Pass Cisco DCACI 300-620 Exam in First Attempt Easily

ACI Fabric Infrastructure

1. Course Introduction

Hello everyone. Welcome to Implementing Cisco ACI. You can see here the paper code is 360 to zero. And we are going to learn various things related to ACI. It is again divided into six categories. So we have to go and learn about ACI fabric infrastructure, package forwarding, external Network Connectivity, integrations, Management, and ACI anywhere means you can take that ACI and it can be deployed inside the private or the public data centers. So let's start with one, the ACI fabric infrastructure. Now, what are the topics we have inside that? You can see that overall weightage for 10 is 20%. And if you complete this particular section, you will be understanding about obviously what is ACI, why we are using ACI as an object model, what is the basic component or the core component of ACI? The hardware, different type of policies, a little bit about their routing and LC mechanism. Okay, so before going deep inside ACI, what I'm going to do that upcoming two videos will just understand. We'll just focus on the concept that, first of all, what is ACI, what is this application centric infrastructure and why we are using ACI? So let's understand those things first in upcoming two videos and then we'll go deep inside all the topics that we are seeing one by one. Yeah.

2. What is ACI Ver 01

Section one: we have to understand ACI topology and hardware. But before understanding those things, let us first understand what ACA is and why we are using this modern data centre approach. In our data center, what are the benefits we have? So for that, I have recorded two videos of the same type, and let's go through one video. After that you can watch the other as well. Then you can conclude what types of advantages we have with using the SDN-based modern data center-based or automation-based policy-based data centre solution or data centre automation solution. Correct. So first of all, this ACI is very much derived from the ACI is derived from the SDN approach. SDN is a common framework that simply represents the decoupling of data and control plans. That's the definition of software defined networking. Correct? Now when you go and learn STM, you'll find that inside software defined networking you have data plane, you have control plane and you have management plan. Now in our case, the management plane is nothing but the Epic application policy infrastructure controller or Cisco Epic. The data and control plane are nothing but the next switch that we are going to learn understanding the upcoming section. Now remember one thing here important that the ACIis not 100% Sdn, it's something like customized STN at all you'll get look and feel of what is the standard Sdn approaches, software defined network approaches. But Cisco has done their own customization, and they are very much influenced by the SDN approach. So what are the key benefits we have here? You can see the key benefits is that we have the centralized automation and policy driven profile. This is the key policy driven. Now it is all this company they are looking driven approach. Either it says data center. So the SDNA approach that we have inside the land that is coming under SDA access or DNA, then over the van that is nothing but the SDWAN or Cisco SDWAN. That's the popular term in terms of data center automation. We have the ACI or datacenter, we have the ACI approach. Correct. Now all these automation or all these STM solution that you are seeing here, they are 100% policy driven, correct? So policy is the key. Not we are looking for the network services but we are thinking that with help of those network services how we can influence the application. That's the key to the policy-driven approach. The following critical point is that ACI provides software flexibility and scalability of hardware performance, resulting in a robust transport network in his networking workload. Now this is the long term, but short is this that either it's a hardware basedserver, the common term is the bare metal. So it's either bare metal or a virtualized world, such as one hosted over any SXI or any type of VMware or any type of content, and so on. ACI is capable of understanding all different types of workload. Even so, they are moving from place to place. Correct. Now the third very important point is that ACI is built on a network fabric that combines time-tested protocols with innovations to create a highly flexible, scalable, and resilient architecture of low latency and high bandwidth. And that's a term, we have that term as a cloth fabric. So you will see that you have the Leaf spine, says leaf spine. And again, suppose leaf one, leaf two, one spine, two spines, three, etc., so it's a clause in architecture where all the devices are only one hop away. So all the leaves actually they are one hop away from this. Fine. And that's why it is providing the resiliency low latency and the high bandwidth. Okay, so how does this look? Here, you can see that your management plane is Epic, and then you have your brain, and then you have your data plan. Again, this is not a completely stable solution from Cisco, but it is still the spine and the leaf because the leaf has some intelligence. So you can't tell that Leaf are dump boxes, but they can also make local decisions. But the Leaf are there at the data plane and then you have the control plane device and then you have the management plane. So you can write everything over the Epic controller, and then you can push to the spine or the leaf, but those policies can be pushed down to the control and data planes. Okay? And let's say you can see that you have the Leaf switch, you have the spine, you have Leaf, and then you have the application policy infrastructure controller. What are the rules? Although these rules listed here are few because we have just started understanding this. So the role of spine is they are working as a route reflector. So whatever endpoint information they learn, they will go and reflect on some other devices with different protocols. We'll go and discuss more and more about that. Here you have the fabric in between. So you have the fabric, and you have the IP reachability inside the fabric. All the links are working as an ECMP equal cost multi pathing. So here are the points that represent the backbone of the ECA fabric. Connect leaf switches. That's true as per the diagram as well. Then represent connection point for end devices. Any type of end devices will go and connect here and connect it to the spine. That's true. So one link going to a spine, one link is going to the end devices. And then you have the common policymaker that is your Epic controller. You can write the policy, and then you can push this policy to the underlying devices. So again, let's do the summary. What are the benefits that you have It workflows and application deployment agility. They are going to support 65+ ecosystem partners. They are secure because they are not trusting anyone. So this is purely working as a whitelist model. Whatever you have to go and specify those things are going to be trusted. You can think like this, trusted and trust us. The thing means you have to create some sort of agreement in between one party to other party and then those parties through that agreement called contract, they can trust or untrust to themselves. And this is true for the application, this is true for the end point, this is true for any type of user and the other user, correct? So like that it's so much flexibility is there but still the entire infrastructure is secure because itis supporting the white listing of the listing model. Okay, so we have the void listing model. This is policy enforcement, it is supporting micro segmentation and the analytics. Great. Then again we have the virtual workload and we have the centralised policy manager. So these are the key benefits that are listed here. Again the summary of the key benefits that we have the deployment agility, they are supporting 65 plus ecosystem partner, they are supporting policy enforcement, micro segmentation, analytic. Everyone is looking for analytics nowadays and then not only it is supporting the physical, but it is supporting the virtual workload as well.

3. What is ACI Ver 02

Let us understand ACI fundamentals. ACI is application centric infrastructure. It is something like you have the leaf and spine structure and all these leaf and spine backbone, they are managed via some controller called Epic controller, application centric controllers. Now, the question arises here that way, in the first place, we need this type of structure. Now, you can think in the perspective of the developers. For developers, what they want for application developers, they want that okay, client has some requirement for new application. They want to write new code, they want to deliver it as soon as possible. What are the constraints these developers have? These developers, by the end of the day, they are depending upon the network infrastructure. And in the traditional network infrastructure, what are the things we have? We have VLAN subnet protocol, sports, routing, routing constraints, security constraints, means we have barrier everywhere. It's not that the traditional network is there for application. It means network is there. You can write your application, the application will flow over the network, but it is not application based network, it's a network. On the top of that you are writing the e writing the applicaNow, considering these barriers into mind, this particular ACI infrastructure that is application in the centre infrastructure has been developed. And if you see that Acai infrastructure, you will find that yes, exactly. We have different type of tiers like Web, app and DB. That is the three tier application infrastructure. And then we have relationship life provider and consumer and the contracts in between them. In general, in short, that now the network is there to support applications, means the application is in the core and then we are designing the network. That's the overall idea we have with the ACI. So what is ACI? In the old IP endpoint based network, but in the new we have application based network. So initially we have IP based network, still we have IP based network. But all those constraints related to VLAN, subnet, port protocol, security, those have been removed. And now you can think network as the application. And even you can change your networklike you have some programming structure or you have some template based structure from where you can change your network structure. Sometimes maybe you heard about some terms like now we have underlay and overlay type of network. And that's true, that's actually true. That you may have any type of physical connectivity in the bottom. So here in the down you may have any type of connectivity, but the view of the network will be changed. So your underlay and your overlay. So this underlay has been abstracted. And when you are talking about the overlay overlay view, maybe change. So on the top or from the top, you can control your physical infrastructure. So here you can see that, okay, I have options that I am doing application based networking. We have new software based network. We have software defined networking management. Now, we are not declarative but we are giving the promise to the customers. So now the new model is that you want to create 1000 VLAN. I will create you want to create 100T names, I will create you want to create 1000 subnets, I will create those things. Like we can guarantee, we can promise to the customer that yes we'll do, and we'll do it very merthat yes Because not only they are supporting various object model, but we can do the programming as well. Now finally you can see that the pure STM thing, that separation of control and the data plane and that's true that your data plane is somewhere, your control plane is somewhere. Even you have your management plane as well. And from your management plan you are writing the code to the control plane and by the end of the control plane pushes all these things to the data plane, just like that. Now here you can see that ACA fabric is an IP based fabric that implements an integrated overlay that we have discussed, that we have under the overlay allowing any subnet to be placed anywhere. That's the true power that now we are not fixed with the network. We have the mobility inside the network. That means that any subnet to be placed anywhere in the fabric and supports a fabric wide mobility. So now you can see we have new capability for the subnets that my subnets can move. So we have the mobility feature for virtualised workloads. STP is not required within the ACI leaf and fabric the spine and the Epic doesn't run STP instances. So now we are in the complete ACI fabric, that is the leaf and spine type of structure. And here in the diagram you can see that you have the cloth fabric where you have leaf, sleeve, spine type of structure. And all these fabric we are managing or we are monitoring from the Epic controller. Now in the core you can see that these devices, they are running Nexus Nine K. So either 9300 as a leaf, 9500 as a spine. This leaf and spine of this Nexus can run either in a standalone mode or in ACI mode. So we have dual benefit. Not only we have the dual benefit, but herein the points you can see that this particular switch actually this is based in the class efficiency. So it's overall throughput it's performance, it's support to the programming means you can programme the Essex. All those things have been very enhanced, very modernized and that's why they are based in the class. So the Nexus 9 K platform has two modes. Standalone ACI utilise an enhanced version of NX OS operating system. That's the key. We have the enhanced version of the operating system to provide a traditional switching model with advanced automation and programming. So you can see now because I have very good hardware, so I can do the programming because I have modern ASIC inside this. So I can do the automation programming, I can enable new capabilities. That's the key we have in the second mode. That is the ACI mode in Nexus Nine K provides an application centric representation of network as a whole. So here we have the application centric infrastructure. If I run in the ACI mode utilizing advanced feature and profile based deployment, abstract the complexity of the underlying network and improving application visibility and greater business agility through the DevOps methodology. So you can see that how many key features we have inside the Nexus Nine K box. Either it's 95000r, it's 9300 because of the advance or enhance that we have inside these boxes, they have much capability to run as an application centric infrastructure to do the programming. It has more visibility and it has much better throughput as well. Now, later we'll discuss about the application infrastructure inside the ACI. So for the moment, let's stop here and we can continue in the next section with some more theories. ACAE.

4. ACI Topology & hardware 01

Let us learn about the ACI topology and hardware one by one. Now, as we already know, we have discussed previously that we have leaf and spine structure and the structure is in a way that we have following ECMP like equal cost multi pathing because all these fabric, so when we are talking about the fabric, the connectivity, the interconnectivity between the leaf and the spine, they are the ACA fabric and this fabric is following the IP based fabric. So first of all, here you can see the diagram where we have the spine and then we have the leaf and obviously then you have the end points as well. Now endpoints may be physical, they may be virtual and that's the whole key we have. So once you have this structure and then you want to figure out that what will be the fabric. So here you can see, let me try to draw here. So in between the leaf and spine we have the fabric and this is IP based fabric. The question here is that okay, you have the fabric and this is IP based fabric. Then what about the eligibility? What are the common terms we have in have in between? So that I will show you in the next slide. But once you have this leaf and spine structure, you have simple and consistent topology, you have the scalability and better use of bandwidth obviously to reach to certain locations. Suppose if you have one leaf and three is fine, then you can see that from leaf to a spine you have three equal path and later on you will understand that from one leaf to other leaf, you have three different tunnel to each. Now, since you have three different channels, obviously if it is an IP based tunnel, then you will do ACMP and that's why we have the symmetry for optimization and forwarding behavior. Again, next section means section number two, you have full forwarding behaviour and all those steps that you need to learn. So whenever we are talking about endpoint learning and forwarding, that's actually itself a big topic and we'll learn and understand in the next section, section number two about the forwarding behaviour now because we are in the cloth fabric, so that's why it's the least cost design for high bandwidth. So now we are talking about the endpoint and now we are talking about the ACF fabric. So what are the common terms and how they are achieving the IP reachability? Correct IP recipe has been done with the ISIS protocol. And good thing about this is100% optimized, this is 100% automated. So we don't need to write any code or any configuration for this fabric IP reachability, this is 100% automated at the moment you go and connect the devices and then the discovery process will start happening. What you need to do, you have to do an assigned minimum thing like the block of internal IPS, the management IP, et cetera, et cetera. Once you have this minimum things assigned automatically system will go and assign these IP addresses to all these devices. So here you can see the very important term that we're going to use. Here is the tip tenl endpoint. And again in the upcoming section more and more you will go and learn about this tenant point or V tips as well. So there are several terms tip, V Tip, P, Tip, et cetera, et cetera, that you will understand in sections upcoming slides. So at least what are the two very important things from this slide is that you have the Is IP reachability, correct? And then you have your tunnel endpoints. They are termed as a retail. So whenever for example, source obviously it is connected with leaf. Then you have the fabric. So this cloud is nothing but the fabric where you have leaf and spine structure and then you have the other leaf. And then maybe you have the destination. So you can see the source will go reach to the tip tunnel endpoint. But tip is forming the VXLAN tunnel. That's the data plan tunnel in between the leaf. Suppose leaf two and leaf three maybe because here the source is at leaf one and destination is at leaf two. This tunnel will be created. And these are the dynamic automated tunnel from source to destination. In between we have the IP eligibility and then this hand off will happen, the source and destination will communicate. Still we have section two where you will go and learn in detail that actually what is happening, who is the gateway, how l two learning happening? How l three learning happening et ppening? How l tSo two take points are that you have IP reachability via ISIS and these endpoints are nothing but the tunnel endpoints. Now we know that we have this fine where you can see them in the diagram that you have the spine switches and spine switches you can think that they are working as a router reflector as well. So what is happening that these leafs, they are learning the end point information and then they are telling the actually they are sending some sort of digest message. Here you can see that you have the coup database. Who is maintaining the coup database, that's fine. So they have the Council of Oracle protocol where they are managing or storing all the endpoint information. So whenever as a leaf I am learning the endpoint information, I am sending that message to the spine that his spine, this is the end point, he'll learn now what your spine will do, they will replicate that information, that endpoint information across all this fine. And they will store those information as the process. Behind this is the coup counsellor for Eco Protocol. So not only that's fine, they are learning it, but they are doing the proxying as well. So think that they are working as a route reflector. Not 100% correct term but as a loose term that they are working as a reflector a moment they are learning if anyone doing the query or they have the capability that they can teach as well so they are learning and then they are teaching like that so in this way they are maintaining the data plane or you can see that the database for the endpoint. So again from last slide you have the important thing that for example what is the dynamic tunnel? What's the nature of the dynamic tunnel? What's the protocol behind that? So answer is that VXLAN but where to where this panel is forming. So from one leaf to other leaf, the leaf are termed as a tunnel endpoint and the tunnel they are forming, those are nothing but the VXline tip VXLAN tenant endpoint et cetera that we can refer great then it's fine they are storing the database or the endpoint information in terms of database and then they have the authority that they can teach as well. So throughout we should have the consistency so we have the IP reachability, we are learning all the information and we are advertising as well those information so whenever endpoint want to communicate with others their endpoint they should not face issues. Again how this learning process is happening we have to go and learn in the next section now behind the scenes is this that you have the data plane protocol.

5. ACI Topology & hardware 02

Let us start where we have left off in the previous section the Vxlan header. Now what is significant here and the important thing here is this that in upcoming section that section I have three videos dedicated to VXLAN. So you will learn more and more about the header format and the use of different sections of field inside the header. But if you want to learn it just to how this VXLAN is working and what are the key component inside that just overview that you have the inner obviously inner header and then you have the VXLAN header then you have the UDP header and you can think of the encapsulation so you have inner then you have the VXLAN and then UDP and then you have the outer header obviously the inner header plus the payload whatever we have those are the actual source and destination. Then we have a data plane header. In terms of VXLAN in VXLAN itself we have different types of fields that you will learn and understand in the VXLAN section in section number two where we have dedicated sessions related to VXLAN. But important thing here is that this VXLAN Mblan so we know that we have the virtual land and in modern datacenters we want in number of lands or in number of virtual lands so we need this VX land concept in VLAN we can create say for example 4K number of virtual land but with help of VXLAN you can go and create up to 16 million virtual lands and that's the virtual land extension we have extensive land correct so here you can see the header format and the capability that the VXline is providing detail obviously we have in the upcoming session where we have the dedicated sessions for VXline. Now what ACI is doing they can take anything as an input and that's a normalisation process either it's Ltwo packet or L three packet or eight zero twenty one Q packet or Nvgia packet or any type of input but they will go and do the normalization. So that means that you have the inner packet and then they have the VXLAN and UVP encapsulation and then we have the source and destination addresses means suppose this leaf one is the source and the other leaf that's leaf six is the destination. So basically in the header you have the source and the destination IP addresses as external header or the outer header and then the packet will be forwarded from one location to the other location. So that's the key we have that's the normalization process we have once they are inside the fabric we have to any communication. So who is the key component, who is the game changer here the key component and the gamechanger is the epic itself that's the controller itself because from the controller we can go and manage the entire data centre devices entire fabric correct now when we are talking about this controller we should know about this controller and hardware. So you can see that you have two form factor, you have the M three, that's the medium range and then you have the L three, that is the large range. They are the UCS type of plate server where you can go and check the processor, memory, data, hard drive, PCI, etc. For if you want to learn more about this, you can go and check the data sheet related to Epic M three and Epic M three controllers. Now, how you're going to join or how you're going to connect this. So you can see that the configuration is straightforward. One of the Epic we are connecting with two of the leaf like that we have minimum, you should have minimum three Epic in a cluster, but you can have more as well. Minimum three is required even to open the tags as well. So here you can see that I have a controller connected with two lips, I have the controller connected to lips and then we have the Epiccluster that's out of band management. Also we can do and configure here. Now again, this is just the front and there are a view of the UCS series blade server. This will look like this, but obviously this is a big heavy blade server because we are going to do so many tasks from these servers. Alright, now while these slides are here that I'm showing you that you may have the cluster size three, five, seven minimum, you should have even minimum Because they are going to use the Sharding. That's the database concept. Even if you want to learn more about the sharding, again you can do a little bit of Google about that and you will get the information about Sharding. There are nice documents and white people's related to sharding capability and Sharding database, how it is working. So here you can see that you have the mirror. So you have the Epic database one, two and three. And suppose if one of the Epic or one of the database will go down still you have two in a cluster that will serve the purpose. But suppose if two of them will go down, then in that case the third will become as a read t case the So that's why actually for high availability and redundancy we have three databases in a cluster because if one will go down still you have a rest of the databases in a cluster who can serve the purpose. Again, if you want to know more about sharding, you can have a little bit of research on this. And we have nice Cisco document as well related to Sharding. So now suppose if we have more controllers, if you want to use more controllers, we can use it. So in that case, what will happen if any of the database or any of the controller will go down? The backup controller will go and take the action as an active controller. So we may have the concept we have the concept of the Active and the standby controller. If anything will happen with the active controller, the standby will become the Active and it will join the fabric. So again, you have the norm of three controller in a cluster. So minimum three controller in a cluster, still you have if you have using few of the controllers as a standby and if Primary will go down, the backup will come into the rimary will go So let's just stop here and next section will learn about the ACI hardware.

6. ACI Topology & hardware 03

The next important section, we have to learn about the Cisco ACI hardware. Now, what I will do that I'll break this session into two parts. So first of all, let us learn about the Epic related hardware and then we'll go and learn about the Leaf and fine hardware. So now we know at this point of time that we have Epic hardware and that is nothing but UCS series Blade server. The important point here to note is that we have generation one, two and three. So here you can see that you have generation one, generation two, and now we are inside the generation three. For this generation we have M three, that's the medium, l three, that's the large. So depending upon the deployment, we have the medium and the large server. Inside that we have preinstalled Epic software and that is coming with the IMC or SIMC. That is nothing but Cisco integrated management controller. Now, if you want to learn more about this, if you want to know more about this Epic hardware, you please refer the data sheet. So you will come to know about different type of capabilities related to CPU Ram processing, how many say for example, tenant it will support, how many endpoint, how many application profile, et cetera. So those numbers you will get in the data sheet and according to that, we can plan that what type of Epic we want to use inside our data centre data centre deploy Let us discuss quickly about the Epic redundancy as well. Now we know that when we are doing the deployment for the Epic, we should have minimum three Epic in a cluster. Now these Epics we can go and obviously we should have out of band management. And then these Epics, they are connected with Leaf as well. So depending upon what type of redundancy we have, what type of design we have related to these Epics is how we are going to do the connection that we can achieve the high availability. So here you can see that you have two connection. One is going towards the fabric and that's very important. And then one connection we can use for outer band management as well. So while either you are going to the fabric or you have outer band management, both the cases, you should have the redundancy. Now here we'll see that the redundancy will be there in the design. And for that reason we have the bonded interfaces. Those are dual home to Tulip and this is the best practise that we are talking about. So we should have bond interface and we should have dual homed connection. Obviously, if one link will go down, you have other link as an, you have other So there are two bond interfaces. For example, bond Zero and bond one. This bond is zero. The connection with bond Zero is with the fabric and bond One is with the auto band management. So, straightforward. First of all, as per your deployment, you have mark out that I'm going to use either Mseries or L series, UCS blitz server, depending upon how many tenants, application profile, endpoints user, export, etc. Once you have done that selection, then when you are connecting the interfaces, those steps are very important before doing the actual deployment. And you will see that there is some prerequisite we should follow certain rule or certain steps before doing the actual installation. So what are that checklist that you can go and refer in the deployment section. Once you have completed the design aspect, your planning aspect. So your planning is done and you know that you are going for L or M Epic, then next you want to do the High Availability redundancy design. For that you have Bond Zero and bond one. One is going to the fabric, other one is going to the outer band management. And here in the diagram you can see that you have Bond interface and then you have Bond Zero and bond one. Obviously one is going to the fabric and one is going for Auto band management. Now containing the same strategy of high availability and redundancy. What we are doing that whenever we are deploying the Epic controllers, we should have minimum three Epic in a cluster. But still, suppose if one of the Epic will go down, suppose the hardware issue or any other issue. So at that time, just to prevent such type of events to be happen, what we are doing that we should have one Epic. So I have one Epic ready as a standby. And suppose if this Epic will go down, this Epic which has the same software version, which has the same firmware, it will go and join the fabric and then it will go and take the same chassis ID, same identity, and then you can say that the replacement of Epic is seamless correct. So that approach also we are using. And then for that approach I have completed steps. So for example, you should have minimum three Epic in a cluster. Let me remove this marking, let me use this spotlight. So here you can see that you should have three in a cluster. One could have a standby Epic. So anyone in that three, you have one standby Epic as well, that is working as a standby mode. And when you want to do there placement, either hardware failure or any issue, that particular Epic, that particular Epic. Whenever you are upgrading the firmware for the active Epic at that time, you are upgrading the firmware to the standby Epic as well. The standby controller as well. So what will happen that if one of the Epic will go down, the backup will go and take the role. So here you can see that the backup has the same firmware version as the Active cluster. Now, each Cisco AP controller, active or standby, has an ID designated. I will show you in the lab section you'll see that these epic, they have certain ID. Now, once this particular Standby will take the role of Active, he should go and take the ID name that is there using the Active as well. And this is the process by which we are changing the epic. We are changing the role of the epic. And then we are achieving the seamless type of you can say that disaster recovery because one of the epic is down. Obviously, two databases are two remaining. Two epic will take the role, but still you don't want to risk with two. So that standby will come and join the fabric. And again, you have the three epic as per the standard.

Hide