Data Center NETWORKS (what do they look like??) // FREE CCNA // EP 7

have you ever been inside a data center it's amazing rows and rows of switches routers servers just buzzing and humming powering the internet we use every day getting goosebumps just thinking about it and as you're learning networking you've gotta wonder how does all that work what does a data sensor network look like is it different from any other network that's what we're talking about today how do network engineers go about designing data center networks and also a huge shout out to bozon software for sponsoring this video they are the official sponsor of this course to help make it free here on youtube and they just happen to have the best prep software out there for the ccna if you need labs exams courseware they got you check it out link below now data centers are different right like if you look at a corporate office you got people trying to connect to things and the internet a data center is all about servers connecting to things and the internet but for a long time we kind of designed their networks the same way with the same model but first i want to start with this what exactly is a data center what's the purpose of it and also when will you get to work on a data center in what situation would you have to step into one and do stuff in one which you probably will if you're going down the networking path you probably it's going to be awesome like i'm so excited for you i remember my first time at a data center and it was nothing short of magical felt like i was at disneyland now i'll start with this pretty much any resource you access over the internet or just in a network period is going to be inside a data center and you've seen data centers before either on tv or what i just showed you earlier i mean here's one i'm actually doing a 360 tour of google's data center just rows and rows racks and racks full of servers routers switches like i said disneyland so like right now you may be watching me on youtube and this is a google data center so you might be watching me from this server right here it's streaming to you from here okay big companies like google have data centers facebook has data centers massive data centers amazon microsoft and if you join their network team you might get access and be able to work inside those but what about other companies that aren't as big as google facebook or amazon do all companies have and use data centers the answer is yes but not all in the same way so let's say we have network chat coffee get yours today network chat coffee sells coffee online through a website a website that lives on a server so for me to offer that service to you i got a few options i could open up my own data center which doesn't always mean having a massive warehouse with the rows and rows of servers it might just be one room with one rack which is what i have in my house i've got my one rack of equipment some routers in there switches servers and you can access my stuff directly from my internet connection a lot of small to medium sized companies do have a situation like this one of my first ever jobs doing network engineering they on their first floor of their building had a little small networking closet that was like a data center it had three racks full of servers routers and switches and that was our web presence that's where we ran our websites that was everything now maybe my home internet connection isn't great or i don't have adequate air conditioning in a raised floor and i don't have everything to meet the standards i need to actually run a company well then i might want to put my stuff in a real data center and they do offer all those things redundant cooling redundant power all things you need if you're going to run a company so i would rent space in a data center instead or rent racks this is what the majority of the companies i work for have done now this will look like one of the larger data centers it'll have rows and rows of racks of servers and stuff but they're from different customers that the data center has you might have joe zebra farm over here bob's biscuits right here in the network check coffee over here you can rent you know two three four maybe a couple rows of racks and so you put your servers in there your routers your switches and the data center would provide you with power and internet connection now what i actually do for network chat coffee is i use the cloud the cloud still involves a data center but it's someone else's data center and someone else's servers so amazon aws google cloud microsoft azure they offer their servers and their data center to you as a service you don't need to rent space or have your own hardware or anything you just say click click i want my website to be right there and it happens so now when we talk about data center network design if your company's fully cloud which is rare if they're all in the cloud you don't really have to concern yourself with how they design the network because they handle it for you unless you're a network engineer for one of these cloud providers then you know all about this and we'll cover how they do it because it's not very different from how you would want to design your your data center up here so if your company has its own data center in its building or they rent rack space in another data center this is where you come in and you have to know how to design a data center network and again a lot of companies have this situation they actually might have a combination of all three of these here the first company i worked for did have this but they eventually did rent space in another data center we retained both of those and we added stuff to the cloud that's called hybrid cloud and that's a whole other video so let's talk about how you should design your data center network so this is what i would do if i were designing my network chuck coffee data center traditionally the old way let's say i rented a bunch of rack space in a data center i have on my racks here each of these racks would have a bunch of servers doing all kinds of stuff websites databases and for the network in each rack i would have one or two switches right at the top we actually call these our tor switches nothing to do with the dark web it stands for top of rack switches tor because yeah it's at the top of the rack and the servers in each rack will connect to their switch now how do all these servers talk to each other and get to the internet well i'd have my next layer my distribution layer if you watched my last video this might start looking a little bit familiar to you our top of rack switches which we could also refer to as our access layer switches will connect to our distribution layer switches redundant connections as we should have and then we'd have another layer of switching our core switch our big daddy beast which i know looks like a building but this is a switch it's a switch a multi-layer switch and that's what our distribution layer switches would connect to axis layer distribution layer or aggregation layer and our core or our core layer now just so you know the distribution and core layer in a real data center would not be out floating in the air like this they would often either be in their own dedicated network rack or maybe resting in some extra space in one of these racks here now this is bothering me with the proper three-tier design you want to have some redundancies let me add another core switch here which again has a beefy massive chassis switch that cost a ton of money but you gotta have it if you're going with this design that way we have some redundancy some resiliency and then if we had more racks in that data center or for like facebook and amazon and we own the entire data center and all that stuff is ours they would also have an access layer with the top of a rack switch with distribution switches and those would connect back to the same core as the other racks did now here we have a huge problem which you might be going well chuck it seems okay right like we have redundancy if a one switch goes down things are still up and you know what you're right about that but we started to notice a problem you see things started to change you see it used to be that we only cared about the traffic that went to and from the internet and we designed our networks that way what does that mean like if me here wanting to buy some coffee because i ran out i'll access network copy through the internet it'll take me to my data center to my core layer my core switches which will then get me down to my access layers or sorry distribution layer switches down to my access layer switch and then down to the server i need to be at and of course vice versa the server will give the information i need through that same path this is what we call north south traffic where northbound and southbound traffic that's what we cared about and that's what we prioritized with our designs but things started to change you see when this design was popular we didn't worry too much about these servers right here talking to let's say servers and storage and resources over here between the racks this type of traffic was not important until virtualization came in with virtualization our data centers became more distributed which basically means that servers in this rack now have a pretty stinking good reason to talk to servers over here in this rack and maybe in this rack and this rack basically servers within the data center in various racks want to or need to communicate with each other that's what we call east west traffic oh and i didn't label this before so this was north south and this is east west kind of horizontal communication now the problem is that we didn't design our network for east-west traffic we designed it for north-south traffic but with the big changes we have with technology east-west traffic accounted for the majority of our traffic in fact it's 80 of our traffic in our data center and it has the biggest bang the biggest impact on the way our network performs so with our three tier design if one server let's say the server down here wants to communicate with this server or it needs to because of our new virtualization technologies with three tier it's gonna go from our top of rack switch to our distribution layer and then it hops over to our core layer and then down to the distribution layer here and then to the top of rack switch and finally to that server now that might not seem too bad but that's way too many hops for communication that needs to happen crazy fast and with today's technologies things have to be quick it has to be crazy fast and this wasn't cutting it we had to go back to the drawing board how do we fix this not only that that wasn't the only issue we also had things like hey we got these redundant links between our topper rack switches and our aggregate layer switches but with spanning tree which you'll learn about later but this is basically what it does it will shut down one of these links and keep the other one active it will do that on each one of these redundant links avoiding what's called a switching loop which will basically blow up your network so three-tier design the campus design not for data centers we need something faster something quicker something a little crazy so let's take this business out of here we don't want this i'm also going to take out the internet for a second and we're also going to take out our core layer we don't need that anymore goodbye core layer actually let's just start over it's kind of messy [Laughter] okay new design here we go we're still going to have our top of rack switches but we're going to give them a new name we're going to call them leaf switches which might sound weird i know but legit l-e-a-f leaf and they're still going to serve the function of being the access layer for the servers in those racks now here's where the real difference is we have a new type of switch it might seem like he's doing the same job as the distribution layer and in a large way he is but we don't call them distribution switches now we call them spine switches and when you hear the term spine i want you to automatically think about backbone keep it all back related and networking our backbone much like our body supports the entire network it has to be able to handle the most bandwidth the most packets going through it it's going to be a beefy switch or a beefy router but in this case it's a bit different because each one of these switches aren't going to be backbone caliber like in the model before with our three tier model the core layer is a beefy multi-chassis beast like don't mess with that guy he'll kill ya but with this model these guys are they're they're powerful moderately so and here's how they connect our topper rack switches are our leaf switches we'll have something kind of weird which might it'll feel weird we're going to have a full mesh and things are about to get messy meshy sorry i couldn't help myself i'm going to use red this tupper rack switch will connect here and here and here and this top rack switch this leaf switch will connect here and here and here what am i doing here notice each leaf switch is connected to every spine switch we're basically creating a full mesh between the leaf switches and the spine switches like i said meshi this is what we want this is what we call our spine leaf design it's often referred to as a cloth design as well this is the standard right now this is how people companies businesses facebook google design their data centers they might modify a few things here and there but this is the base design they use but why why is this better this looks messy why would we do this this is not clean at all cabling nightmare and you're right that's one of the downsides is when you do a design like this you better have a lot of cables you better have some fiber some ethernet whatever better you know fiber fiber you're in a data center use some fiber told you things will get meshy last time i'm going to use that i'm sorry now let me show you why this is a powerful design remember we care about east west traffic we want this server to be able to talk to this server as fast as possible and with this design here's how it happens he says yep i'm ready to talk how does he get there well using any of the spine switches he can go bam one hop bam two hops two hops and that's always the case all the time every time no matter where you are in this design if i'm this server right here and i want to go to this server over here i'm one two hops away just like that whereas before in the three tier design it could be three hops or more it's unpredictable and that's the key with the spine leaf design it's predictable and it's reliable because again we care more about the east-west traffic server to server traffic than we do the north and southbound traffic to the internet to the user because east west is 70 to 80 of our traffic in the data center now quick recap of things i want you to remember these right here leaf switches these right here spine switches spine it's our backbone slightly beefy can handle a lot they are our workhorses now they themselves don't have to be these massive multi-chassis switches because they share the load amongst themselves and that's the other thing i want you to know the traffic is low balance across these equally which is possible because you know we have a full mesh here every leaf switch is connected to every spine switch also notice the spine switches themselves don't connect to each other and yes with this design we go a bit cable crazy but man a maximum of two hops for any server to server communication now let me show you what these switches look like so you have a visual here we have what's called our cisco nexus class switches if you're just getting into the cisco world the main switches we use for like access distribution core and like the campus environment are our catalyst switches the nexus switches are for data center they are specifically designed for data center they're crazy the throughput is mind blowing like look at this this two ru unit which means two rack units it's a thick boy um 12.84 terabits per second of bandwidth or throughput that's a pretty beefy spine switch and then right here we have a smaller one which is our 9332c it's one ru or one u but it's still amazing 6.4 terabits per second of bandwidth here's a list of their nexus switches now they do have these massive chassis look at this guy this guy will make you cry he's so massive and crazy but i want to show you this real quick if we get back down to our 9300 series here notice it says it can be a leaf or a spine and that's up to how big your network is if you've got 30 to 50 000 servers you're trying to accommodate you're gonna need some big stuff now one more crazy thing i want to show you about the spine leaf design and that's how these leaf switches are connected to the spine switches this connection right here most often is not layer two it's layer three you may recall from the other videos in this course layer three is routers dealing with ip addresses layer two switches deal with mac addresses typically when you connect two switches together that's gonna be layer two and that's a lot of what you would see with a three tier design and we'll cover more on layer two and how we do all that later on in this course but i wanted to point out that these connections switch to switch are not layer two they're actually layer three which means yes this leaf switch and this spine switch they're both layer three switches or multi-layer switches and this does two things for us first is we don't have to worry about any one of these links being blocked by spanning trees spanning tree is a loot prevention mechanism at layer two we'll cover more on that later but in a spanning tree world two of these links would be blocked by default only one link being allowed to prevent a loop but because we're using layer three and we're routing we could low balance across these links keep them all up and use the full bandwidth we have and that was number two yeah we can we can low balance between these now data center design does not stop here it gets a lot crazier this is just what we call the underlay kind of the foundation network on top of that we'll have what's called the overlay network and this is how all the big players design their networks facebook amazon google whoever they are the spineleaf architecture design just gets our stuff connected the overlay does some insane network automation cisco aci stuff i'm throwing buzzwords out at you and i'm not going to go into it right now because we will cover it later on but just know we're scratching the surface that's it hey were you paying attention because right now we're going to quiz you we're going to see what you know we're going to pull some questions out of the boson xmax for ccna which is the best practice exam for ccna i highly encourage you to check that out in the link below anyways let's see what you got which of the following devices cannot be connected to leaf nodes in the cisco aci architecture now this might throw you off a little bit but let's see what you got ready set go now before i show the answer i do want to say this seeing cisco aci might have thrown you off a little bit but you may remember that i mentioned aci as one of the network programmability or network automation platforms that depends on the spine leaf design as its underlay or network underlay so if you saw a leaf node you could have made a jump to hey maybe they're talking about spine leaf and if they say leaf node maybe that also means the leaf switch which it does so given what we learned which device cannot be connected to a leaf node now right off the bat we can say uh spine nodes that can't be true because spine nodes do connect the leaf nodes that's how the architecture works now you haven't learned about epgs or apex and we haven't talked specifically about application servers that leaves us with leaf notes can a leaf node connect to a leaf node well in that architecture no they can't let's see it for right select it show answer and we are indeed correct if you got that right fantastic job you're one step closer to getting your ccna and not available we're going to think hey spine leaf let's think about that let's think about that anyways get set go alright welcome back let's see how you did so we have to select two choices here what's true now things we can take off immediately first spine nodes must be fully meshed which might have been a bit confusing but what they mean is are the spines connecting to each other in a full mesh well we know that spines don't connect to spines so that's definitely not true of the cisco aci architecture or the spine leaf design what about b each spine must connect to every leaf node hey we do know that to be true let's select that sucker yep now see we haven't talked about so that might be like i i don't i'm not sure d leaf nodes must be fully meshed no they don't because leafs don't connect the leaves let's take that sucker right out and then e each leaf node must connect to every spine node hey we know that to be true i think that's right so you know we got two answers that we know to be true let's um let's check it out did we get it right yeah we get it right there's bozon's excellent explanation if you got that right man you're killing it again one step closer to ccna and these questions are often considered harder than the actual exam that's how boson is they try to model the exam as close as they can and often and many people will agree it's more difficult so if you can pass bozon i think you're ready well that was ccna episode 7 we talked about the spine leaf architecture it's how we design our data center networks you know what honestly it's how we're starting to design our campus networks as well but i'll get into that later it's a crazy world now with networking everything's changing you're coming in at a good time trust me anyways let me know what you thought about the video below if you like it like it if you haven't already subscribe we just hit 500 000 subscribers which is i mean i don't know what to say so thank you you guys are amazing and if you want to help me do more of what i do here making courses making content that's educating you helping you get that next job helping you advance in it consider joining my membership either join the youtube membership the join button below or go to this is it.io that's a membership a mission that me and david bomb will run we both have committed to producing free content we're both doing a ccna right now which is crazy we're also working with jeremy chara on a security plus course completely free here on youtube and joining this is it helps us do more of this you also get access to all of our stuff on this is it before we release it here on youtube so check it out anyways that's all i got i would love to hear what you guys think of this video below i love seeing your comments i try to respond as many as i can i just love seeing you guys and if you need community if you need help if you just need support an encouraging word consider joining my discord server below a ton of people in the industry who are willing to help you out and encourage you anyways that's all i got today guys i'll catch you guys later [Music] you

Channel: NetworkChuck

Views: 252,409

Rating: 4.9496908 out of 5

Keywords: ccna, cisco certified network associate, ccna 200-301, free ccna, ccent, icnd1, icnd2, network chuck, networkchuck, ccna training, ccna course, network design, collapsed core, ccnp, cisco training, spine leaf topology, spine leaf network architecture, spine leaf architecture vs traditional, spine leaf data center, spine leaf ccna, spine leaf cisco, spine leaf vs 3 tier, spine leaf architecture benefits, spine leaf campus network, data center, data center network

Id: 6-66D9J5PkY

Length: 20min 17sec (1217 seconds)

Published: Sat Oct 10 2020