1 00:00:01,740 --> 00:00:07,340 What do you need to know about cluster networking for exam AZ‑801? 2 00:00:07,340 --> 00:00:10,530 Well, for this part of the exam, if you're asked about it at all, 3 00:00:10,530 --> 00:00:14,530 it's just what are some inbox Windows Server features that 4 00:00:14,530 --> 00:00:17,640 can make your network run more efficiently? 5 00:00:17,640 --> 00:00:20,150 Because, as I mentioned in the previous module, 6 00:00:20,150 --> 00:00:22,810 high availability is about redundancy, isn't it? 7 00:00:22,810 --> 00:00:28,170 So, ideally, your cluster nodes have redundant hardware network interface cards, 8 00:00:28,170 --> 00:00:29,540 or NICs. 9 00:00:29,540 --> 00:00:34,230 If those NICs are fairly current, they should support RDMA, 10 00:00:34,230 --> 00:00:36,090 or Remote Direct Memory Access. 11 00:00:36,090 --> 00:00:40,740 That would allow those NICs to access remote storage almost as 12 00:00:40,740 --> 00:00:43,300 fast as the server can get to its own storage, 13 00:00:43,300 --> 00:00:43,970 you see? 14 00:00:43,970 --> 00:00:45,380 Now, 15 00:00:45,380 --> 00:00:50,590 we can combine RDMA compatible network interfaces with 16 00:00:50,590 --> 00:00:52,280 the Server Message Plock protocol. 17 00:00:52,280 --> 00:00:57,740 The SMB 3 protocol and SMB Direct allow you to take 18 00:00:57,740 --> 00:01:00,670 advantage of that RDMA capability of your NICs. 19 00:01:00,670 --> 00:01:04,850 And again, in the name of file share in particular, 20 00:01:04,850 --> 00:01:08,640 accessing file data on shares using the SMB protocol, 21 00:01:08,640 --> 00:01:14,440 can be very fast and efficient in terms of internode communications. 22 00:01:14,440 --> 00:01:17,940 Lastly, in Windows Server we have NIC teaming, 23 00:01:17,940 --> 00:01:22,550 which is where you can create a logical network interface that 24 00:01:22,550 --> 00:01:26,540 actually maps to two or more physical interfaces. 25 00:01:26,540 --> 00:01:29,470 This allows you to do bandwidth bonding, 26 00:01:29,470 --> 00:01:31,560 which gives you increased throughput, 27 00:01:31,560 --> 00:01:36,160 but also gives you high availability at the network interface level because 28 00:01:36,160 --> 00:01:39,630 if you lose a NIC due to a planned or unplanned outage, 29 00:01:39,630 --> 00:01:41,880 as long as you have at least one other network 30 00:01:41,880 --> 00:01:44,340 interface in the team that's available, 31 00:01:44,340 --> 00:01:46,300 then that server can continue to communicate, 32 00:01:46,300 --> 00:01:47,640 all right? 33 00:01:47,640 --> 00:01:50,770 So we could go beyond this, we really don't have to though, 34 00:01:50,770 --> 00:01:53,490 but we want to think about things like multi‑path 35 00:01:53,490 --> 00:01:55,620 I/O and redundant connectivities. 36 00:01:55,620 --> 00:01:58,260 You might have multiple network paths. 37 00:01:58,260 --> 00:02:00,580 You might want to separate, for example, 38 00:02:00,580 --> 00:02:05,040 your storage, network traffic, your cluster internal traffic, 39 00:02:05,040 --> 00:02:08,700 that is your heartbeat traffic, and then your data traffic. 40 00:02:08,700 --> 00:02:09,620 If you can do that, 41 00:02:09,620 --> 00:02:12,340 if you have the network infrastructure and the 42 00:02:12,340 --> 00:02:14,570 hardware to support that topology, 43 00:02:14,570 --> 00:02:17,570 all of those things are good because they increase 44 00:02:17,570 --> 00:02:21,950 the resiliency of your cluster, they increase the performance of your cluster, 45 00:02:21,950 --> 00:02:25,940 and give you as much high availability as you can get. 46 00:02:25,940 --> 00:02:28,140 Cluster Workload Options. 47 00:02:28,140 --> 00:02:32,160 All of this setup is well and good, but ultimately it comes down to well, 48 00:02:32,160 --> 00:02:36,130 great, now that we have the failover cluster up and running and it's functional, 49 00:02:36,130 --> 00:02:40,430 what exactly can we do as far as making applications 50 00:02:40,430 --> 00:02:41,810 and services highly available? 51 00:02:41,810 --> 00:02:43,220 What is supported? 52 00:02:43,220 --> 00:02:44,760 Well, no surprise, 53 00:02:44,760 --> 00:02:49,090 many of the most common Windows Server roles are supported 54 00:02:49,090 --> 00:02:51,190 out of box for Windows Server clustering. 55 00:02:51,190 --> 00:02:55,270 The distributed file system, or DFS, 56 00:02:55,270 --> 00:02:59,740 you can make a DFS namespace highly available. 57 00:02:59,740 --> 00:03:00,500 This, again, 58 00:03:00,500 --> 00:03:04,300 it brings up the idea of redundancy because if you know much about DFS, 59 00:03:04,300 --> 00:03:08,740 we cover DFS in great depth by the way in AZ‑800, 60 00:03:08,740 --> 00:03:12,980 DFS allows you to aggregate file shares from across your environment and 61 00:03:12,980 --> 00:03:16,670 create a single logical tree view of those shares, 62 00:03:16,670 --> 00:03:19,950 and you can replicate those shares, you can replicate your namespace, 63 00:03:19,950 --> 00:03:25,670 so it's nice that as an option you can put your namespace in a failover 64 00:03:25,670 --> 00:03:29,640 cluster to give yet more high availability to it. 65 00:03:29,640 --> 00:03:33,050 File server is another built‑in clustered role. 66 00:03:33,050 --> 00:03:37,070 This would be an active‑passive file share, and then there's the 67 00:03:37,070 --> 00:03:42,140 Scale‑Out File Server, or SOFS, option as well for virtual machines and 68 00:03:42,140 --> 00:03:48,200 for SQL Server databases. You can take Hyper‑V VMs themselves and make 69 00:03:48,200 --> 00:03:51,040 them highly available as clustered roles. 70 00:03:51,040 --> 00:03:54,700 You're using Hyper‑V Replica to replicate virtual machines and 71 00:03:54,700 --> 00:03:58,180 you want to replicate VMs that are in a cluster, then you'll 72 00:03:58,180 --> 00:04:00,130 want to have the Replica Broker. 73 00:04:00,130 --> 00:04:04,090 In fact you'll need to have an instance of the Replica Broker as 74 00:04:04,090 --> 00:04:08,360 a clustered role. Dynamic Host Configuration Protocol can be 75 00:04:08,360 --> 00:04:12,460 critical for a business whose endpoints received their TCP/IP 76 00:04:12,460 --> 00:04:14,840 configurations dynamically. 77 00:04:14,840 --> 00:04:15,470 Again, 78 00:04:15,470 --> 00:04:19,810 this is one way. I mean Microsoft's DHCP server does have its own 79 00:04:19,810 --> 00:04:23,980 failover capability, so I don't want to lead you down the path to 80 00:04:23,980 --> 00:04:27,260 think that there's only one solution to the problem of high 81 00:04:27,260 --> 00:04:29,840 availability. There really never is. 82 00:04:29,840 --> 00:04:33,550 Now you'll see that in Windows Server failover clustering there's a 83 00:04:33,550 --> 00:04:38,440 generic catch‑all for generic app, script, or service. 84 00:04:38,440 --> 00:04:43,360 So this is to remind us that we're not limited only to the clustered roles 85 00:04:43,360 --> 00:04:48,690 that Microsoft gives us that it hosts as you see on the slide. You can always 86 00:04:48,690 --> 00:04:59,000 make your own application service or script highly available by using the generic template in Failover Cluster Manager.