Introduction to SAN and NAS Storage | Neil Anderson | Skillshare
Play Speed
  • 0.5x
  • 1x (Normal)
  • 1.25x
  • 1.5x
  • 2x
13 Lessons (1h 30m)
    • 1. Introduction

    • 2. How to Make the Most of this Course

    • 3. Benefits of Centralized Storage

    • 4. NAS - Network Attached Storage

    • 5. CIFS (SMB) Configuration

    • 6. NFS Configuration

    • 7. SAN - Fibre Channel Part 1

    • 8. Fibre Channel Part 2

    • 9. Fibre Channel Part 3

    • 10. Fibre Channel and FCoE Configuration

    • 11. SAN - iSCSI

    • 12. iSCSI Configuration

    • 13. SAN - FCoE Fibre Channel over Ethernet


About This Class


Today's data center and cloud environments have converged infrastructures and management tools, and with the advent of virtualization technologies like VMware and Microsoft Hyper-V, external SAN and NAS storage is becoming standard. It's no longer possible to get by as an IT professional by knowing only servers, networking or storage, you need an understanding of all three for your career to thrive.

This course will give you the knowledge you need to understand SAN and NAS storage. You'll learn the theory of the CIFS (SMB), NFS, Fibre Channel, iSCSI and FCoE protocols, and see how to configure the storage system and connect from the client with hands-on lab demonstrations.

The lessons are vendor agnostic so they're suitable no matter who the manufacturer of your storage system is. I've used NetApp as the storage system in the live demos but the information is relevant to all vendors.

Thank you for taking the time to read about the course. I hope you now have enough motivation to get into the learning right away. Click the blue button to enroll and get started!

What are the requirements?

  • A basic knowledge of computer networks
  • No previous storage knowledge is necessary

What will you be able to do after taking this course?

  • Understand the basics plus more of Storage Area Networks and Network Attached Storage
  • Understand the difference between SAN and NAS
  • Understand the theory of the NAS protocols - CIFS (SMB) and NFS
  • Understand the theory of the SAN protocols - Fibre Channel, iSCSI and FCoE
  • Perform a basic configuration of SAN and NAS storage on a Storage System
  • Connect to CIFS (SMB) NAS storage from a Windows client
  • Connect to NFS NAS storage from a Linux client
  • Connect to iSCSI SAN storage from a Windows client


1. Introduction: in today's DIA Center on Cloud Environments. It's no longer possible to get by as an I t. Professional, knowing only servers, networking or storage. You need an understanding of all three for your career to really thrive. Welcome to the introduction to sign and not storage course where you will learn the basics , plus more off storage area networks on network attached storage into these of your world environments. My name is New Anderson a name our professional cloud on data center technology strainer, who's being entrusted by net up for the last name years to train their storage engineers on the engineers of their partners, as well as performing knowledge transfer in the classroom on online. I also did Ellen new training courses for now up on other companies as well, like Sir School, IBM on Verizon. In this course, you'll learn the CD of how the SIFs SMB NFS fibre channel ice cozy on FC Weepu echoes work on. She hoped to configure distorted system on connect from the client with Hyung's on wild demonstrations. I'm not going to just teach you the concept I'm gonna show you with real world Demo is how things are done the city in the lash ins is vendor agnostic, super suitable Neumar who the manufacturer of your storage is. I've used nap az restored system in the life dam was. But the information is relevant to all vendors. Thank you for taking Italian. Teoh, watch this video. I really want to give you this knowledge to help your career. So if you're ready to get started, please click the green bun to your right on Dwight School. 2. How to Make the Most of this Course: high on Welcome to the cars. Thanks very much for enrolling. I really do appreciate having you here in this video. I'm going to assure you how to get the most of the cars on that involves going to my Web shaped much. Do that now. So here I am on my blawg, which is at www dot flat box dot com, and you can see that I'm posting regularly on here with new updates of being focusing on these introduction to sign and nice of it yours. Andi Netapp started surefire the in future. I'm going to be covering the wider range off clothed and data center technologies. So I'll also have articles about school on VM ware. Except I've been doing to tour the old so far. But I'm also planning and doing some use our records a swell. So if you go to the tutorials menu up at the top here on Click on Sun and Nice Storage B six, you'll see that I've uploaded all the videos for Vicks course. I've also included tax descriptions for all with those videos as well, so I picked one at random. See, I'll go to ice cozy sign overview here on when this opens up, You see that? I've got a link to the video on YouTube on underneath. I've also got a full tax description of the article is Well so if after you've watched the videos, if you want to get out of fresher on a particular topic and you don't want to watch the whole video again, you can come here on You can see that the tax Darko. Instead, it covers all with the same information. Other things I have in here, you can click on the free cars in the top, right? To get a different view off the course on something that you're definitely going to want to do is to connect with me on social media using the bonds over on the right here. So if you click on YouTube that will open up my YouTube channel, you can subscribe here, and then you can get a notification of every time I upward and you tutorial another one is my LinkedIn page. So if we have a week in here and if I go under my network on connections, you can see that I've got newly 2000 connections and linked in under all highly targeted for all working in the cloud and daycare center industry on the benefit you'll get from this is when you connect with me. All of my connections will become your second level connections, which makes it easier for you to build your client and day ascent or network when you connect with me on LinkedIn. If they ask you from a email address, it's Neil are flat box dot com. Looking back at the website again, I've also got links year for my Twitter and Facebook accounts. I haven't been too active on there in the past, but that's something I'm going to fix this year. Under is also a link for my RSS feed. If you'd like to subscribe to my IRA course that way as well. Okay, So much everything I wanted to tell you. There have a look at my website. You can see tax descriptions for all the video is there. You can also connect with me on social media. Thanks 3. Benefits of Centralized Storage: in the later lessons will be getting into the how of storage area networks on network attached storage. We'll cover how the different protocols work on how to configure. I were started systems, but first I want to give you the why why we have from bite. We would use sun and non politicos in the first place. The first benefit I want to tell you about is for centralized management and these benefits on mean about competing sign and knives to when we're using dust director touch storage director Touch storage is where we have storage, which is dedicated to one particular server, so that could be a local hard drives and saying that server or it could be an external radar in okay. Our first benefit off external centralized shortage, which is sign and knives, is centralized management. If we've got 50 servers, it's much easier to manage the storage for them all. If thoughts on one centralized system Robert and distributed individually to each of those 50 servers, the next benefit we have on which most people would probably say is the main benefit upset toe I storage. Is this utilization? If you're using traditional director touch storage. You'll maybe get this utilization off 2030 40%. When we used centralized storage, we can get a figure of it's going to be closer to 80%. The reason is what? See, we had booze 50 servers again. And let's see some of those servers did acquire 300 gigabytes off storage Speaks Well. If you're using director touch storage, you're not gonna put in exactly 300 gigabytes of space in there. You're going to put in probably 500 gigabytes of space in there because, well, first off, you can't buy 300 gigabyte disks. And secondly, you want to leave a bit of room for unexpected growth is, well, you don't want Teoh change a hard drive and a server because it's really inconvenient, and it's going to require a knife edge. You would have to power of a server off tick Iraq, open up the Shashi and put more disk space in there. I don't want to be doing that with centralized storage. We don't have to go through all of those problems. We have one centralized pool of our storage on weaken, slice it up and distribute it to the different servers exactly how we want to. So if if I've got a couple of servers that require fee 100 gigabytes of disk space, I give them 300 gigabytes of disk space. If only our on it turns out that I was being too conservative there, and we need more disk space than I can easily change the night of this space. Affordable servers and typically Aiken do this non destructively as well. It's not going to require a night John the server. And for sure it's not going to require me a pirate diving antique out of Iraq. Another way we can benefit. Disc utilization is through. Thin provisioning went. See that? I've got 10 servers, and I think they may need up to 500 gigabytes worth off storage species. So 10 servers, times 500 gig abates by is five Ted abates worth off speech in a traditional environment, I would need to buy five terabytes worth of disk space right now, but when I'm using centralized storage, I can implement thin provisioning with thin provisioning. Let's see that the servers are only going to require actually only going to be using 200 gig abates right now, to 10 times, 200 gigabytes. Aiken Just by two tatter Bates worth off physical this space. But I make it look to the servers Likely they actually have the fuel 500 gig abates. So it looks like they've got faith. Tear debates, refer disputes tool. I've only actually appeared for to tear debates if we do the choir more especially, or than I just go in baby disks at that time. So I'm moving from just in case model when I'm purchasing my storage to a just in time, I get good cost benefits from that. Another way that I can improve on my desk Utilization is with D duplication on compression with D duplication. I've got multiple servers, all using that same centralized storage. If there's any blocks on disk there. Besim. I couldn't move voice duplicate blocks, so this cuts down on the amount of storage space I'm using again Cost savings Aunt. Similarly, I can also use compression on the discs as well to get the same benefit. Next thing is disk list servers. If I'm using its on Pruitt call, I can have my server's boot up from discs on that remote storage. They don't even have to have a single disk drive in the server themselves. This gives me cost savings on the hardware folks, who my servers are going to require. Last rock space, West Power and last cooling. I can implement stories tearing on my story system. Went See, I've got discs which have got different levels of performance. I've got high performance SS these on. I've got high capacity but lower performance CIA drives. I can keep my hot data. The other is accessed frequently, which is probably new on the SS D's, and I can keep my older archive data on the drives. Most centralized storage systems also have features that can automatically move the data from the high performance storage to you. The whole performance storage as it's being actress last frequently resiliency. Our centralized storage systems are always built toe have very high degrees of resiliency because pretty much always we're going to be mission critical systems. So if a disc feel was much taken cared off by red, if a disc shell feels I'm mirroring my this shelves if he controller fields, I've got multiple control MERS in my storage system so it will feel over to another controller, and it can replicate the debtor between store insistence and different sites. So if I have a huge disaster on the whole site blows up. I can feel over to the other geographical location, centralized back up back to booze. 50 servers again. If I got 50 servers and they've all got their one will call Tip Drive, then managing by cops is going to be very inconvenient. Maintain consuming it for all Using sheared centralized storage. I can centralize may back up solution as well. It's going to be much easier to manage on my back. Ups can be run more quickly. Snapshots, snapshots are any short term backup. Snap shorts are super fastly happened pretty much immediately. What snapshot is It's a point in time copy off the failed system so I can take snapshots. It's a copy of the failed system. They happen pretty much immediately, and then if India gets corrupted, or if somebody accidentally deletes a fail, then I can very quickly and easily recover from the snapshots. No snapshots of the don't replace your traditional backups, because received in this implication is the storage. But they're great for short term quick and convenient backups and restores disaster recovery. I mentioned earlier we could have a storage system in our main site on we could replicate the data to a backup system in our disaster recovery site. Very easy to do that if you lose with me and say you feel over to the disaster to cover the same. The last huge benefit that I want to talk about is virtual ization support for software such a jvm wear on Microsoft hyper V with that arm order, Um, Dia centers Now pretty much always. We're going to be using virtual ization software. This means that Robert and having dedicated hardware for exchange server dedicated Hardware foot are sequel server except dropped. We can draw on multiple virtual servers on Massimo underlying physical server box. So Ah, but laproscopic here from three m where once we get this over here on the left up goal a physical server on went see, I'm running virtual machines on their off an exchange server, A sequel database server on a Web server. That's the pictures you see up here with the M Were the underlying physical hardware like the box on the left in the box and the right here that's used for this CPU on for the manatee. For our virtual machines, where the data is stored is stored on external storage, like a sign or a nice system. That's what you can see dying here on what this enables is the motion, which is the killer feature off the M, where with the motion. If I want to take the serve it on the left here, dying for maintenance that I can be motion all the virtual machines from it over to the server on the right. So my exchange server, my sequel server, my Web server, which is all running his virtual machines. I can move them on the fly from one physical server to be different. Physical Server and his new wife, Bridget. All voice service clients don't even know of it. The server was moved, so this is great if I want to take a physical box thing for meat, and also the same thing will happen if I physical box field. So if the box blows up, all my virtual machines can non destructively move over to another physical underlying server box. This is only really possible. If you're using external storage, the M where supports all of the sun through Icos for based on this support NFS as well. 4. NAS - Network Attached Storage: in this lesson, we're going to do a compaction off the night on the sun protocols. I know folks assure you what Nagy looks like from the client's point of view. Nice stones for network attached storage. And it pervades fail level access to the storage system but will meet more sense of it. What fail level access is when, I assure you this in the lab from the client's point of view with knives of the management of the failed system, resides with the storage system so the client connects to the started system. The started is already there and configured on the failed system, is mining by the storage system you're going to see that is different on San Protocols. In a second, Naz typically uses the existing data network infrastructure, so the existing network cards, cables and switches that would already there in your local area network and being used for data will also be used for the storage as well. Son Stein's first order JD, a network. It pervades block level access to the storage resources soon as his failed level access sign is block level access with sign. The management of the failed system resides with declined. So the client will connect to its storage, which is on the storage system, and it gets a more block level Access to that storage The client Will CSI it shoeing up? Just wake us if you'd I did a new hard drive in the client and the client will have to form Are the storage with the fireworks. Kristen before it can use it. Sign typically has a dedicated started network infrastructure, so we don't use the existing network cards. Cables on switches that we have for our normal data network typically will sign. We will put in a dedicated network just to be used for storage. Many modern storage system support boob knives on Sign Pruitt Coast back in the old days. Typically, your started system would support either sign or knives, but not both at the same time. Some started systems still work like that, but I want in your start assistance to support all of the knives on all with design protocols as well. On the same system, the feature is only available in Sign is boot from John signed pervades block level Axis, which appears we seem to decline, is on local in Stone drive. It's possible to boot from Sun where boot from sign is deployed. It's possible toe have disc list. Servers should see we've got a server on. We've configured storage for it, which is sound storage on remote storage system. The server can actually boot up from the remote storage, and we don't even need to a local hard drive in not server, we can get really good cost savings from doing this boot from Sign is Only Supported on Sign That's why it's called. But from sign it's not possible, Overnights went summarized. The different storage protocols I were Naz protocols are sips the common Internet Fail system on NFS, which is the network fail system. Our son crew echoes are FC Fibre Channel I scuzzy the Internet Smoke Computers interface on EFS Yui, which is fibre channel over Ethernet. We're going to go into those protocols in a bit more detail. Strips is the common Internet file system. It's a version off ash and be the server message block the two terms ships on SMB. There are slight differences, but the terms are often used interchangeably. Sips was developed may maker soft, and it's used for windows based clients Samba alos or UNIX based clients to access if shares so it was designed for Windows clients. But UNIX and Winnicks and other clients kind of also access it shares with strips. The servers share storage and the clients use or mop with that share. Let's have a look at how shifts network attached storage works with a wide dental. I'm on my Windows client here. I've already configured storage on my storage system to connect it. I opened up Windows exploder, and then I'm going to right click on this PC and choose the option to mopey network drive. I'm gonna use drives ad for this example. So after I have mapped a network drive, it will show up with a Dr Layer. On the path to this year is backslash, backslash and in the I P address of my storage system, which was 17 to dock. $23.3 dollars. 31 with the story. So I just configured on a box slash on Bashir name is financed. I enter my user name and password andan connected to the finite share, and it's showing up as my zed drive on and conceive it fails and folders that are in that share. I didn't need to format the failed system. I've got immediate fail level access to the storage much take a look at and their fashion. NFS is the network file system. It was developed by Sun Microsystems originally for UNIX based clients. So it's a good fit from Lennix as well. Just like other operating interestings can access if shares other clients like windows on VM ware. Yes, X. I can also access NFS exports with sifs The servers share on the clients use or my share with NFS servers Export on the client's meant the export. So an export in NFS is equivalent to a share in sips. We can also have a look at how NFS works from the client point of view in the lab. So I'm here on at my Lennix client on. I'm gonna open up a terminal window on here. First off, I'm going to meet a directory that I'm going to mount the storage into so I say Mick directly or put this in the main directory on this is going to be used by seals, put my password in and then to make and use the storage. The command ISMM ain't storage system is named storage to Indian s on the export name. Is Shiels in this example on? I'm gonna make that to my mount Slashed seals folder. I need to run Mrs Route. So let's try that again with putting shoot all at the front of it. I'm very go. You can see that. I know. Managed to make the storage. I'm gonna change directory to moments Weiss Shields. And if they do an ls to West, the content I can see that I've got directly is in there and again, just like the six demo I've got fail level access to the storage. I didn't need to format it with a phalanx. Kristen Bachelor taking kid off on restoring system shade. Looking at the network topology for our Nagy politicos. Two bulbs shifts on an effect. You can achieve that. The client access is its storage over the seam McLeary a network infrastructure. But it's also using to access its other servers and services on the Internet. We don't need to put in a separate network just for storage 5. CIFS (SMB) Configuration: Let's take a look at how to configure shift on our storage system here. I'm on a netapp system, but the configuration stamps are going to be similar. No matter which vendors started system, you're configuring net up. The storage is implemented in storage virtual machines and each storage virtual machine will look like a separate started system to the clients. I'm gonna create a dedicated SBM for myself storage. So I'm in the storage virtual machine page and click on Create. I'll use the SPM name off shifts for this. Damn you. The D A protocol I'm gonna use here is chefs. It's for Windows clients. So I went to use the NT F executed Ishmael. The aggregate is reset with hard drives in my storage system that is going to be used for fish storage on. I have my DNS configuration pointing at my windows the mean control mur that is also being used by my clients on a click on submit and continue. On the next page, I configure my ship sayings I give the storage virtual machine and I p A drinks. In my job, I'm using $172.23.30 dollars sir It one shop Nick Mask to 55255255 Doll on a get way of one ship in to dock. $23. 31 dot to 54 Click or kit to that? No, I put a tape when the Gateway, the Gateway and I P address obviously need to be in the same sub net through or I'll fix lap and click on OK, deport is the network card on the started system that the clients are going to reach their storage over and my wife. That's easy. Roedean. Click on OK to that. When I set this up, my storage system is going to show up as a server in the active directory, demean in the shift server Name field Here I say the name. I want it to be new in eyes in the demand How user seemed and I used for the spn which waas shifts. Then I enter my active directory demean, which is flat box. Don't lump for my example on the administrator name on password of a user that is going to be allowed to join this computer to the demean. Then I provisions a volume which is what the clients used to access for storage. I'll use a sheer name off. Damn one. Oh, me at small size cause we're just in a live environment year. So 50 maga. Bates Real What old? Typically, this would be a lot bigger, and I consent. What? The permissions are going to be a new volume as well. He and I'm just gonna leave the everyone full control a real world environment. You might well wanna walk it down more than not. And I click on submit and continue. I can skip the next page about SBM level administration and in a c make somebody ph Andi, I'm done. But over was to it. I've no configured slip storage for my Windows clients. Other operating systems will be allowed to use this story here as well. If I have enabled that if I have a look at my windows demean controller on open are active directory users and computers and go to the computers or you, you can see there's my sip server. That is the storage system, which I just joined to the domain 6. NFS Configuration: to configure NFS on a netapp storage system. It's similar to how I did. Six. I quick on the create button in storage virtual machines to create a new storage virtual machine. I'll name this one NF ash on enable the NFS Protocol on here. Persecuted. Ishmael is now UNIX because NFS is typically used by UNIX on when textile clients it's also commonly used baby m where I'm happy with the aggregate I'm using I two get one match a set of disks that are going to be used here I click on submit and continue after checking the DNS configuration is correct. I then get onto the configure NFS Pruitt Coal page. I configure the I P address that clients are gonna connect it on. I'll give this I p address. 17 to dock. $23.3 dollars. 32 Subject masked $255.255255 or on the gateway is $172.23.3 dollars to 54 and click on OK and then specify the port that the clients are gonna connecting on its e zero d for that example. Okay to that then have a section to configure an eye ash to authenticate the users that are going to be accessing this storage. I can either use our local user database configured on my started Tristen, or they can use an external and I s server or an external held up server Such a Z active directory were just using local users here so I can leave this section blank. I'm going to configure an export for my storage and export is the NFS equivalent of a share in windows. What's name our exports demo to ah Nomiki small size again, just 15 aggravates again For your world, it would typically be a lot bigger than next on. Specify the clients that I don't like to connect into the storage 0.0 dot 0.0 slash tzeitel means it will accept any clients on any I p address. I click on submit and continue. I don't need SPM level administrators so I can skip this page, then out. See the so many page I can click on OK, here and that's it. All done. My NFS clients can no main to their storage 7. SAN - Fibre Channel Part 1: in this Washington we're going to start covering I our son protocols on the best one to start with is the original Shine Protocol, which boys fibre channel. Before we start getting more in depth foreign fibre channel for let me give you some general shine terminology. First we have a on 11 is a logical unit number that represents a disk that will be presented to a host through the Klein connects to its along. It sees the one as if it was a local hard drive. Ones are specific to our son, not our knives protocols. Other terminology the client is new in as the initiator on the storage system is new in as the target. Favorite channel is the original shine political, and it's still ready, popular today that uses dedicated adopters cables on switches. And it's different than Ethernet at on layers of the U. S. I stock, including the physical level so you can't use an Ethernet network card or an Ethernet switch for fibre channel. It uses dedicated favorite channel infrastructure. This is different than Arnaz Prue Icos, which run over Ethernet. Theo. The FTP Fibre channel protocol is used to saying scuzzy commands over the fibre Channel network. So if your client hide a local hard drive, it would send Scotty commands to about local hard drive with son. It's sending scuzzy commands for seeing commands, but over a network now favorite channel is a very stable, Andhra Viable protocol, so it's very popular with old school storage engineers. Favorite channel is waterless, unlike TCP Andi UDP with TCP the Sander Sands traffic to the receiver, the receiver will pd optically sand acknowledgements back. If Ascender doesn't get an acknowledgement back then, it knows that the traffic was lost in transit and it will be sent the traffic with UDP its best efforts. We rely on the upper layer protocols to be sent traffic if anything gets lost. So that's how traffic works over Ethernet networks is different in favour channel. It's built into the protocol that we never lose any traffic that we're gonna have to recover from. Favorite Channel currently supports bind which of 24689 16 gigabits per second, depending on the hardware that you've got deployed. Favorite channel uses a dedicated network for this storage. So you see here we've got our standard a local idiot network running over Internet, which gives clients access to their servers. In the example here, the server is going to be using favorite China storage. For that, we have separate dedicated network infrastructure through putting them both together. Went See that? The server that you see in the middle here is a Web server on the client. Up at the top is going to be, ah, kissing a webpage or not Web server. The client hable access for server over the normal Ethernet local area network. Then, to fetch the actual Web page from its storage, the server will connect to its storage over the Faber Channel network. So the network you see down here on the server it has got HB A's, which our host bus adopters about to the equivalent of a network card in Ethernet not will connect to dedicated favorite channel switches. And then our started system also has HB is to connect to our favorite channel network. Favorite channel uses worldwide names Don't be w ends for its addressing the W W. Ends are it by addresses that are made up of 16 hacks of national characters on. You can see an example off their format here. It's a big, long hex of national address. The world weighed in ord, named the W W N N, is the same too annoyed in this storage network and the same W W nn can identify multiple network interfaces off a single network. Note. So we've got a horse. It's got maybe multiple HB A's or multiple parts in the HB A's the W W N and signifies that one Horst W W N N stands for world wide. No would name you may. Sometimes she even also being referenced as N. W. W. And the new would worldwide name. That is exactly the same thing. So don't be W nn or an www and they're the same thing, just two ways of seeing it. Our hosts also have worldwide port names. W W P ends a different W W p and, as we're saying to every individual port on annoyed. So if we had a multi part H B a in machine host, each port on that HB A would have a different w w p and w w p ends are the equivalent of market rashes and Ethernet. The W W. P. N is burned in by the manufacture all that HB A, and it's guaranteed to be globally unique, just like a market ration. INTERNET W W P ends are saying to HBs on both the clients on the storage system as well. So the ports on the client on reports on the storage system they'll all have unique w W p ends were primarily concerned with w W P ends. No W w n ends when we're configuring favorite channel networks and just like W W N ends can also be known as N W W ends. W W P ends are oaks or sometimes knowing as P W. W. And again it means the same thing. Aliases can be configured to make configuration on troubleshooting easier on your storage. For example, we could create an alias named Exchange Server for the Exchange server, which has got that big long W W p end that you see there by configuring an alias. It means that when we are configuring settings for about server, we can specify the alias Roberton, VW WP and Vicious is more convenient and also makes it less likely that we're going to in any typos for favorite channel. You're going to need to configure the favorite channel switches on distorted Justin. Aliases can be used on both of them. 8. Fibre Channel Part 2: for security. A zoning is configured on the switches to control which horse I don't like to communicate with each other. The servers, which are clients of this started system, will be able to communicate with the started Justin, but servers will not be allowed to communicate with each other over Fibre channel. This increases security and also mix the favorite channel network more reliable and stable . I've got an example of how to configure joining Hear the commands in tax in this example is for Easter School favorite channel switch. Brockett are another popular manufacturer of favorite channel switches. In the example. Here go a couple off servers at the ball, which are clients off the storage system. Each of the servers have got an H b A with a W W p n on have configured an alias with the servers name for about particular W W p n the started system up the tall also highs. It's HB, and it's W W P N on. We've configured an alias for it as well. With the zoning, we've got zoo name, Server one So we can figure is June, which is dedicated for Server one on. We see member FC a liest server one on member FC, alias Netapp Controller one Visualize Server one to talk to its storage. We also configure a zoo in for server, too. So I've got zoo in name server, too. Remember FC Alia Server to a member FC alias Netapp Controller one that joined Allies Server to to talk to its storage so I can figure is June for each of my different servers, which are clients office storage system. Then I group all of those June's together into a zoo in set. So I've got June sent Name my zoo in, sent on actually a member server one on a member server, too. The zoning configuration is done on the favorite channel switches, and with this configuration, server one can talk to its storage and server to conduct to its storage that the two servers can't talk to each other over the favorite channel network, as well as configuring zoning on our switches. We also need to configure one masking on distorted system. It's critical that the right one is presented to the right host. If the wrong horse was able to connect to online than it would be, why a bowl to corrupt it. So if you look back example here, I don't want server to to be able to connect to server ones and one viz owning on the switches. Make sure that the servers can't talk to each other that it can both talk to the storage. So how did I make sure that they can't connect to each other's lungs? That's where one masking comes in. Joining prevents unauthorized hosts from reaching the storage system, and it prevents horse from talking to each other over the favorite channel network. But it doesn't prevent Ah Horst from accessing the wrong line once it gets to the storage system. One masking is configured on this storage system toe walk in London to the host or hosts who are authorized to access it. To secure your storage, you need to configure, is ruining on your switches on one masking on your storage system. Here's an example of how we would configure one masking, so I've got my storage Justin up top and I've configured a boot one for Server one on Evokes who configured a boot one for server to for might server one but one. I see that the only initiator that can connect to that. It's server ones W w p n and I say that the server to boot one. The only initiator that can connect to that. A server Tuesday. Lbw PN ac you Dokes who have used aliases here, Robert and typing in the W W P n in my configuration. Next thing to talk about is switched. Damien I ds. Each switch in the fibre Channel network will be assigned a unique demean i d. I think the name of these quite confusing because I think of a demean i d. I think that would mean an I d for the entire diminutive switches, but it doesn't mean not at all the demean I d means a unique i d. For each switch that is in that favorite channel network One switch in the network will be automatically assigned as the principal switch. It is in charge of assigning Damian ideas to the other, switches each switch and learns about the other switches in the network and how to wrote to them based on their Damien i. D. When our servers or started systems HB a powers on it will send a flunky which is a Phablet Morgan request to its and locally attached favorite channel switch. The fibre channel switches it's plugged into. The switch will then assign it a 24 bit FC i D, which is the favorite channel idea dress. So what's happening is the server is powered on the each Be a comes on lane. It sends a flaw g to the favorite channel switch. It was plugged into seeing he This is my w w p n Please air saying me an f c i d so that I can communicate on the Fibre Channel network, the F c I. D. Are saying two hosts is made up of switches. Demand I D that it's plugged into onto the switch port. The F. C I. D. Is similar to an I P address in Ethernet. It's used by fibre channel switches to road traffic between servers on their storage switches. Maintain a table of FC I D to W. W p n address map ings on what port The course is. Well, kid on so every switch in your favorite channel network it learns about the Damien I ds off all with the over switches in the network It also learns about the W W P ends of all the horse to the attached to that network and based in the f C i. D. It knows the demand i d. Of a switch that that Horst is plugged into because we have all this information very able to write the traffic between the hosts. Here's a picture off the fabric longan process working. So I've got Server one tone at the Bolton, which is a client off my storage system up at the top when Server one Powers on its HB a book, send a fabric Morgan a flaw G to the switch that is attached to, and the switch will assign it an f c i d. If I know did a sure foggy database or not switch, I would see the interface that the servers plugged into, and it's F C I. D. On it still be WP and the same thing happens with our storage system up at the top when it powers on its HB A will send a flaw G to the switch for its plugged into the switch will assign that port on F C I D. And if we do a sure flunky database or not switch will see the interface that the storage is plugged into its F C I. D. And it still be W p n The favorite channel switches share of the flaw G database information with each other using the F. C. N s, the favorite channel name service. Each switch in the network learns where each W W p n is and what it's FC ideas and how to route traffic there. Here's an example of issue FC N s database. So the shore ful Auggie Database Command that will only work on the switch where the clients are directly plugged into the switches share that information with each other through using FCS. So if we do a sure FC and ask on either switch, we will see the f c i d. Under W w p n off all of the horse to the out and our network because of the f. C. I. D. Is derived from the Dominion I D. Which is how we identify our switches. Switches no know how to road traffic to any horse than the network. After the flaw, G fabric Morgan process is complete. The initiator books and Dipolog E the Port Morgan. Based on designing configuration on the switch, the horse will learn it's available. Target W W P N's on its storage so you can see the Port Logan process here. Server one Done it. The Boston sent in a flaw g a fob Dick log in, and it was assigned its f c i. D. From there. When that process is complete, it will send a pla G a port log in to its locally attached switch. The switch will check it zoning configuration, and, as you can see from the example here, it will allow you the server to talk to its storage. Finally, we have the pl r I The process. Longan. The initiator horse walks into a pl r I process Morgan request to its target storage. The storage system will grant access to the host to its lungs based on its configured one masking serve activist. See an example again after the server has done its flaw. G on its plug e. It walks into the process. Log in directly to the storage system, and this will allow it to connect to its lungs. 9. Fibre Channel Part 3: servers access to for storage will invaded Lee. Be mission critical for the enterprise. So we're not gonna want toe have any single points of failure. Redundant Favorite channel networks should be puke in place. New it is fab decay and fabric be or sunny on, son. Be each server and started system. Horst will be connected to bull fabrics with redundant HB A porch fibre channel switches distribute shared information to each other, such as Damian I ds the FC N s database on zoo inning. When we configure zoning in a fabric, we only need to do it in one switch and it will then be distributed to the other switches from there. This mix things more convenient for us, but results your a potential. Deng stayed here as well. Where if we make a miss configuration, it's going to be replicated between over switches and no fabric. If an adder in fabric E was able to propagate the fabric, be, this would bring Deng bul fabrics and it would drop the servers connection to their storage . This would be disastrous. For this reason, switches and different sides of the fabric are not cross connected to each other both sides of the fabric are kept physically separate. This is different than how we often do things in Ethernet line networks, where switches will be cost connected to each other. In fiber channel networks, we have two fabrics, fabric game and fabric B, and hosts are connected to both fabrics that the switches are not. Switches are dedicated to either fabric A or fabric. Be went to have a week It what that looks like. So you see here I've got Server one dining at the ball pond. It's good to HB a parts for redundancy. The first port is connected to fabric. A. The second port is connected to fabric. Be I ought to do that on my storage. My storage has also got redundant. HB a port one is connected to fabricate a. The other is connected to fire, but it be the two fob dicks are can't strictly physically shot but it from each other. You can see of signified that they drawing a big red line up the middle here so you can see my hosts are connected to bull far bricks. But the far breaks are kept physically separate from each other. This means that if I have a mis configuration in fabric, a fabric a could be taken down. But that miss configuration cannot be propagated to fire. But it be my server would lose connectivity to its storage over five e. But it can still get that over fabric Be so I don't have a knife edge. OK, but wait, We're going toe have at least two controllers for redundancy off our started system. So our network topology is actually going to look more like things. We've still got the fabric in and defy Bigby favorite channel networks. Where are switches are kept physically separate from each other. In the example, I've got a couple of clients server one on server to which are connected to bull fabrics and no up at the top. I've got to set. But it started Justin controllers for redundancy. The controllers just like the servers again, actors and hosts. So my storage controllers are connected to bulls favorite channel fabrics the way that I'm going to configure my joining on my switches Now, on one of the switches which is in fabric A I can figure is due in first forever. One Sisy Zone name server one member FC Alias is Server one A. That is the H B a port on Server one, which is connected to the fabric a network. And then I also see a member FC alias Controller one E and member FC alias controller to a booth controller. One on control or two are connected to my father. Gay network on my server can reach its storage through either controller actual or not seem fiber. A switch out configure Izu in for server too sexy zoo in name server, too. Member FC Alias as to a. That is the W W p n all of the H B A, which is on server to and connected to my fabric a network and then member FC alias Controller one A on member FC alias Controller to A. The HB is on control, or one and two are connected to the fabric. A network on I tie it all together into a zoo incent. ICI's ruin sent names when she a member Server one on member server to I can figure that out on one of the to fabricate switches, and it will propagate it to the other switch so it saves me having a do a duplicate configuration on both. I also need to configure my fabric be switches. So on there. I also configure a zoo in for Server one has his own name Server one member FC Alias s one . Be the H B A on server one, which is connected to my fabric. Be switch on member FC alias controller one B A member FC alias controller to be for the HB is on controller one and two, which are connected again to the fabric be network. I also need a zoo in for server to so I do a similar configuration there. Zoo name server to member FC Alias as to be the HBO on Server two on member FC alias Controller one be on member FC alias Controller to be the W W P ends for the HB is on. My controllers are connected to fabric be on Tiote together in my zoo incent I got zoo inset names. One cent be member Server one and member server to I can figure that out on one of the two fabric be switches on it will propagate it to the other fiber be switch. So that takes care of May zoning on my switches. I also need to configure one masking on the storage system on my storage. I've got a boot. One for server one unable to go a boot, one for server to I see that the W W P ends for server one E on for server one. Be our bull of able to connect to that one. Server one will be able to connect to it over either Fabio A or fiber to be on further server to boot. One idea the same thing. I see that the members of its one masking our server to E I'm server to be next topic to tell your boat is TPG Target portal groups. All of the parts undistorted system which initiators can access for storage through are added to a target portal group and the example that you see on this light here ports controller oney, controller one be controller to a and control or to be are added to a target portal group. Each of those parts will have its own unique w w p n. I add all four of those Dobie WP ends to a target portal group and the hosts are gonna learn that they can connective or storage through any of the w w P N's. In not TPG. Lula is asymmetrical logical unit assignment. It's used, baby. Start interest in to tell the claim to which other preferred pops for it to use direct pops to Norge. A node is another name for a controller. In distorted system owning the one are marked as optimized pops. Other pops are marked eyes known optimized parts. What? How this is gonna work? So we've got the same example we were working out earlier where I've got a started system which is made up of two nodes which are controller one on controller to controller. One wins of it disks where the one for Server one is currently located. Server one can get Teoh. It's one through either controller one or controller to, but it would be better for it to go Teoh Controller one. Because that is a direct path. The storage system can give the server. All of this information lets it know almost the parts, but it can take to get there, on which are the preferred Pops and users on were to do that so the server learns about optimized path one which is going through fabric A and terminates on HB a controller one a . It on shell learns of a optimized path to which is going through thigh but be on which terminates on HB controller one Be path one and path to are optimized paths because we go to control or one which is where the line is. The server will also learn about non optimized Path three, which goes through fibre A and terminates on the H B, A controller to a unknown optimized. Part four, which goes through fabric be on which terminates on controller to be service server, has four different paths of it it can take to get to its storage, and two of them are better optimized types doing the log in process initiators will detect ports available to connective or storage on in the Target portal group, and allure will notify which of it preferred Paps. Multi puffing software on the initiator will choose which path or paths to take to the storage. All popular operating systems your flavors of windows UNIX, Linux, VM with except Trump have multi path thing software, which supports active, active or active Stein Bay. Perhaps the claim will automatically feel over to an alternate path if the one is using fields. So, considering our example, where we had the two optimized paths and the tune on optimized perhaps on our multi passing software, we could choose to do active, active load balancing overboard optimized parks. Or we could do active standby where we send the traffic over one of the optimized pops. And if it feels we feel over to the other optimized path, for example, here's an example of what multi path thing software looks like from me client. A couple of popular manufacturers for each B. A s are Emulex onto logic. Here we're looking at a screen shot off Q logics on Surfer, and this is software which is installed on the server. It's going to be a client off the storage. We get a nice graphical interface, and you can see the server is over here on the left. This is the server that the software is running on. It's plugged into favor. China will switch one, and it's detected four possible paths that it can choose to get to its storage. We can go on the server run this software, and then we can configure it with the way that we want to get to the storage. So as you've seen client Connick to the eighth of sign, storage is fundamentally different. Toe How Ethernet Networking Works. I already had a lot of experience in Ethernet networking. When I came to start Ege on, I found this pretty amazing. An Ethernet. If you want to connect a client to a server, you have to point declined at the servers i p address, but with Faber Channel. Because off the log in pro sex, the client will auto magically detect its lunge, you can go onto the favorite channel switch that decline is connected to, and you could do a sure foggy database from their toe. Learn what it's W w p n is. You could then configure viz owning on your switch shoes. You can also again using that w w p n create the one on your storage on configurable, unmasking when your storage you can then powered or not client you could boot up from the operating installation she D, and it will automatically detect its and one without you having it configure anything when the client and you could install the opening insisting on its remote one. An Ethernet networking, overdosing and switching decisions are handled by network infrastructure devices, all of them. But in sign started motive Path thing, intelligence is enabled on the client and Horst. 10. Fibre Channel and FCoE Configuration: how much you can look at how to configure fibre channel storage in the lab. Turn back on my netapp storage Justin. I'm to create a new storage virtual machine furnished, so I click on the create button. I'll give the SPM name off FC. I don't choose video protocol off FC slash f shui. There is just one bulks for both pool echoes here because favorite channel on fibre channel over Ethernet work in exactly the same way the use of the fibre channel protocol. It's just encapsulated in an Ethernet header. When I'm using FC Yui, I use a secure age style of UNIX on I'm happy to use the discs of everyone on aggregate one . I checked that my DNS configuration is pointing at my DNs server on a click on submit and continue. Next, it will take me to the configure FC or actually protocol Page and I need to configure my interfaces so arctic the check bolts here on, I'm gonna accept the default of one lift preneurs each new. It is a separate controller in the storage system. Typically, I'll have multiple new words making up about one started Tristen, I want to have an accessible interferes on each of those Nords. So that will give the client multiple paths to get to its storage. Then I configure the 1st 1 for my first client. I'll give it in one of one gigabyte again. Real. What old? This would typically be larger expense if I the open interest in, is running on that client. My client is running Windows 2012 server in the lab here. So I'm happy with Windows 2000 and eight or later, and then I enter the host. Initiator, this is the W W P n off the host. By doing this, I'm configuring my one masking. Only that one particular host is going to be allowed to connect Division One. The one is dedicated to that server. I don't want any of our servers connecting to it by accident. Now, if the one was being used as a the M where data store and hosting all my different virtual machines, then I would have multiple host initiator addresses in here for each off my VM where he s ex hosts. They are going to be able to run those virtual machines I can find the w w. P n looking on the horse itself or by checking the flaw G database on the fibre channel switch with its plugged into. So I've got my w w p n on. I'm going to push the in Their newest the sayings for favorite channel over Ethernet are exactly the scene. It's great out right now because I can We can figure one or the other. If I was configuring after you, we hear I do exactly the same way as I would do favor channel. I click on submit and continue. I don't need to configure SPM LeBeau administration. So I skip this page. Then I'll see May somebody page and I can click on OK, I have no configured Faber Channel on my storage system on. I've created the one for my first client so it can connect and I would get to its storage 11. SAN - iSCSI: next we're going to cover ice. Cozy I scuzzy is the Internet Small Computer System interface protocol. It runs over Ethernet networks and was originally seen as a last expensive alternative to fibre channel. It has higher packet overhead on has traditionally being seen as having lower reliability on performance than favorite channel does. However, you could argue that is just a perception thing. As it runs over Ethernet, it can share the data network or have it stolen. Dedicated network infrastructure tool cards, which times for TCP offload engine are specialists. Adopters, which can be used to offload the TCP I P processing from a servers sheep you there oaks or sometimes called ice cosy. HB is each being starts for host boss a doctor and it is the sign equivalent or being nick , a network interface card for normal Ethernet India traffic. Let's have a look at I were different network topology auctions here. Server to storage on server to client is bull of using a shared Ethernet network through our normal daytime. On our stories, traffic is both going three vaccine switches and also using the team network cards on our end hosts. If you try to do this back in the day when we just had one Giger that you thrown at, then you could have problems. But now that we have 10 to get the Ethernet and even faster speeds, this is more feasible. Another way we can deploy. Our network is using a dedicated storage network in the example. Here we got Server one in the middle went Say that about a Web server on We've got a client up at the Top, which is opening up. A Web page that is hosted on Server one declined at the top sends its traffic to Server one over our standard Line data network. Up the top, the server reaches its storage down the bottom using a separate, dedicated Ethernet network that is just used for storage. And the server here has got redundant network cards. It's got to standard Ethernet network interface cards pointing up to the network at the top fishing decline. It's also go appear off Ethernet ports going down, fishing the storage network as well. Those could be on standard Ethernet next, or it could be on boost specialized tool cards that are used for our ice cozy traffic. No, is that we've got double the infrastructure here. If I go back to the previous side on the server, we've got two parts on. We've got to cable so that are connected to two switches with the second option on the server. We've no got four parts on four cables which are connected to four switches serviced Second option as a more expensive way of doing it. But because we've now got this dedicated network for storage, that should give us more reliability. Another thing we can do is we can have a dedicated view on network for our storage. Use a different view on foot, are normal detail and have them running through the seams. Switch infrastructure showing the picture here. I need to have connectivity between decline on the server guys provided over my date of yuan. I also need tohave Cornick to the from the server to its storage that has pervaded over E set, but it's storage view on when we put it together. You can see that the debt avilan on first Orangeville on our bull of using the CME switches , but we're separating them into separate view ons for performance on security on the server , we could have the data on the storage villain trunk down into the same physical port. Or we could split them into two different physical parts on the server. We'd only have improved trunked into the same physical poor if we had enough. Bomb went there, and we were confident that this was not going to give us any performance issues onto the addressing now. So you saw an earlier Lassen. That Fibre Channel uses worldwide names to identify initiators on targets. I scuzzy uses i Q N, which is I scuzzy qualified names or once commonly, that can use you. I extended unique identifier for its addressing the I. Q N can be up to 255 characters and long and has the format that you see here, for example, the i Q. And could be like, um 0.1991 dash 05 dot com. Don't make herself call on Cash Horst. Doing what? If you're looking an I Q. And and it's got a really old day and they're the date. There's not really anything to do. The host, the i Q N, is derived from when about vendor applied for the I Q. and addresses the I. Q N is the same to the horse. There's a whole similar to the way that the W W N and is applied to the horses are hole in Fiber Channel I scuzzy runs over Ethernet, so individual parts at Addressed by I p addresses Roberton W W P ends that we use in favour channel ice. Cozy does not support Faber channels flog eep log e and P L. R. I log in process. So an administrator must explicitly point the initiator at its target by specifying one of the I P addresses in the target part or group. It will then discover the targets like U N on the other ports in the TPG multi puffing software on the initiator can then choose which path or pops to take to a multi path thing in Nice, cosy it works. Regime is Moby Path thing in Fiber Channel. Although it runs over Ethernet ice cosy, it's still one Sean Pruitt call with multi passing intelligence on the initiator. One masking is also configured in the same way as in favour channel, this time using the i Q n Robert W w p n on restoring system to identify the client. Zoning is not supported in I scuzzy on our switches. Possible based authentication is typically configured on the initiator and target to guard against spoofing attacks on end to end I. P. SEC Encryption can also be enabled if you need to enhance security further, I'm back in the lab and we're going to have a look at how I scuzzy works from the client point of view. I'm on a Windows host here on I'm going to open up the ice Cozy initiator. It'll prompt me to start the Ice Koji service show. I'll click Yes to that, and then the ice cozy initiator window will pop up. I've already configured in one for this host on my storage system. So to connect to that, I click on the Discovery Top in here and then a click on Discover Portal. I enter one of the I P addresses that has being configured for ice cozy on my storage system. That's $172.23 dollars, $3. 33 for this example, and I click on OK, and now if I click on the target's top, I conceive it. It's discovered the i q n off my storage system. Next I click on Connect on Enable mohe path on. Click on OK, and I can see that I'm connected through I'm done in the ice. Koji, Initiator, I'm gonna click OK to come out of here. True, I'm connected to the storage. But if I open up Windows exploder you see that I've still only just got make she drive that I had already on my DVD drive. I can't see my shine storage in here. The reason is with ice cosy it see John Pruitt call. So my client gets block level access to the lung. I need to make it ready for use. To do that, I need to go to disk management Batch. No cold. Fail on storage services in your versions Off Windows Server. I click on disks on I can she the one that I've connected Teoh over. I scuzzy to use it. I need to right click and bring online. It gives me a warning masters to check, but it's no old Eddie and use by another server. It's no so I can click Yes to that. Then I'll see a message to create a volume Start Menu volume Wizard, I'll click or not. Link in Visual opened in You Volume Wizard, which will help me format my lung under saying a drive letter to continue. I click on next. I'm provisioning it for this Horst, which is Winnie on its disc one, which is my eye scuzzy lung. I click on Next. It tells me the disk will be brought online and initialized. I click OK to that. The size of the volume will, by default, used to the maximum size I click next again. The local drive in Mist Server, which is Windows installed, is this she drive? I have a disk drive is Dr E. May. Next available drive letter is e through. I'll accept that and click on Next. I'm on a Windows server so the default file system will be the NT file. Kristen. I'll accept the default for the allocation unit. Size the volume label Alcohol. Vicks I, Scotty one. And then click on Next, actually a confirmation page that some anizers my selections on. I click on create. Once it's done, I can close this window and I can click on the volumes page up in the top left I don't see my eye. Scuzzy one showing up there as ve drive. If I open up windows Exploder, I see my e drive in here now, and I can now use May storage. It acts just like it would if I had a local drive in. My server here on the server has got block level access to that desk. 12. iSCSI Configuration: I'll show you how to configure. I scuzzy on the storage system. Now here I am, back on our netapp system again in storage. Virtual machines on. I'm going to click on Create. I'll give the ESPN my name off. I, Scotty Andi enable the ice Cozy protocol. I'm happy with security style of UNIX on using the discs. That makeup aggregate one. My DNA ash configuration is pointing at my DNS server, so that's all good. I click on some and continue on the ice cozy configuration page. I'll use a target alias off. I scuzzy. I'm gonna use one wife Per knew it again. The CMAs I did in Favour Channel. I want an interface to be available on every Nord that makes up my storage system so that the clients will have multiple passively can use to reach for sign storage that gives us redundancy and also enables active, active load balancing from decline. I'll configure a one for my first client. I mean, use a small size again off. One gig abates. This client is Windows 2000 and it or a letter. And for the host initiator, I'm going to enter that hosts I Q n Fibre Channel users w w piento Identify the hosts So I use the W W P M here when I was configuring fibre channel ice Cozy uses I Q ends. This will set the one masking with my storage system, so only the correct Windows server is going to be able to connect into this one. I get the ANC U N from the ice. Koji Proper is page on the host on I paste it in here, then click on submit and continue. No, I forgot. Teoh ashamed. An I p address. I'm getting a remainder, so let's do that on will use I p Address 172 door, 23 door, $3.33 Subject Mass to 5 35 to $45. All Onda Gateway 172.23 2 or $3 to 54 in my lab environment here. I've only got one node in this storage system, so it's only asking me it. Configure one I p address. In the real world deployment, you would have multiple. No, it's for redundancy. It's like the broadcast, a mean which is just Hey, the interface news. Which other interfaces it can feel over to if it has a failure on click on, submit on. Continue. I don't need to configure SPM level administration. I would only need to do that if I had some administrators who I was going to allow to administer this SPM but no alive them to administer my other ass. PM's I don't need to do that here. My storage administers. I don't want to configure all the different SV ends, so I don't need SPM level administration. I can skip this page. I then see the somebody on I click on OK, I've no enabled X cozy on my storage system on have created VA one for my first client. 13. SAN - FCoE Fibre Channel over Ethernet: The last John Pruitt coat to cover is FC We Fibre Channel over Ethernet. F shui is the newest son critical. It became possible with the advent off 10 gigabits per second. Ethernet, which is enough bandwidth to support booth. Data on shortage traffic on machine adopter. Really importantly, EFS Yui uses F C p the favor channel protocol the exact seem favorite channel protocol is used with native favorite channel with EFS Yui The favorite channel protocol is in captial e it in an Ethernet header to run over Ethernet networks. She'll FC Yui It works exactly the same way as neat a Faber Channel does. It's just that we encapsulate about seam tape of traffic inside and Ethernet header so we can run it over. Ethernet networks que s quality of service is used to guarantee even acquired buying whipped to Pistorius traffic to stop me. Normal data traffic from hogging too much of the bind width negatively impacting our stories traffic and it retains video liability under performance Off. Need a favor channel One of the main benefits we get from efs eui is deceiving. Zweden, get in our network infrastructure The slight here. We're looking at the network infrastructure in Native Favor Channel. The server is connected to the Ethernet local area network with Ethernet Knicks. We're probably gonna have to your boys for redundancy. And it's also connected to the storage network using fibre channel HBH, horse bus adapters and again will probably put two of those in for redundancy again. Verse four adopters four cables onto four switches when we use neither favor channel competing that to fibre channel over Ethernet networks. Server two stories Traffic on server to client traffic users were seems shared. 10 gigabits per second Ethernet network. So now we just have to adopters two cables on two switches. We've cut infrastructure down and half. We save on the hardware costs because of X, and also because we've got half the cabling and half the switches that's going to require Last rock speaks whence power on glass queuing more cost savings. As I said earlier, favorite channel over Ethernet works exactly the same way as need to favour China. All those f c p. The favorite channel protocol works exactly the same way as it does in near favorite channel. It's just encapsulated in an Ethernet header so it can travel over any three net network. We still have W W P ends on the initiator on target, and we still use flunky foggy on pl r i. Morgan process in f shui Bovis storage under the A traffic is going at the CME Shared physical interface. The stories. Traffic users f c p. It requires a w w p n the Ethernet Data traffic requires a market dress the way the Ethernet via traffic and F C P stories. Traffic works is totally different. So how can we support them both or in museum physical interface? The answer is we virtualized about physical in your face into two virtual interfaces. A virtual nick with a mark address for the Ethernet DEA traffic on a virtual h b a with a W w p n for authorities, traffic and distorted and India traffic are split into two different violins. A data view on Onda storage Feline went to take a wicket that in the diagram here, we've got a single server server, one over on the left. It has got to physical interferes cards in it, but she anyone and she any to for fibre channel over Ethernet. Our interface cards are called She Aeneas. A converged network adopter booth she, Aeneas are going to be split into virtual adopters for data on storage. On the slide here we see how we're doing that for data. So she anyone on CNY to have both got virtual Knicks on them with Virtual Nick one on see anyone Virtual Nick to on CNY to Envoy's Virtual Knicks will both half mark addresses as saying to them. On the switches were trunk ing the date Avilan dying to physical port on the CNI on no ish in the middle here that were crossed, connecting our switches for the data villain traffic. We also have virtual HB A's on the converged Network adopters. So we've got virtual HB a one on See anyone on virtual HB a Tu on CNN A two on we have W W . P ends on our virtual HB is we're trunk ing, the storage villain doing from the switches to the virtual HBH on the converged network. Adopters the newest that this time we do know cross connect our switches because we need to comply with best practice for our favorite channel storage traffic that says that we have a fabric came on a fabric B, which are kept physically separate from each other. If we put the whole thing together, you see that we're running, Guv, our data on our storage traffic over the scene shared infrastructure. We're trunk ing both the data on the storage villains dying to single physical ports on our she Aeneas. And then that traffic is split out into a virtual nick for the D A traffic with a mic address on it on a virtual hb a for a restorative traffic with a w w p n on it. Another thing I need to tell your boat is the lost source. F Shui Fibre channel is a lossless political. It ensures that new frames are lost in transit between the initiator on the target. It uses buffer to buffer credits To do this. Ethernet is not lost. Lex TCP uses acknowledgements from the receiver, but to the sender to check the traffic reaches its destination. If in acknowledgement is not received, the sender will resend our pocket FC. We uses F C p, which assumes it wants source network, so we need a way to ensure our storage pockets are not lost. to well traversing the Ethernet network. The PFC Priority Full Control FC Yui Extension for Ethernet is used to ensure about lost. Let's deliver T p f. She works on a halt by hot basis. It's no end to end, so we don't just need to support it on our end hosts each nick and switch in the path between the initiator and the target must be after you. We capable So you can't use just standard 10. Give the Ethernet Ethernet switches. The switches have to be after you. We capable. Okay, so that was f shui. I want to wrap up the introductory son lashings by giving you an adapter So money The Nick is a network interface adaptor. It's a traditional Ethernet network card. It's used for Naz Pruitt calls on ice Scuzzy Tool is a TCP off would engine. It's used to offload the TCP I P processing from a server CPU. It can enhance performance from Naz Prue echoes on Thai scuzzy. H B A is ah, host bus adopter that is the fibre channel equivalent of an Ethernet nick. Ah, nice cosy H B A is an Ethernet tool card optimized for ice, cozy she any is a converged network adopter. It's a 10 gigabit Ethernet card, which supports F Shui. UTA is universal Target a doctor, but the Netapp proprietary card, which supports F shui or favorite channel on our storage system. On lastly, what's her? Don't look at the portico stock can partisan between our different sign protocols. Which son The storage system presents a one to the client, which looks to decline just like a ziff it hard. A local hard drive installed in it so and chassis at the lowest level of the client. Sends, reads and writes to scuzzy to the desk with Neetu Faber Channel, but is encapsulated in F C P, the favorite channel protocol, which runs over favorite channel networks. Favorite channel networks are different than Ethernet. Over. We don't only here one is the USA stock. So we did acquire favorite channel switches on HB is in our and hosts favorite channel over Ethernet. Oxo users favorite channel political, so it works exactly the same way as need to favor channel does. But the F C P traffic is fervor and cap surely it inside efs Yui we put any throw net header on there so we can know Run our favorite channel protocol over Ethernet networks. And we can use a converts network adapter in our hosts to run data traffic on stories. Traffic bull Very Besim physical part I scuzzy is very different to need a favorite channel and f shui. It does not use the favor channel protocol discuss the region. Rights are encapsulated in the eye Scuzzy Pruitt call that has been fervor and capture Lee it in TCP and I pee on it runs over Ethernet networks.