1 00:00:00,000 --> 00:00:01,080 In this lesson, 2 00:00:01,080 --> 00:00:03,780 we're going to cover some datacenter topologies. 3 00:00:03,780 --> 00:00:05,700 Now, when we're talking about a datacenter, 4 00:00:05,700 --> 00:00:08,310 this is any facility composed of network computers 5 00:00:08,310 --> 00:00:09,930 and storage that businesses 6 00:00:09,930 --> 00:00:13,410 and other organizations use to organize, process, store, 7 00:00:13,410 --> 00:00:15,900 and disseminate large amounts of data. 8 00:00:15,900 --> 00:00:18,060 Now, that is a pretty generic definition, 9 00:00:18,060 --> 00:00:20,820 but that is because a datacenter is used to describe a lot 10 00:00:20,820 --> 00:00:23,250 of different things these days, from very small 11 00:00:23,250 --> 00:00:25,350 to massively large areas. 12 00:00:25,350 --> 00:00:27,990 For example, at one of the smaller organizations I used 13 00:00:27,990 --> 00:00:30,090 to work for, we had a small datacenter 14 00:00:30,090 --> 00:00:32,610 that was roughly 150 square feet in size, 15 00:00:32,610 --> 00:00:35,130 basically the size of a standard bedroom. 16 00:00:35,130 --> 00:00:37,680 Within it, we had a single rack of networking equipment 17 00:00:37,680 --> 00:00:40,620 and about five racks that contained our various servers. 18 00:00:40,620 --> 00:00:42,810 On the other hand, one of the largest datacenters 19 00:00:42,810 --> 00:00:45,540 in the world is located in Bluffdale, Utah. 20 00:00:45,540 --> 00:00:48,360 This datacenter is known as the Utah Datacenter, 21 00:00:48,360 --> 00:00:50,130 or by its official title, 22 00:00:50,130 --> 00:00:51,930 the Intelligence Community Comprehensive 23 00:00:51,930 --> 00:00:54,480 National Cybersecurity Initiative Data Center, 24 00:00:54,480 --> 00:00:57,930 and it is 1.5 million square feet in size 25 00:00:57,930 --> 00:01:00,870 and it spreads out across 20 different buildings. 26 00:01:00,870 --> 00:01:05,340 This datacenter is so massive, it uses around 65 megawatts 27 00:01:05,340 --> 00:01:07,620 of electricity just to power it. 28 00:01:07,620 --> 00:01:10,080 This datacenter actually cost the US government 29 00:01:10,080 --> 00:01:12,810 about $1.5 billion to build, 30 00:01:12,810 --> 00:01:15,420 and they spend over $40 million each year 31 00:01:15,420 --> 00:01:17,310 just on the electric bills. 32 00:01:17,310 --> 00:01:18,570 Now, personally, I have worked 33 00:01:18,570 --> 00:01:20,400 at some really large datacenters, 34 00:01:20,400 --> 00:01:22,140 but nothing in comparison to the size 35 00:01:22,140 --> 00:01:23,880 or scale of that one. 36 00:01:23,880 --> 00:01:26,910 Amazon, for example, uses datacenters that range in size 37 00:01:26,910 --> 00:01:31,140 between 150,000 to 215,000 square feet, 38 00:01:31,140 --> 00:01:33,930 and each of these datacenters can host about 50,000 39 00:01:33,930 --> 00:01:36,180 to 80,000 servers inside of it. 40 00:01:36,180 --> 00:01:37,620 Now, as a network technician, 41 00:01:37,620 --> 00:01:39,450 you may very well end up working at one 42 00:01:39,450 --> 00:01:42,300 of these large datacenters one day, so it's important 43 00:01:42,300 --> 00:01:43,860 that you understand the different topologies 44 00:01:43,860 --> 00:01:46,320 or architectures that they're going to utilize, 45 00:01:46,320 --> 00:01:49,710 including the three-tiered hierarchy, a collapsed core, 46 00:01:49,710 --> 00:01:51,267 the spine and leaf architecture, 47 00:01:51,267 --> 00:01:54,210 and the different traffic flows that are used inside 48 00:01:54,210 --> 00:01:56,760 and into our different datacenters. 49 00:01:56,760 --> 00:01:59,910 Now, first, let's take a look at the three-tiered hierarchy, 50 00:01:59,910 --> 00:02:02,100 which consists of the core, the distribution, 51 00:02:02,100 --> 00:02:05,310 or aggregation layer, and the access or edge layer. 52 00:02:05,310 --> 00:02:08,340 Now, the core layer is going to consist of the biggest, fastest 53 00:02:08,340 --> 00:02:09,630 and most expensive routers 54 00:02:09,630 --> 00:02:11,400 that you're going to end up working with. 55 00:02:11,400 --> 00:02:13,770 This core layer is considered to be the backbone 56 00:02:13,770 --> 00:02:15,000 of our network, and they're used 57 00:02:15,000 --> 00:02:17,430 to merge geographically separated networks back 58 00:02:17,430 --> 00:02:20,010 into one logical and cohesive unit. 59 00:02:20,010 --> 00:02:22,320 In general, you're going to have at least two routers 60 00:02:22,320 --> 00:02:24,240 at the core level, and they're going to operate 61 00:02:24,240 --> 00:02:26,100 in a redundant configuration. 62 00:02:26,100 --> 00:02:28,530 After all, if you only have one core router 63 00:02:28,530 --> 00:02:31,470 and it goes offline, your entire network is going to grind 64 00:02:31,470 --> 00:02:33,780 to a halt and stop functioning. 65 00:02:33,780 --> 00:02:37,020 Next, we have the distribution or aggregation layer. 66 00:02:37,020 --> 00:02:39,330 This layer is located under the core layer, 67 00:02:39,330 --> 00:02:40,860 and it provides boundary definition 68 00:02:40,860 --> 00:02:43,320 by implementing access lists and filters. 69 00:02:43,320 --> 00:02:44,340 Here at the distribution 70 00:02:44,340 --> 00:02:46,830 or aggregation layer, we're defining the policies 71 00:02:46,830 --> 00:02:48,330 for networks at large. 72 00:02:48,330 --> 00:02:51,030 Normally, you're going to see layer three switches being used 73 00:02:51,030 --> 00:02:52,770 at this distribution layer to ensure 74 00:02:52,770 --> 00:02:54,660 that the packets are being properly routed 75 00:02:54,660 --> 00:02:55,657 between different subnets 76 00:02:55,657 --> 00:02:58,440 and VLANs within your enterprise network. 77 00:02:58,440 --> 00:03:01,140 Finally, we have the access or edge layer. 78 00:03:01,140 --> 00:03:03,300 This layer is located beneath the distribution 79 00:03:03,300 --> 00:03:05,370 or aggregation layer, and the access 80 00:03:05,370 --> 00:03:07,260 or edge layer is going to be used to connect all 81 00:03:07,260 --> 00:03:11,130 of your endpoint devices, like computers, laptops, servers, 82 00:03:11,130 --> 00:03:14,310 printers, wireless access points, and many others. 83 00:03:14,310 --> 00:03:16,260 These access or edge layer devices 84 00:03:16,260 --> 00:03:18,120 are usually going to be regular switches, 85 00:03:18,120 --> 00:03:19,050 and they're going to be used to ensure 86 00:03:19,050 --> 00:03:21,030 that packets are being converted to frames 87 00:03:21,030 --> 00:03:23,310 and delivered to the correct endpoints. 88 00:03:23,310 --> 00:03:24,750 Now, you may be wondering, 89 00:03:24,750 --> 00:03:27,210 why do we need this three-tiered hierarchy? 90 00:03:27,210 --> 00:03:29,400 Well, by using this type of hierarchy, 91 00:03:29,400 --> 00:03:32,580 we can get better performance, management, scalability 92 00:03:32,580 --> 00:03:34,680 and redundancy from our networks. 93 00:03:34,680 --> 00:03:37,200 This also gives us a better way to troubleshoot our network 94 00:03:37,200 --> 00:03:39,990 because normally if there's an issue, we can isolate it down 95 00:03:39,990 --> 00:03:42,690 to a single access or edge layer device. 96 00:03:42,690 --> 00:03:45,450 Now, once we've isolated it down to that single device, 97 00:03:45,450 --> 00:03:47,160 we can work on fixing it while the rest 98 00:03:47,160 --> 00:03:50,070 of the network can continue to operate unimpeded. 99 00:03:50,070 --> 00:03:51,870 Now, normally in your datacenter, 100 00:03:51,870 --> 00:03:53,790 you're going to find your core layer devices, 101 00:03:53,790 --> 00:03:56,250 as well as your distribution or aggregation layer devices 102 00:03:56,250 --> 00:03:58,470 for the local network in that building. 103 00:03:58,470 --> 00:04:01,380 If you have remote branch offices or other locations, 104 00:04:01,380 --> 00:04:03,360 each one of those will have its own distribution 105 00:04:03,360 --> 00:04:07,170 or aggregation layer inside of its main distribution frame. 106 00:04:07,170 --> 00:04:09,840 In both cases, your network will then branch out 107 00:04:09,840 --> 00:04:11,850 into the intermediate distribution frames 108 00:04:11,850 --> 00:04:14,970 where you're going to find the access or edge layer devices. 109 00:04:14,970 --> 00:04:15,930 Now the second thing we need 110 00:04:15,930 --> 00:04:18,510 to talk about is the collapsed core. 111 00:04:18,510 --> 00:04:20,070 In datacenter topologies, 112 00:04:20,070 --> 00:04:22,110 a collapsed core design is going to refer 113 00:04:22,110 --> 00:04:24,060 to a network architecture where the core 114 00:04:24,060 --> 00:04:26,220 and the distribution layers are being merged 115 00:04:26,220 --> 00:04:27,780 into a single layer. 116 00:04:27,780 --> 00:04:30,180 This basically gives you a two-tiered system, 117 00:04:30,180 --> 00:04:32,370 also known as a collapsed core. 118 00:04:32,370 --> 00:04:34,290 Traditionally, network designs consist 119 00:04:34,290 --> 00:04:36,570 of that three-tiered hierarchy we just talked about 120 00:04:36,570 --> 00:04:39,720 with the core, the distribution and the access layers. 121 00:04:39,720 --> 00:04:42,120 But in a collapsed core design, the functions 122 00:04:42,120 --> 00:04:44,790 of the distribution and core layers are going to be combined, 123 00:04:44,790 --> 00:04:46,860 and this simplifies our architecture. 124 00:04:46,860 --> 00:04:48,330 This is often seen in smaller 125 00:04:48,330 --> 00:04:51,150 or medium-sized datacenters with a scalability offer 126 00:04:51,150 --> 00:04:54,480 by a full three-tiered model is just not required. 127 00:04:54,480 --> 00:04:56,460 The collapse core model is going to reduce the number 128 00:04:56,460 --> 00:04:58,860 of switches that you need, and this can lower your costs 129 00:04:58,860 --> 00:05:00,810 and simplify your management. 130 00:05:00,810 --> 00:05:03,390 The collapsed core model also reduces latency 131 00:05:03,390 --> 00:05:04,680 by decreasing the number of hops 132 00:05:04,680 --> 00:05:07,110 between devices and the core of the network. 133 00:05:07,110 --> 00:05:10,140 However, while a collapsed core design does offer simplicity 134 00:05:10,140 --> 00:05:12,960 and cost savings, it may not be suitable for larger, 135 00:05:12,960 --> 00:05:15,780 more complex networks where the distinction and scalability 136 00:05:15,780 --> 00:05:17,910 provided by separate core and distribution layers 137 00:05:17,910 --> 00:05:20,190 are going to be more beneficial for us. 138 00:05:20,190 --> 00:05:22,380 The choice between a traditional three-layer model 139 00:05:22,380 --> 00:05:23,430 and a collapsed core 140 00:05:23,430 --> 00:05:25,170 is going to depend on the specific needs 141 00:05:25,170 --> 00:05:27,240 and size of your datacenter. 142 00:05:27,240 --> 00:05:30,150 Third, we have the spine and leaf architecture. 143 00:05:30,150 --> 00:05:32,700 Now, the spine and leaf architecture is an alternative type 144 00:05:32,700 --> 00:05:33,900 of network architecture 145 00:05:33,900 --> 00:05:36,570 that's used specifically within our datacenters, 146 00:05:36,570 --> 00:05:38,580 whereas the three-tiered hierarchy we covered 147 00:05:38,580 --> 00:05:40,710 was going to be used to connect the core layer down 148 00:05:40,710 --> 00:05:43,110 to the distribution layer, and then down to the access 149 00:05:43,110 --> 00:05:45,450 or edge layer with all the endpoint devices. 150 00:05:45,450 --> 00:05:48,210 The spine and leaf architecture is simply going to be focused 151 00:05:48,210 --> 00:05:50,790 on the communication within the datacenter itself, 152 00:05:50,790 --> 00:05:53,310 and specifically looking at server farm portions 153 00:05:53,310 --> 00:05:54,753 of that datacenter. 154 00:05:54,753 --> 00:05:56,880 Now, the spine and leaf architecture consists 155 00:05:56,880 --> 00:06:00,180 of two switching layers known as the spine and the leaf. 156 00:06:00,180 --> 00:06:03,060 The leaf layer is going to consist of all the access switches 157 00:06:03,060 --> 00:06:05,580 that are going to aggregate traffic from the different servers 158 00:06:05,580 --> 00:06:06,720 and then connect directly 159 00:06:06,720 --> 00:06:09,300 into that spine layer or network core. 160 00:06:09,300 --> 00:06:11,070 This spine connects switches 161 00:06:11,070 --> 00:06:13,350 that are going to interconnect all the leaf layer switches 162 00:06:13,350 --> 00:06:15,450 into a full mesh topology. 163 00:06:15,450 --> 00:06:16,980 This leads to increased performance 164 00:06:16,980 --> 00:06:19,140 and redundancy for all these servers 165 00:06:19,140 --> 00:06:20,610 that are going to be connected to the leaf layer 166 00:06:20,610 --> 00:06:23,070 and in turn to the spine layer. 167 00:06:23,070 --> 00:06:25,140 By using a spine and leaf architecture, 168 00:06:25,140 --> 00:06:27,750 we can actually get faster speeds and lower latency 169 00:06:27,750 --> 00:06:30,450 than using a traditional three-tiered hierarchy. 170 00:06:30,450 --> 00:06:32,370 By using the spine and leaf architecture, 171 00:06:32,370 --> 00:06:34,320 we're going to be able to take shortcuts in getting data 172 00:06:34,320 --> 00:06:36,870 from one place to another and this happens best 173 00:06:36,870 --> 00:06:38,880 when we're using software-defined networks 174 00:06:38,880 --> 00:06:41,640 in combination with a spine and leaf design. 175 00:06:41,640 --> 00:06:44,130 Now, if you're installing a spine and leaf architecture, 176 00:06:44,130 --> 00:06:46,020 normally you're going to install two switches 177 00:06:46,020 --> 00:06:47,940 in each of your server racks. 178 00:06:47,940 --> 00:06:50,280 This is normally called top of rack switching 179 00:06:50,280 --> 00:06:52,320 because the switches are physically installed 180 00:06:52,320 --> 00:06:54,210 at the very top of your server rack, 181 00:06:54,210 --> 00:06:56,610 and each server in the rack will have a connection to each 182 00:06:56,610 --> 00:06:57,690 of those two switches 183 00:06:57,690 --> 00:07:00,030 that are located inside the server rack. 184 00:07:00,030 --> 00:07:01,890 These switches are essentially the leafs 185 00:07:01,890 --> 00:07:03,660 in our spine and leaf architecture, 186 00:07:03,660 --> 00:07:05,220 and they connect back to the spine, 187 00:07:05,220 --> 00:07:06,420 which serves as our backbone 188 00:07:06,420 --> 00:07:08,640 inside of our datacenter network. 189 00:07:08,640 --> 00:07:11,100 Now, alternatively, you can actually connect this spine 190 00:07:11,100 --> 00:07:13,140 and leaf architecture in combination 191 00:07:13,140 --> 00:07:15,240 with the standard three-tier hierarchy. 192 00:07:15,240 --> 00:07:17,637 Under this model, all the servers in the datacenter 193 00:07:17,637 --> 00:07:19,470 are going to connect to leaf layers, 194 00:07:19,470 --> 00:07:21,780 and the leaf layers will connect to the spine. 195 00:07:21,780 --> 00:07:24,540 That spine then connects directly to the core layer 196 00:07:24,540 --> 00:07:26,100 of your three-tiered model 197 00:07:26,100 --> 00:07:28,770 where all the other non-datacenter devices 198 00:07:28,770 --> 00:07:31,350 will instead connect back to access or edge layers 199 00:07:31,350 --> 00:07:33,630 and then to the distribution or aggregation layer 200 00:07:33,630 --> 00:07:36,180 before going back into the core layer. 201 00:07:36,180 --> 00:07:39,090 Fourth and finally, we need to discuss traffic flows 202 00:07:39,090 --> 00:07:41,010 in relation to our datacenters. 203 00:07:41,010 --> 00:07:43,320 Now, there are two main types of traffic flows. 204 00:07:43,320 --> 00:07:46,380 We call these north-south, and east-west. 205 00:07:46,380 --> 00:07:48,060 Now, these two terms are going to be used 206 00:07:48,060 --> 00:07:50,040 to describe the direction of the traffic flow, 207 00:07:50,040 --> 00:07:54,060 either into or out of a datacenter or across a datacenter. 208 00:07:54,060 --> 00:07:56,700 When we have north-south traffic or communication, 209 00:07:56,700 --> 00:07:58,350 we're referring to traffic that enters 210 00:07:58,350 --> 00:08:01,380 or leaves the datacenter from a system physically residing 211 00:08:01,380 --> 00:08:03,360 outside of that datacenter. 212 00:08:03,360 --> 00:08:05,610 When we specifically talk about north traffic, 213 00:08:05,610 --> 00:08:08,280 this is the traffic that's exiting the datacenter. 214 00:08:08,280 --> 00:08:10,530 Southbound traffic, on the other hand, is referring 215 00:08:10,530 --> 00:08:13,020 to traffic that's entering your datacenter. 216 00:08:13,020 --> 00:08:15,240 In both of these cases, the data will be exiting 217 00:08:15,240 --> 00:08:17,310 or entering the datacenter through a firewall 218 00:08:17,310 --> 00:08:20,730 or other network infrastructure boundary, such as a router. 219 00:08:20,730 --> 00:08:22,560 Conversely, we also have something known 220 00:08:22,560 --> 00:08:24,210 as east-west traffic. 221 00:08:24,210 --> 00:08:26,490 Now, east-west traffic or communication 222 00:08:26,490 --> 00:08:28,080 is instead going to refer to the data flow 223 00:08:28,080 --> 00:08:29,820 within your datacenter. 224 00:08:29,820 --> 00:08:32,940 For example, if we're using a spine and leaf architecture, 225 00:08:32,940 --> 00:08:35,970 any data flow between various servers in the datacenter, 226 00:08:35,970 --> 00:08:37,919 even if it goes between different leaves, 227 00:08:37,919 --> 00:08:40,440 would be considered east-west traffic. 228 00:08:40,440 --> 00:08:41,640 Now, due to the increased use 229 00:08:41,640 --> 00:08:44,100 of software-defined networking, virtualization, 230 00:08:44,100 --> 00:08:46,140 private cloud and converged networks, 231 00:08:46,140 --> 00:08:48,300 more and more traffic is being classified 232 00:08:48,300 --> 00:08:49,770 as east-west traffic 233 00:08:49,770 --> 00:08:52,320 because it's all existing within that datacenter 234 00:08:52,320 --> 00:08:54,090 or that virtual datacenter 235 00:08:54,090 --> 00:08:57,150 if you're using cloud computing like AWS, Azure 236 00:08:57,150 --> 00:08:59,340 or the Google Cloud Platform. 237 00:08:59,340 --> 00:09:02,400 Simply put, if your data is entering the datacenter, 238 00:09:02,400 --> 00:09:04,470 it's considered southbound traffic. 239 00:09:04,470 --> 00:09:05,970 If it's leaving the datacenter, 240 00:09:05,970 --> 00:09:08,310 we consider this to be northbound traffic. 241 00:09:08,310 --> 00:09:10,140 If it's moving within your datacenter, 242 00:09:10,140 --> 00:09:12,810 we consider this to be east-west traffic. 243 00:09:12,810 --> 00:09:14,760 So remember, there are many different ways 244 00:09:14,760 --> 00:09:16,110 to architect your datacenters 245 00:09:16,110 --> 00:09:19,290 and networks using either a three-tiered traditional model, 246 00:09:19,290 --> 00:09:22,410 a collapsed core model, or a spine and leaf architecture. 247 00:09:22,410 --> 00:09:24,840 And in any of these cases, you still need to understand 248 00:09:24,840 --> 00:09:26,700 how the traffic flows in, out 249 00:09:26,700 --> 00:09:28,770 or throughout your datacenter. 250 00:09:28,770 --> 00:09:30,780 When you're choosing a datacenter topology, 251 00:09:30,780 --> 00:09:32,970 which one you decide to use will really depend 252 00:09:32,970 --> 00:09:35,550 on your business case and your organization's needs, 253 00:09:35,550 --> 00:09:37,550 so make sure you consider those as well.