NetApp ONTAP 7-Mode Hardware Architecture Tutorial (new version)

in this lecture you'll learn about the on top 7 wood architecture 7 wood is a legacy operating system from NetApp that isn't available anymore so if you're wondering well why am i covering it here then there's a couple of reasons for that first reason is that there are still a fair amount of seven wood systems out there in the field so it's good to understand them because of that and the second main reason is that to understand the current clustered on top architecture and why it works the way that it does it really helps if you know about the seven wood architecture first [Music] on tap used to be available in two different moods at the same time that was seven more and clustered data on top which is also known as cluster mood or C dot for cluster data on top the current implementation of on top is the clustered version both moods did exactly the same job but is acting as the operating system and controlling everything on the storage system and when you bought a Fire's hardware platform you had the choice you could either buy it running as seven mode or running as clustered because it's the operating system it's one or the other you couldn't run them both at the same time you had to make a choice seven mode evolved from naps original operating system that was data on top 7g and cluster data on top evolved from the acquisition of spinnaker networks in 2003 and data on top GX which came about because of that now you're hopefully wondering why would now have two different operating systems doing exactly the same job running on exactly the same hardware that seems kind of strange right well the reason is that rubriz own limitation with both versions cluster mode was more scalable than seven mode but the early versions of cluster data on top were limited in features compared to seven mode in versions earlier than 8.3 nap released on top in both modes and customers made the choice the either chose to deploy the full feature set support of seven mode or get the scalability of cluster mode but they couldn't get the best of both worlds we had to make a choice at that point now work towards having one operating system which did provide the best of both worlds and the worked on achieving feature Part A for cluster mode in each new software release and from version 8.3 of on top they achieve that so now there is just the one version which is the clustered on top that has got the scalability and it's also got the few feature set as well so now you don't make the choice best of both worlds you're going to be running cluster data on top and because there is only one version now they've simplified the naming and it's simply knowing as on top now okay let's have a look at the architecture of seven would and then you'll understand why it's got limitations as far as a scalability goes so we've got the controller the controller also known as we had is the brains of this system that's where the CPU and the ROM is it's also where the operating system lives and on the controller we've got our physical ports there that can act through our clients and also to our dis shelves with some models you can have internal DeChellis in the chassis you can find out the capacity by adding additional external disk shelves so let's have those next the connection to the shelves but uses a SAS connection and the connection gets daisy chain so we have a part on the controller is connected to a part on the back of the disk shelf and then that this shelf gets connected to the Nets the Shelf going down in the stack the maximum amount of shelves you can have in a stack depends on the model of controller and shelves that you're using now if you look at the diagram here you can see that there's a single point of failure there as far as the connectivity to the dish Elvis goes if any of the cables fails then we lose connectivity to all of the disks below that so to get redundancy for the connectivity to our dis shelves we have a second connection going down to the bottom shelf in the stack now if any one single cable fails we can still get to you all of the disks in the system there okay next thing is there's another obvious single point of failure there which is the controller so you can run on some models all five systems with a single controller but real Watteau typically your di your storage is gonna be mission-critical you're not gonna want to have any single points of failure so for that you will put in another controller to give you redundancy for your controllers now a quick note here looking at the diagram in most of the modern platforms two controllers fit inside the same chassis but in the diagram here to keep things simple each controller here is a single controller okay so it's one chassis but there's a single controller in the chassis so over here on the Left we've got controller one and it's the shelves and over on the right we've got controller two and it's got its own dish selves as well now obviously we're gonna need to connect them to get the redundancy we want the the desks that they are on the desks to be available if either controller fails so to enable that we can connect controller one to control our twos the shelves we connect it to the top shelf from the stack that gets daisy chain down and again we've got that second connection that's our mph a multipath high availability connection and we also need to do the same thing from controller to going to controller one as well so now if either controller fails we've still got connectivity to all the disks with all the data on there we put another connection in between our controllers that's the HJ high availability connection the controllers send each of our keeper lives over there so if the keeper lives aren't received the controller will realize the other controller must have gone down and it will take over okay let's look what happens under normal operations so you can see here under controller one we've got a set of the idea set one which is owned by controller one in on top both seven would and the current clustered on top as well the disks are always always owned by one and only one controller now the ownership can move if need be but the disks are always only owned by one controller at a time so you can see here for our date set one it's on disks which are owned by controller one controller one is connected a to the network where we've got the clients whenever the client comes in to use via the incoming network connection is going to terminate on controller one with seven words for a single set of data the networking is active standby the connections are always served by controller one for the disks that it owns on this controller two is always going to serve the connections for the disks that it owns for day a one we do not have active active load balancing with connections here in controller one on controller 2 they always hit controller one if controller one fails it will feel over to control or two but it's active standby and here you can see that controller 2 also has it's there as well so we've got data set 2 which is owned by controller 2 and obviously whenever clients come in to access their - that is going to be served by controller to controller 1 is assigned by unit ok but what happens if we do have that controller failure so here controller 1 fields when that happens controller 2 is going to stop perceiving the keeper lives over the hea connection so it realizes controller 1 is going down and it needs to take over ownership of its disks on data so when that happens controller 2 will take ownership of controller ones disks so you can see before but they a 1 was owned by controller 1 and day 2 was owned by control or 2 when we have a failure of control or 1 controller - orange dr1 and dr2 as well it takes ownership of the a1 so that it can serve it for the clients the network connections also feel over to controller 2 as well now because of us because we can have failures it's recommended that you do not put more than 50% wood on either controller because if you did if you had the controllers running a hundred percent well when there was a failure it would be trying is serve at 200% it can't divide so performance would be severely degraded so it keep acceptable performance recommended doing on more than 50% would on your controllers okay so that's how the architecture of seven mode works you'll see when we do the next lecture the clustered is very similar with one major difference which overcomes the limitations of seven woods and what those limitations are is there's a maximum if too fast controllers can be configured as a high availability a cheap air and managed as a peered system with seven modes you can only have two controllers in the CM system this on vacant ruler one will always be accessed through controller one it does not do active active would balancing for the same set of their it does active actively balancing four different sets of their but not for the same set there's a limit on the maximum amount of disks and throughput that can be handled by a single note so it was a performance limitation there as well and you can't scale out a seven word system it's a limited to two controllers you could scale out by purchasing additional each he appears but they would be managed as separate systems which is obviously going to be inconvenient from an operational point of view it's easier to manage one system than it is to manage multiple different systems they'll also appear a client's or separate systems as well which again is going to make it more complicated to manage from an operational point of view there can be moved easily and non disruptively between disks one the same controller in seven mode but moving they are between the two different controllers in the h-shape here or between different seven mode systems is disruptive to client and more complicated okay so those are the limitations of seven mode you've seen the way that the architecture works as well in the next lecture I'll cover the cluster wood architecture and you'll see how it overcomes with scalability limitations thanks for watching if you want to get hands-on practice with now up storage you can download my free how to build in that up lab for free ebook it's got full step-by-step instructions on how you build a complete net out lab and best of all you can run it all for free on your laptop and if you want to get my complete NetApp course which covers everything you need to know about NetApp storage you can check out the other video that you can see here - thanks

Channel: Flackbox

Views: 2,961

Rating: 4.9245281 out of 5

Keywords: ONTAP, Flackbox, NetApp, Data ONTAP, 7-Mode, FAS, Storage

Id: EYCAIKLb1uk

Length: 12min 22sec (742 seconds)

Published: Thu Jan 03 2019