 18 months old so it's not brand new it's it's exciting new it's actually nicely majority over time and we want to tell you a little bit about it and we want to tell you what we actually put into this new thing in DC and m11 and as you can see here it's not anymore just one DC NM it sounds strange no we didn't create a multiple product product but it has different modes you can use so we have different install modes which you can select actually the same software type in their respective operational mode you want to see now classic LAN is what we call TCM land in the past it's basically you click something and it executes something on the switch it's a very boring so it's a CLI abstraction if you want so we have Sam that's where the original part comes from and then what we did very recently is what we call LAN fabric and that's what we're going to talk about more extensively in the next couple of minutes or so and I hope my demo works demo gods are good than me now paired with the land fabric we also came up with what we call a compute node you can guess that today you have multiple applications running in your data center network manager so in a management operation side and some of you want to host them that can tell you you go and install hundreds of servers and new your open source deployment and so on so forth or I can give you kind of a turnkey solution from from an infrastructure perspective from an appliance perspective and run that stuff on top of that so DCM extends naturally into compute nodes and you install it as it would be a DCM just in a different mode last but not least on the bottom media controller we're very extensively working with the production networks multicast network specifically professional media and that's where the media controller comes in it's basically a smart multicast controller which understands how much bandwidth is available in your network where you want to have your streams actually set up and make sure that there are no oversubscription and specifically no drops in that multicast network so that's the media controller but all of this is DC NM so it's the same bit it's the same install you just select what you actually want to do with it and then you go forward and run your network managed out of that - the question for you on that because actually work a lot in the multicast space in the IPTV space and ru is this only like for a streaming media and multicast applications with like over the top inside the data center are you also seeing this for like headends broadcast video type a didn't work yeah so it's for production networks so we actually run a couple of very famous TV studios on top of that okay so it's the production of work stream which are transforming from the classic in kind of approach into an IP multicast approach okay and that's the controller's side on this Roger yeah yeah it's just a little bit there but sorry that I don't talk more about that today no that's good actually but I'm driving into the LAN fabric now just yours of course please please speak I said land fabric is very we want to focus on and when we look at DC nm it's a classic inventory provisioning and resource visualization which we which we have up there and then with the LAN fabric mode we actually went into the full lifecycle in full network deployment life cycle or management life cycle and that's where we started with a zero day one day two you guys are aware of these terminologies right they zero I want to build my network they want and want to continue expand and then they - I want to do today the operational part they - ops part which then Tim will extend we talked about that so this is all this same DCM and you see here actually the compute nodes coming out sliding out and the application framework over the top which is providing us the ability to run on that single operational infrastructure tooling different functions and different tools which are required to basically complete that lifecycle ending so you'll see a lot of this not because it will be under the hood but this is actually how it looks like DCM the compute cluster you will probably hear me talking about the application framework which is one of our micro-services infrastructure we built and you can actually run it on top of an appliance which we are going to provide what's the difference between a CI and the DCM I know you know a CI at least from what I could see so far has a bit more capabilities but why create two different products oh that's fantastic question so when we look at a CI a CI is changing the network consumption model so it's that whole good I intent based networking approach that you don't really care what's running underneath now there there's still customers which are very in the native or the legacy kind of operational mode or didn't want to make that jump over to two Sdn solution as a whole and that's where DCM comes into into the pictures so DCM runs and XO s switches so you're still your individual switch it's your DCM over the top which takes care of the configuration and the management of your network you still have access to CLI if you want to it's still individual nodes and it's still the classic feature set which runs on the switches down there it's DC am actually doing overlay as well or is it just fabric underlay sort of manage that's fantastic question as well so DCM does overlays DCM introduces overlays not on the server side but on the network side so we're sticking to our switch capabilities so like if our next two switches absolutely we we run V taps and let me see if the next slide makes actually sense because that's where we are here at that Third Point so vir we're building networks in DC nm and we're trying to abstract these networks to a certain extent so these this new install mode specifically Al and fabric mode helps me to have workflow driven network creation I I don't want to say in ten dollar buys I need to do take two shots but it's you basically saying what you want to have instead of configuring the OSPF process when we go back to things like land classic which I mentioned before you would say I want oh s PF process 200 on this interface a very low level of intent when we're going here we are a little bit higher level we say I want to run a fabric and that fabric should run an overlay or that fabric should run a complete routed approach like the ebgp RFC 7490 RFC sorry the b2b be to be routing mstc fabrics however you want to call them right or you can even go there and use it same as land classic but in a closed-loop automation meaning what you sent is enforced and if somebody goes into the switch and changes the configuration compliance comes back and tells me about that so it's a kind of a half intent if you want to call it from that full lifecycle there but yeah that's that's exactly what we're doing we are modeling configurations we are deploying these configurations and we're ensuring that these configurations are really there in place so you can go destructive absolutely but we will tell you about that and we were able to give you a reconcile going forward and I said probably everything about this here already on the slide so we go a little bit forward one step forward in land fabric mode or in the land fabric install mode there is a class set of tools which we are providing and beer you're putting that that umbrella under under the name of easy fabric now you guys know what I meant with how many times I'm going to say fabric right it's the land fabric the easy fabric and how our now this whole easy fabric construct should help you you do the plumbing at first I mean you have to go and plug it in right the switch needs power needs somehow a connectivity referral about a fan but in band works equally and then we basically can bring up the network by bootstrapping it based on complete empty switch and the beautiful thing here is I don't need to know a serial number I don't need to know an IP address if the switch is in that power not a provisioning mode he will recognize that there is a DC nm in the network and he will register against DC nm and you will see the list of switches a Weibull from there you can select and go on at that point this is not the only install mode and I think as you see here there is something we call Greenfield discovered so if you don't like you have a DHCP server in your out-of-band network or your any in your invent network and you decide to put the IP address first on the switch and then you want to discover it so we can do a full Greenfield discovery of these switches and that's what I'm going to try to show you guys later in the demo last but not least which is I think it's the most complicated thing to do these days is brownfield so having an existing network and trying to assimilate all of that configuration into the CNN right so let's take for example we did this with the x90 VPN you configure it by hand you guys know how many lines of CLI you probably could have and we're going to I'm trying to understand this we were taking IP addresses for point-to-point routing and populated into the resource management we have in DC nm we're going to take your OSPF your your i/o size configuration we're going to look at your EVP and configuration you have in there your via nice right the the networks and verbs you deployed and we show you that subsequent in DCM once we imported that and that was one of the first iteration we did with DC nm 11 that was actually when we took the nexus fabric manager functions and migrated them into DCM 11 and allowed the migration to be completely zero-touch by not taking one tool database into another tool database and try to model stuff around it's by pattern recognition and by template and the DCM configuration function is all based on policy construct it's we call them micro policy so it's some little snippets if you want to little config lat snippets we ship them with the product so you don't have to create them or you can create them on your own if you think some of these configuration don't make sense to you you want to have it differently but overall we build a whole list of these templates together concatenate them take the resource management in and create a configuration at the end which will be deployed on the switch and now when we go back to the configuration compliance we're not matching one config file against a full order config file it's a subset an intent template a policy template if you want so against what is actually in there so we understand what an interface means we understand what the routing protocol means we understand what VLANs mean and try to not only abstract them but also move them into DCM and make them available for you make sense cool boring you guys want to see you know guys don't believe it I'm sure that you don't believe it so I don't believe it I didn't believe it when I saw it first but actually works still data centers might have multiple tools in there and when I want to integrate domain controller let's call this e and I'm a domain controller for a set of infrastructure in my data center maybe I want to hang that into another domain controller maybe I have each other maybe but I'll try to normalize my deployments in a way I want to have the same IP subnet or network whatever communicating together in a certain way that's where the northbound API to the discussion so DCM is actually all what you will see is running on top of the REST API so even the UI of TCM itself is running against the REST API so everything what I click you can actually do on top of that you can build your own workflows and run them against the API sand you're pretty much happy - on what the outcome is because an abstraction what this networking fabric or the switching fabric does against what you want to do when you have built in your logic so I'm gonna question for you I'm just curious Cisco's had a few different products over the years that have dealt with data center automation orchestration and some of these things what differentiates DN CM from things that you've done in the past I don't think it differentiates it in from the past it's actually a thing from the past which got just modern over time okay so as DC nm I think I started working on DC m7 that was probably eight years ago okay yeah it's well decent m6 is even older and before it was San fabric manager which was the core of it and then we merged it into datacenter Network manager so it's actually around for quite a long time and the toolset which is in there in the in the core of the platform has being revamped over the time some what are nice micro services as as one example which we did in the very latest version but we we kept all of this knowledge of how to interact with certain systems from an MDS fiber channel switch all the way to the most modern Nexus which we have in our portfolio okay so is this something that would be considered complementary to like a CI absolutely yes so I said we we have a lot of customers still running an XO s running a 5k or 7k or even a 9k in annex OS mode this is absolutely complementary to this when you think of data center course even in a CI you might have a couple of boxes which are running on X OS complementary to it right you can manage your network there or you can just run in a complete standalone fabric mode or nx-os fabric mode would be excellent a VPN bgp whatever you want to not everyone is is there and wants an AC i turnkey solution so there there is actually a turnkey solution for Denix always on top of it okay yeah I feel like their capabilities are really similar what would be a good use case for why you would implement both I don't want to look at them as competing in a way on one versus the other it's really what is your operational model on that side what what way you want to go as a customer and look at it there are a lot of vendors out there in the market saying hey that's my way or highway and we we have a certain set of customers which are with us for a very long time I want to make sure that these customers are happy and make sure that other customers which have an order view of building networks like the massive scalable data center customers that they have also an operational model they like for that we have or remotes on our searching system but it is really what you like and how you like to operate your network earlier on in one of the slides you talked about how this product is supporting a hybrid and multi cloud yes is that something your guys are building like the overlay into various cloud providers or something and if that's the case and has that done is there are you guys doing like virtual router type function that you deploy inside a cloud yeah yes so Jerry was talking about that yes so real hybrid cloud integration so we have a set point in the clouds we have Google Asher as well as AWS the cloud service router 1000 v is pretty well-known and DC nm actually managed stats entity out there and allows you to extend the VR FS you want to you want to have up there and Schurz the security the IPSec part there and make sure that it comes actually directly from your data center on Prem fabric all the way into the cloud and this also may be to complement what you said before happens on the ACI side as well as on the standalone nettings outside so we're trying to make automation pervasive that there is no gap between one or the other so the operational model you want to run your network with besides when you go versus capabilities and trying to to do the matrix and say ok I get not there is on the left and that not there is on the right I like it on the right so I'll go to this model over there we want to make the functionality more pervasive there where is this headed with DNA Center and feeding into that is it no it is it is so system is a very strong view on multi domain multi domain is campus status and service provider as such and the domain controllers that's what we call them domains are there with very specialized know-how and we want to keep them there with very specialized know-how but there's interaction in between them and that's where the api's come in and that interaction on the top with the DNA Center for example there is pair of eyes integration which is DNA Center to order domain controller or there is more abstracted model where you have a single pane sorry of glass which you which you can go to and operate that we do that actually for quite some time with Anna so network service orchestration which was an acquisition we did at a left some time ago so there's a there's a lot stuffs that stuff going on okay all right you guys want to see how it works let's see let's see if that goes so I'll have to split the demo in two parts and the reason is very simple I don't want to let you guys wait too long so for the first part I'm going to discover a little kind of network here look very fancy a couple of switches I will create a fabric and we'll show you what fabric options I have I will go in import these switches I will discover them you remember I mentioned that discover green fields and the switches are empty they just have an IP address set after discovery I will set some roles and then I will make sure that the configuration respectively is prepared staged and then pushed and that will take a couple of minutes so I will switch over while this is going on and show you what we do with an existing fabric so this fabric is built looks pretty much the same as the one before it actually has one switch more here which you will see then it's a slightly different type of stop switch but it's the same always running on top it's a couple of Nexus switches abano for cloud scale portfolio here and what I will do on top of the existing created fabric I will create VPC so you guys remember VPC virtual port channel multihoming or something like that how many lines of configuration you guys did in order to do that in a couple and how many times you missed all the best practices yeah more than compatible okay so we have a presidency what there are the best practices Hey here we go I will run a VX land on top of it so I'll create a Verve and I will create a network network is for us a layer to V NIR layer to segment with a first top gateway so we use that distributing gateway and such and then I will attach actually the VMS are already there I will just create a poor channel here a V PC make sure that the connectivity happens there and that these two VMs can ping to each other you guys probably need some stronger glasses because my my screens are very small for the pingers just got them then and that's basically than the end for them only can do more question and answer I'll try to show you a little bit of how the policies looks like where you see the policies those options you have in fact you can build this in probably less than 30 minutes without a single change or you can go crazy and change every single knob which is available on the CLI and wrong good yes alright let's see if that goes let me switch over to data center network manager 11.2 doesn't switch let me see [Music] alright so we like our britches around here yeah let me see if maybe I just sit down that's probably use a little bit more real estate you guys can see it right it's big enough okay so you'll ignore what data center network manager splash screen I increased a little bit the size so that's bigger than what it normally is but it should work well I'll have a fabric already configured there as you mentioned I said I have a demo fabric being built I leave them just on the side so ignore that it even exists and I'm going to create a complete new fabric together with you guys fabric builders our tool we have our menus slightly structured between control monitor and administration topology will show you after it's a topology then but the fabric builder is where a start control my my entities in Vera try to build my new fabric and that's what I'm trying to do here I need to give a fabric name not not very magic I guess everyone likes a name to don't name something and here you see for the first time different fabric templates so this is actually the magic behind easy fabric which creates all this modeling of configuration put the policies together and make the switches at the end working as they should it it is a template you can edit there's a full template library and you get access to the template library whenever you install these CNM it's under control template library you'll find it in this specific case I'm going to create a zz fabric here and it asks me to enter a couple of attributes we were following the mantra of the minimal amount of input to get the maximum amount of output it's great for customer who just want to try something very fast so I'm I want to hear an a s and we'll try to do private a s I want to use point to point so I'll allocate out of an IP subnet out of a scope point to point here's a slash 30 I could change that to a slash 31 I could do OSPF I could do is is I could do to route reflectors I can do for out reflectors so this is already in in mind with with a V excellent evpn fabric when we when we see this here I can go on with more modes I can say I want multicast replication I want ingress replication here so all of this multicast complexity goes away from me I don't I don't care on it and you see here I literally have pretty much every knob you could think of how you like it to turn but I don't have to right you saw me I entered an a s number but the very first one I change to ingress replication just because I don't want to enter a lot of order things or think about multicast and I can say safe and add switches and I'm done okay I'll do one slight change simply to speed up things and we have here a way to not write a raise two switches there are already empty but I'll go and import them as they stand but did this is my empty fabric container I created so let me go just very shortly back you saw before we had demo 11 only that was my built fabric now I have NFP which is a completely new built fabric if I don't like any of the settings I did I can go and change them here so it looks you're using it to create like a green field yes right can you use this to import a brown field that's already established absolutely okay absolutely so let me actually go exactly to that point once we create it the construct of what we call a fabric or container whatever you want to call it it's a it's an empty bubble right so it's something very store something in it from there on I can add switches and it can add switches in multiple ways I can just go into power and order provisioning and look what switches were discovered so who'd talked to DC animator history against DC nm so I get the whole list of them there's none of them there I can actually create empty switch just a placeholder with a serial number and whenever that switch comes on line over power and automation I will take it into the fabric and we'll deploy it with the respective personality configuration I also have the possibility to discover existing switches either by seed IP or by list and I'm going to enter here a seed IP and a username and password I'll stick to two hops I will not preserve the configuration which brings it back to the question from you I can actually preserve the configuration that means that I will import everything which is on the switch compare what I import against what the fabric actually supposed to be and I tell you what matches and what doesn't match and you can reconcile when I reconcile it can be disruptive that depends on what your diff is for example I have a IP one and IP two on that link but I want four and five you have to change IP s that's a disruption so that's not really the tool but the tool but recognized so let me go and discover some switches here I will not preserve the configuration yeah there's a little red thing which tells me it go in search for and you see here it discovered all of these different switches I have two border Leafs I have two spines here and I have three Leafs and I import that yes I want to erase my configurations and don't want to preserve my configuration in that case and how are we going through the discovery process but this is not magic yet right it's a little bit of log in with SSH little logging against NX ap a little bit of SNMP just to get these objects and to understand what these switches are about capabilities they have I on purpose have here a different switch in there when we would go a little bit further in this whole configuration of fabric you would actually see that that switch has a certain function missing in hardware so you would not be able to in do even not be able to enable that function you have a Broadcom chip is that what it is this is one of our first generation Nexus 9000 switch is with the tandem chips the Cisco and Broadcom chip and yes exactly there maybe you can do the layer 3 V tap on we actually can do the layer 3 V tap there but there's some multicast a restriction stare we we have a mode or an overlay multicast which is called tenant router multicast which is like M VPN these switches don't have the ability given the time them capable the tandem approach which is there ok so you you see here this the switches are slowly being discovered and and you see very very slowly here there's there is a little red tab which has phase migration mode if if you would preserve the configuration that migration mode is actually an indicator that I'm going to parse all of these convicts and import them into my DCM as we go here and I promised you to show a little bit more of building this fabric I have here a switch and this is a simple top of rack single rack unit switch which I normally would use as a as a leaf or an as an axis and I want to actually give him a different role I want to use him as a spine and by selecting the role he will subsequent get a complete different configuration than a leaf or a border node in that in that case so these this tagging of these switches these capabilities they have and with that additional no notion of their it belongs in in the fabric topology will be used to generate the convicts because there's an assumption here if you're gonna take a switch and put it into a completely different role that you you have to have an auto out-of-band management network setup that this is interacting with so that you can interact with the control plane without you know without losing access to this which yes so I'm using an out-of-band management here I'm using management Bureau and a completely separate out of magic now dependent Network I don't have to when we do in band bootstrapping we we are connected at least to one switch and then go in bands with respective DHCP relays and relaying off of the Pope information we required through so we can actually do inland as well and so you would not lose that this is specifically interesting when you go in an overlay underlay network the excellent network for example you use your underlay as an embed network for that and then we'll go there here I'll give them here a border role you see I have various different roles various different functions DCAM Maps pretty much all the functions we have in nx-os to build such fabric constructs we bring into DCM and it comes piece by piece whenever you upgrade the software on the switch the software on DCM will recognize some of these pieces and then enable them whenever necessary so you see me here I have a hierarchical Network here leaf spine and a border I have neighbors our links discovered these links are plucked in I can also create my own links between switches if they would not be plugged in so I can build that I can model that Network if you want to and at that point I would say go and generate my configurations you haven't seen me entering any like router OSPF or IP address right I don't have to enter an IP address I entered an IP pool and it arrives out of that IP pool the necessary subnets wherever I need so I'm creating loop f0 loopback one I'm creating point two points I'm creating a rendezvous point if I need to I'm creating a route reflector if I need to but I haven't entered it at all right the system does it for me in a best practice kind of manner that's what we deliver out of the product with with fabric builder or easy fabric what's the underlay protocol in this case I think I selected OSPF okay you have a choice I have a choice easy fabric supports OSPF and eius eius and we're using the overlay control plane namely evpn with ibgp that's where the tradition comes from of VPN rest families and and service provider networks the we also have the capability to build ebgp networks and then we can manually add a VPN if you really want to and butcher it into something it was never intended to do right Scott you remember that discussion but you can't do the you can't do the edge ERP for the underlay I could do ERP I'm not providing you with any template for EHR B but you can create an ERP template if you want to so I'm going to push this here yes this will make a couple of seconds and then you see the horse race of of these individual switches which are being published and made ready but I don't let you guys wait until it's done I'm going to switch over to another fabric before I do this I want to just give you a short overview of how the configuration actually is being built so you see this list of templates these are all little pieces which then as a result to build your show run all of these pieces can be viewed or edited let me say for example I took here policy which was set set with border and can go and change that if I want how does such a template look like you see here my template library this template library will be with the product will be update as the product goes I'll can open up these templates and you see it is a set of form language which defines the input and a set of CLI so guess how difficult it is to create that stuff on or modulate that stuff if you really want you have you have basically an open set of of a two library to build fabrics and to build networks if you want to so that was day 0 good you guys like that seen it hundred times doesn't work anyway when I want to use it actually works and I want to show you that it works so this is my second fabric and you you see it looks pretty much similar as before I I sent convicts down I can actually go and look at the convicts sorry what did I do it right now did I click something bad just a curiosity question while it's thinking when you were looking at the templates and one of the one and I don't know that one you picked or one of the other screens but it mentioned that it was just good for the the Nexus 9 K yes was that something that you had selected previously because the commands I saw in there was kind of like those would work on other devices too yes so when you create a template you have actually the Liberty to select for what platforms this template is good so you can literally create a template with the same name which is but assuming these were built in is there a reason that you would create a template that could be used on all of those but just fades for the 9 k sure yes so we do certain fabric modes or certain ways of installing which we are narrowed down to one set of platforms but you're free to change it's it's mostly how we built the product from a quality assurance and system testing perspective we want to deliver you best practice templates and this the best planning module says excuse me it just made it really modular for me I made it very moderately ok let me see let me close this here let me go back into fabric builder and I think I told a switch to go into maintenance mode which I actually didn't want to do but that's fine so let me go here let me put him back into active operation I mean that's something DCM also does so software upgrade is one which you can run through that but maybe you want to take a switch first out of operation so you can do equally put it into maintenance mode graceful insertion and removal and make sure that no traffic is being impacted there but out coming back to what what I actually wanted to do here for the demo I wanted to give you a V PC domain can anyone give me the CLI commands for V PC anyone has it by hand we've tried to stop with the madness of CL eyes and we just say I want to do a V PC domain I want to do a V PC pairing and the even specifically in VX Leni VPN removed the peer link so you're you're don't need a physical peer link anymore use you see me here I'm selecting a year leave 83 I said I want to do V PC pairing I want to use a virtual peer link and actually you see now there are three viable switches which could be a V PC neighbor for that domain I used 83 so I will use 82 but equally it could be 81 or 82 because there is no physical link anymore in between this system okay so I actually created right now with that single click and with this safe and deploy a V PC domain and made it and staged it way I'm gonna miss something but why are we doing V PC if you're doing evpn is that necessary yeah when you want to do all attached hosts mm-hmm we are still using V PC under the hood under the hoods the reason why we use V PC in again still is right the multihoming approach is linear with V PC then using EBP and multihoming okay very you're not doing EVP and multihoming you're doing EVP multihoming but we are preferring to VPC we have experience that it performs better in the markets okay even as we write a lot of the EVP and multihoming code itself okay working actively as part of the IETF there we are still very strong on the VPC side and have a we have a little bit of a head start so what if you are in a mixed environment where you wanted to do evpn with other vendors are you saying you can support that EVP and multihoming approach or you can it's just not the preferable way yes so we support EVP and multihoming it's actually to be very very frank and open it's very very hardware dependent it has their in specifically in IP overlays it requires a function which is called Mac multipathing which is not pervasively there in every hardware so it depends on the hardware you select but we have platforms doing this so we're doing EVP and multihoming these cases we do EVP an interoperability actually for the last six years at ent see we're there with the two other original authors since basically day one and yes we're doing interoperability testing and it works actually pretty well we can share the white paper as part of it it's it's not too hard to find but I'll go and definitely make sure that you guys get it so I configured a V PC domain right you saw me that I clicked and I actually deployed and it's now there so how many lines of configuration it is again I don't care it was around 50 or something like that I'm done right my V PC domain is there I want to create a port channel down to a host I'll go to my interfaces still it's an interface I want to create I want to create a Lee PC and you see here under interfaces there is a V PC based on the selection that I want to do a V PC it tells me there are actually V PC domains accessing I select that I can change my ID I don't want to I just leave it as it is and because I prepare it a little bit here I know that my server is connected on Ethernet 1/2 on both of the members of that may be Seto main I'll save it I deploy it and that pretty much is my interface being done right so I did my fabric at him a VPC domain I did my V PC interface or port gel and the only thing which is missing is now a service which is the overlay right I don't have the X line right now running my pings are let me see if my pings are working well you guess it works doesn't work doesn't work so we have to go and figure out how we create that that overlay now in this and for for that specific we have here what we internally called a top-down deployer it's creating a network and again minimal input I want to use VLAN 10 because my server is already in VLAN 10 I want to use one and two one six eight ten at 1/24 for my host facing subnet and and I done know I need a verb you guys like me arrives worse yes mm-hmm we are f it's a verb okay configuring a verb lot of lines not here I just created a verb so what do you see and I stick to the word what do you see here again you have templates these templates are in the template library you can go and edit them this is nothing else than template of CLI commands and an interface or an interview of how you enter data here I create the network and I'm ready to deploy this now and at that stage I can select my switches I can select my interfaces and then at a given second later or so I should have my communication layer to working between these two VMs so I'm going here leave 83 actually going to select only one why do I get all this - it's a V PC domain right there paired together so I will not be allowed to individually configure them I want to go to my portal one eye on purpose I only save or select one interface if the other interface is the same let me show you by accident it fills it in if it's a different one I can edit it and give another pour channel there if I think there should be additional configuration deployed I always have something called free form free form is let me add a CLI command which I would like to have there let me similarly do it here this is my order leaf where my order server connected it's even at one one it's a single attached one I'll do safe and now I have staged three switches which are getting a layer two segment or is it let's write it down here using a VLAN 10 to the host it uses vni 30,000 between the switches when I want routing I route in my vrf with a VN i of 50,000 and the wharf is called my v RF and I'll go just here and deploy that and this is pretty much how simple you could build the excellent EVP and network from scratch deploy a couple of networks on top of it and let host communication happen guys believe me that it works barely I would wait till it's green okay it's green let me see I actually don't know if it works I haven't tested yet but here what's going on at a given point these interface will come up actually they will not come up because I did a mistake have you guys seen what I did on the interface on a V PC when I created it I said LACP I don't do LACP down here so I probably should change a computer it's like pedia I should have known to put mode on so it's let me see can prove you that it actually is the mistake but I don't want to steal too much time from my buddy Tim we're still good to him see that's my interface there we go here it is yet it huh see I even know my mistakes mode on safe deploy you're not new active to both sides my server doesn't do active and I do suspend individual to avoid network loops when some of them happen I think somebody teaches you that is best but not sure if you did any classes which I attended at a point anyway but here we go we have a ping and you should be able to see that ten twenty nine and ten twenty seven they seem to be in the same subnet and so that bridge over that router network is actually working here so VX Lonnie VPN is in shape and running and then we now go to the topology up here as a final view for you you can see DC nm supports multiple fabrics you're not tied down to one single instance of a fabric per DCAM instance you can actually run multiple of them you see me here having that fabric demo eleven which I did previously you see up here the monster I created which doesn't look very nice here and I can go and zoom in into their respective one and all the way down into the switch information if I want to which which brings it back into inventory life cycle of the network itself backup restore row box and so on and so forth this fabric aver you have the template them access if you want to and that's what we call easy fabric all right with this any questions are you seeing slower load times at everything when you add more fabric instances it's gonna say say I have 50 fabric and surface is that gonna make my load time lag no you should be perfectly fine on that one DCM has a limit of somewhere 250 350 switches so that's your operational domain which he can manage soar DCM can manage you can install multiple instances of DCM it's OVA or it can be a physical server if you want to and then if you want to create a multi DCM manage we actually can import these cinnamons DCMS and show you a single view of of these different things so there's always a CPU involved the CPU is only involved for the management part this piece over here the network itself you're you're on a different scale basically the scale of the hardware for the switch itself

Channel: Cisco

Views: 9,913

Rating: 4.9069767 out of 5

Keywords: Data Center, Networking Field Day, NFD, DCNM

Id: DGs9qOkVb5c

Length: 47min 24sec (2844 seconds)

Published: Thu Oct 03 2019