Dell EMC Unity Product Overview and Architecture with Kaushik Ghosh

so I have a two guys here with the with me from my team I'm Pierre Luc echad Leafers I'm I'm a a product manager lead product manager customer operation for the mid-range I need to explain to you that the mid-range now in a Dell EMC include everything was aerated Dell storage and heritage EMC storage so we have in my portfolio you have the PS line also known as equal logic the SZ line we have also the unity line the vnx line and all the other things that you know from the past but for today we decide to focus on one of the product and because it's one of the product that is optimized for flashing our portfolio with position always for flash is our mid-range and also is our unified product is the unity and we want to speak about that with the two important people that they are more important than me probably and unfortunately they work with me from many years I don't know how they feel about that so coffee gosh here is my project manager lead for unity was the lead also for BMX and also is leading the SC line so I try to combine the team so he leads both things we have an ivory team between the two hm is the smart guy that is going to do demo they told me to not doing them anymore so I'd like to go to in the lab but that's a problem we're going to spend a little bit so I want to make sure that you understand right unity has more than 250 feature is an enterprise storage so we have replication we are all the things but we want to focus on particular things that they are new or they are important was in this moment for our customer and also I saw a lot of things that we discuss in the community so we're gonna do a new unity overview very high-level overview but also giving some idea of how successful is this product in the market today also some architecture overview and then we go into Unisphere and cloud IQ cloud IQ is our service in in in the cloud as marked service in the cloud and Unisphere is our element manager onboard of the store edge I saw yesterday Twitter I don't remember who was but you guys was a there was a grave elapid from a grave with the storage admin is dead we don't think it's completely dead but with unity we try to make the things very simple right the days where we play a lot and create a raid10 with 50k drive and 10k and I can have a stripe vertical striping and try to do all these things unfortunately it's gone in some case but for the benefits will accelerate the adoption and the ability of ease of use so that's what unity is so I let a cauchy going in through this I'm the guy that is going to answer the question if you have question I hope you have a lot of question and they said we cover only this but if you have caching question everything the replication and everything really has a lot of programming product so it is possible to kind of get also curve any idea where unity fits yes rodders especially between the beginning sadena yes so unity is really the replacement of vnx right is our next generation we did a lot of things there as I said that is a unify the v-max is the I n so the I n is the multi controller unity the dual controller autumn eyes is the best damn storage in industry in the sense that this start very small in its growth so it's really mid-range scale at architecture right we have the FC line as well the SC we optimize that for hybrid because as you know the FC was the inventor of data progression so it's very granular data progression we can add dedupe and compression on spinning media so we optimize them to like that and that's what we define as our mid-range in our portfolio as the LMC v-max is the high-end scale I owe that was just presented is also di n um be annexed what was VN x unity start very low as C as well start very all right so that that we have the our portfolio we are the mid-range so in the mid-range the other things is we have a lot of use case so that's why we have the unify we have all these things that we offer embed a replication migration and all these things okay I answer the question I hope the perfect link is it all right good thank you very look good morning everyone so I'm partial coach of The Gazette site I manage the product management team for for for the mid-range group products of that unity SC b NX b NX c but today we are very talk about unity unity all / how many of you here have been unity presentation or know about unity some of you have okay great right so so that's basically what we go over so we'll do a little bit of the overview some of the high high level features and then we'll do a little bit of the architecture of your day as well right but before we do that let's take a quick look at some of the stats and little stats so we launched the unity last year second of may to be precise just before le MC world and since then we have done a the product has done really well right so from a revenue perspective just short of a billion dollars now in this one-year time frame ten thousand-plus systems and five thousand five hundred plus customers today with this unity product and just just completed a year basically of this product we actually at this year's delhi MC world we actually refresh the platform the unity platform we have we have the new hardware there with the with the you good at more memory more CPU power recently but the key part of this story is basically a unity from an architecture perspective you can scale independently capacity as well as as well as as alice if you right so so as an example a customer could buy us a huge a 350 f as an example all our boxes starts as a compact q form factor so you start with the 350f as an example and then as there as the need more capacity they can just add drives or if they need more throughput more CPU more bandwidth more performance or I ops they can score scale horizontally and go up to us to a larger system right so from a capacity perspective as you can see we can do all 16 petabytes of raw flash this is before taking into account efficiency the impact of efficiency the customers trade back in the controllers in order to scale up that is correct so what what we what we do is do you offer online data in place upgrades so they buy the new controller and then you can do an online replacement of the controller to the new controller while your while you are still doing i/o everything is running you do an online upgrade to the to this controller okay so unity basically you can try it Stan okay you can buy as a as the hardware like I was showing you a or a purpose-built hardware you can also buy it as a converged system this is our V block VX block system unlike other other solutions this is a product sort of solution but what I mean by that it's it's likely integrated together and there's one Dell EMC and um the support system that's standing behind it right you have to call up multiple vendors and so on and then the piece which is sort of new we introduced last year is this is this virtual software-defined unity so this is basically the same unity software but without the hardware right so so what's the use case for this so saying customers who want to try now test it out you can do that you have a project short-term project and you do need a solution you can you can use this software for that project when the project is over you can delete it it gone right and the other use case which is coming up recent these days is the remote office use case so in remote offices custom nowadays going with hyper-converged type of solutions right so in that type of a situation we're trying to reduce footprint you can put this Software Defined Software Defined the unity in that environment and and the added advantage that the solution provides us because it is the same unity software stack you can replicate it to say for example of converse unity or a or a standalone unity in your main primary data center I know that's a huge advantage that we provide over other standard alone Software Defined solutions right and you can for the storage side of this you could use a for example us of disk which is inside the server or you could use like a scale io type of storage and a be send under behind this so it's very flexible from that perspective so the other things there that I like to give you some secret here right that you guys like this the secret is also we use these we change the way we developed our systems because we actually use vehicle instance in our test environment so we can accelerate our development using actually a completely stack of unity so that allow us to bring to market more weekly things that in any case the will require more investment from ardell custom right so that's another big thing and as I mentioned we have more than one of the key features stuff like that you can download that and try by yourself then you can't call me up if something you don't like because the next question is what's a little support on the virtual instance yes so the level of support we give that for free and then we give a capacity based support so you can buy 10 terabytes us up to 15 terabytes and basically what we offer you is the normal support this is support contractors pedometer yes and the good things that's another interesting things you can upgrade a let's say a customer download this on you download this you use the order free 4 terabyte and then you want to purchase and make that a professional you don't need to destroy your data you just apply the license and we expand the capacity immediately and just react to what we look at this is this is a subscription-based license so but it gives you a nice flexibility so you need it for a couple of years you don't need it after that and delete it and don't have to pay any it's sold standalone you don't have to own that off grace - mines first threat in terms of developed life cycles you're talking about how quickly are you rolling a new version internally after you releasing customers and did they deposit themselves or are you able to remotely deploy on their behalf so generally from a software perspective this aligns with our software releases so on software releases we have major releases coming every six months and then you have finer religious coming every in the middle basically between so every quarter so we develop obviously releases every week basically but then we test and we make sure that the other obviously needs to have the height extender so we have two weeks cycle and we doing nine to eight cycles for the major release then we have service packing stuff like that at the deployment of this is a VSA it's very easy you just next next next we already pick the CPU memory that require and you just deploy you present the basically data store the human supported VMware ESXi to know the a file such please wait for it okay you can have an AWS version as your version something like the run of their own so went out only for ESXi yoga and sorry but in terms of actual deployment what so you've got in the build custom store either visible yes rules there's any upgrades on in a six-month cycle you pushing that customers we are learned the customer that there is an upgrade available you can pre stage the upgrade it the VSA works in the same way of the physical so there is no difference you get alerted and then you decide when you want to do have a gray then also we notify the cash money this is Master EV upgrade or we're going to speak about that later the cloud IQ allow us to target specific customer based on the configuration that you have and tell them you know you should upgrade another catchment that they not need that feature or they not use that use case we not we not force them to operate right all right represent your unity base that current release level sorry what percentage is the unity basis at the current release level you think we so we previous release we are about I think 60 70 percent that current release we just release the service pack I think we are about I will not say probably 20% the adoption is pretty quick you know depends on what feature we boot I mean obviously if there are very important fission people waiting for that obviously they can adopt more much faster but this alert the automatic alert I'll allow us to force adoption of new release very quickly right fine so let's let's let's build upon the time all right so so with unity when we launch it one of the things we did was try to make it simple so you'll see the scheme card come across the presentation simple and flexible so it's part of the simplicity what we have done is basically all the software on the box today is now included but like the logo said there are tons of features out there and all of that because I'm not going to touch all of them here today in this presentation but that all of that feature functionality is included when you buy buy a DVD box right and that applicable utterly to the box but also to the software-defined version as well as to the conversion so let's now look into this box what does it look like so it's a it's a 2-u form factor with a very compact box right like as recited brain there are simple if I open the cover basically there are two controllers out out here drag stack one on top of each other so the sba SPB that's what we call them the two controllers it's connected with the CMI bus on the back end of the on those controllers there are 25 drive slots right so when way does his demo he can he'll show you there's a carousel in the units where we can actually rotate it and see what it looks like to the other side so that's what it is unlike other architectures in in our in this situation both controllers are active both are running right so both controllers survive both controllers of four lungs it's a unified so we can do lunch we can do file can do revolves open stack everything basically runs in parallel across those two or across those two controllers the cache is mirrored using that CMI bus so in any if any of the controller fails something you're all your caches that level only or the other on this mirrored controller right in addition to that there's also battery backup builded if both controllers have books the entire system has to shut down then the battery backup picks up of all the imposed everything that's in memory so 12 gigs at back-end right like we have built-in ports built in 10 gig built in 16 gig private channels built in 10 gig optical so these are all building ports and then if you want additional i/o modules then then we can purchase those and we have some slots that you can fill those with those populations and regional i/o modules so support 32 gig fibre channel that's not right now how here sometime next year is very looking ok so so again so active active unified designs if you are familiar with the vnx architecture so that this is it these two controls vertically serve you files they serve the block evolves everything there is no other additional deselect review for a block will be assigned to one of the controllers or I'm sorry one will be assigned to one controller or the other that our data correct and a filesystem would be assigned one controller and not right that is absolutely correct but you can have multiple lines multiple file system for a single controller so it's actually it's actually completely embedded so you can oscillator when we do in the day mode you can provision v-ball file system online in the same way that we did the user doesn't and we share the same resources so we use the same snap technology we use the same pool of resources this is a completely different is a containerized architecture that we create for you it so it is before the container was created they are device container that we create where we have blocked and file and vivo are running independently using the same underlying storage premise block asset right so to your point right a file system you're not gonna provisional and you provision advices but we underline with doing all the things that we require for them all right so I think I talked about did you say about a total block and like a LAN and a file system can share same storage yes right they can share the same storage pool remember you only an axe you need to provision in storage and then create the path difference then all of this is not here yeah it's gone yep it really do it as part of the demo as well just for sure yep so it moves quickly on that okay and then the other part of it is the density right so with those 25 drives you cannot fit in almost like a half a petabyte of data in this to you form factor all right so the retail unity is architected for file flash right so we do get some questions phase this basically is this basically a regular system if you just populated with flash drives called it all flash what did you do something unique some special things the distant for all flash right so so so wanted to show you some of the unique capabilities that we have in this product for all flash so one of you are all tell all the drives in the system are now three details in and right we are at one right but it today we looking to go going through a map right per day before the end of the year right everything will be half right per day now as you understand going to half right per day may sound simple but it is not because we do have to do a lot of good work on the cold in the in a in IO stack to make sure that the drives don't wear out commercially so there's a lot of work that has to go in in the product now we are full multi-core optimized so every process everything that runs basically runs across multiple cores right we are not pinned to specific force so that ensures that basically all your cores are evenly balanced evenly distributed right so sort of maximizing the resources that I have as a mid-range system occurs in the top line system okay I 1214 I just had it in the largest line for portunity then look at them that's okay yeah Tom for the next P oh yeah yeah we do have a automatic flash we're balanced right so so what what it does is basically if we we monitor every block right so if there's a block we find that it's going to get worn out or coming here so we'll make sure we redirect the aisle to another block not only on the same drive it could be perhaps in another drive in the pool all right so we do that we all flowed off a lot for garbage collection to our custom firmware that we build so that what that means is that we are not using up our CPU cycles to do garbage collection we are not requiring customers to allocate excess space in order for us before doing this garbage collection we also have something called in-memory log and I'm going to talk about that in the next slide or that works right polishing the supposed right right of good stuff in line compression I'll talk about that and the other thing is in this pool you can mix different type of flashback say if you started with for example a 3 terabyte black stripe 3 write poetry terabyte by Skype economics mix it with like a like a seven terabyte one right video flash drive in the same pool but we allow you to mix these drive tips and we rebalance the aisle not only from a wear perspective but even from an i/o perspective we do that rebalancing across these drives alright but now let's just get on the inline compression P so are you doing that in my MB RAM or you right in to some free space so we in memory the amount using memory it's a we use the de-rank that we have been past them right that we use for not only compress so that's what is gonna explain every expensive and then this is where the the inline compression happen another things up right them so so from an i/o perspective so what happens is we have this log file right so think of it you brought up nvm so this is basically similar to mimics 10 nvm behavior so we have a log file this log is basically coral on business so everyone has its own you log file sitting in this in our in our memory right io comes in on just like with any data bases right you first like to a log before you write to the database of i/o content we write to this log file right as soon as we write to the log file we acknowledge it back to the host yep iOS taken care in the past like you know previous architectures beneficent amen I would have to go back and allocate space make sure space is available all that stuff sort of all those operations which I used to do before I don't have to do that anymore right now you may be adversely memory them specifically so are you right you're writing into a single memory demand then acknowledging the right are you writing into a couple of them in several days so it's our a global memory space yes we have a so if we're learning the engineer of detail obviously we have memory budget that we are located for these specific things so we have our MCC is our core cache memory for the underlying memory this is on the top is an elevated cache at the top that basically use a memory log that we give from them but the data is replicated to the - yeah I'm hearing that me I mean this is is mirrored also is protected by the world so is in our DRM protection right so this can have percent good system right so so important thing to remember is there is no specific separate memory card or anything like that right this is just sitting in your regular - which is mirrored backed up by the battery rack right so that's another beauty of this architecture is that you didn't very back up knock that that Super Cab backed up no he's better than sorry there is two battery and we do the bolt on a MC the device inside of there is a two mSATA device inside of each processor so so as we do these rights we start at the right accumulating this in the log files right our sorry and then our inline compression comes in say we're asking this question where it happens so this is before the data has written has been written to drive with this while it's still sitting in cash at that time our compression logic starting in kicks in and starts compressing this data and then we would Wallace these rights and write it together right wherever we can find space of participation or space then we'll have to split it up but as long as this space we can they'll try to do this post type writes the right and bring multiple small rights by due to one a fewer larger rights they don't necessarily wide stripe the data across multiple flash drives sorry we we do not wide striping the data yes there are yeah we try to white stripe when we when we can or where is the mosque on equal space also we have a process that is a we call internally space make up they make sure that we have all these slice of able so we did the beauty of this that it is not only for in my compression if you allow us to water the i/o that comes in and always try to optimize this is present also on our I've read version so you get a lot of benefit also on the average when you do food right right that's right and our would basically include what drives with stripes and striping is all building into the food right so when you write to the food earning all the data gets the memory log is really on container level so it's a virtual volume or a lot under two thousand seven sir so it's separated from the disk under the nourishment but remember the MCC that behind that is the global K so we have a very good granular we know what this piece of code speak the rattle with the MCC so there is an API that will tell them what is the last incoming what I should keep and and we not this stage we the stage the pages but we have if you look at the MCC design we have something that is unique that we keep we say this page is dirt but we give that page until what you want to use that so if some coming in and want to reuse that page we can provide that paper so and then the last piece of this is that space maker we talked about it right with our background process this is looking at the blocks I talked about earlier right for where if there's a particular block being worn out we balance we balance it we distributed similarly we do the same thing from performance perspective as well why this is important right because you are mixing different types of last types right so you have a 3/7 a 15 to bits like that all things being equal your fifteen to edge is probably getting hit more time than the Twitter event right so that's why it's important to understand these blocks right if there are blocks on a $15 that are getting hit let me move some of them to the three terabyte that sort of it's sort of key try we try our best to make sure it's balanced across these different types of types so so benefits so one consistent response time right we lower flash consumption make sense right with the inline compression longer flash endurance right so the two things we are doing out here one we are writing less data and we write less frequently and then lower dollar per gigabyte now one thing I didn't mention earlier and I just want to make sure I guys understand is that I pay for example we write something in cash and we acknowledge it back but when it comes to write to the disk we find that a all the space is used up what happens then we can't go back and delve level or take back your IRA so what we do is a B under the covers for every lon we actually reserved two blocks of space right and which is enough for this particular loss we do have such a situation it is backed up by storage behind where I can upload it it's not a common thing it's not that the all the doing that but just letting you know that there is this option out there all right so as part of these part of this story three critical component of this is TOS in an all black story right why it's important because in a hybrid environment for example all right you had different types of dry drives for an application that wasn't running frequently it would have fallen to the near line sass so automatically sort of did would have done a little bit of would have done some sort of like they're able to have collect hearing of the I write with all flash everyone's equal write an application which is running every day and then I've been a bad job which picks up at the end of the reporter would sort of use up the same CPU cycle so to us is very important us is building into the product we have it for blocks we have import a snapshot as well we have it for V Vols so on right we can limit by I UPS we can limit it by bandwidth and then one of the new capabilities that we introduced is this something called both support so bus gives you a little bit of flexibility so you can say that hey you can say that hey in the morning for example when everyone is logging in I would allow allow the u.s. limit to be exceeded by like say 20 verse for about an hour from nine to ten right so that gives you a little bit of flexibility now with regard to the limits so and this is a I want to use a real Q s but is a completely rewrite of us so we use our you're tagging so it's it's it's very granular in that sense and again you can try to download the vaporization in water source on the beautiful adhesion and I can apply a different water like snap and clone and everything it's it's a very we not take on a cue that we're not using the cue that it's really a your tag so that's why we can guarantee when you apply the policy that that does what was then the same taneous sort of reactionary narrative is a guarantee rely on no there is no getting on oh so we guarantee via your for the products that you said we not guarantee a you at the real entire system level right you need to know what are your system capacity what's happening when you actually hit a limit are you dropping right are you just everything Lane yes we obviously you do unless I hope so latency goes up right I mean now you need to tweak the fix you don't want the latest to go to the creando millisecond sometimes but but that depends on his best practice of how to apply these right full alright so so I mean we do is give us a snapshot yeah and I'm gonna show you an use case for that okay there for example a test abuse case right you take a snapshot you gave it to your you mark men with it you masculine snaps with Austin and you start to do testing them and you have your primary land that you runs in SQL and you say okay my primary line is the one that needs the best service I want to give to the test guys only five thousand I hope it is be doing tests I don't want what fat the entire system lies metric that's a copy on right now no it's a good recommend right redirect them right we change this as I said we have a things difficult they are going to do show you that aspect of it we have this feature called end loans right and we're going to hear them get put them of that yes so just to quickly show you like from a density perspective you have a great density story right 3.8 for these are old one right for the flashcards by the way right 15.4 we launched introduced it last year right the thing which is new we are we announced that del you see what is this dense shell but then she'll fit a lot of eighty drives through your palm factor fits of standard rack so custom another buyer of deep rack I just put it in the regular act and if you take that don't mind it with the 15 terabyte with the 2x efficiency right you are looking at almost like a half a petabyte part you put your compression level into the water to down and everyone so it said we access everyday everybody use compression it's actually it's a two to one really effective on the our installed a obviously if we look at there are people they get both one and more but which one is having Robert so and I'm going to talk a little bit about the pools I touched upon it earlier so the pool is basically the thing so we have changed it so this is because it the dynamic pools we talked about this earlier from the same pool you can carve out lands you can carve out files you can carve out B bones right so it's an unified pools are you require a minimum of six drives in the pool so it you can start with a really small capacity right depending on what your Escape is so my cost perspective very affordable you can expand it as low small as like one drive right why this is important because as the pledges are becoming larger right you don't want to be every time you add capacity I can't expect my customer to add like tens I right entropy job I Drive like so this is important difference do one of course if you want to do more you can go all the welcome to do that so sorry so I have so how does that work in terms of in the old days you had private records houses damaged field right now what does that look like on the recoveries right so under the cover is we call this method rate so basically we we use the vr-12 chance from from an n we used hours and extents and we use a route our rate technology so the extent basically it's represented drive but is actually a chunk of that drive and that allow us to carve out the space for the on spare from the top of the tribe also allow us to do parallel rebuild right so there are two advantage here is one I can add one drive at the time right because I never actually carve out the space and I can also reveal all the drive at the same time from in part of the video so the rate is completely in software now there's no physical rate card yet so we have a definition of great and either our base be it's a single protection of work thing and we we allow the customer to choose that we call that rate five or eight six but and of the color is completely beautiful and even the rain with earlier like for example you would specify whether you wanted rate five eight plus 1 or twelve plus one now it's all automated right so we will decide the rate weight based on how many drives you are putting in and so on like so but you only the only choice that we provide you is do you want to eat my production to Andre tricks and everything else will take care of it as a rebalance as well when you add drives like oh you might actually deliberately change that rate level underlying but well we don't want it like we don't want to go and make dual protection that you choose to a single protection that we a problem what's your point like the surface area of each other we deal with the width of them can change right but if you choose to all potential we want to make sure that you have dual protection otherwise you're going to call me that Williams if you okay what's going on we actually mix great weights right so for example if you had like three terabyte drive and you had say for example eight by twelve plus one and we added 15 terabyte drives we might decide to change it to like range by four plus one for those 15 what's the impact if I'm then trying to design for capacity because if I know what my rail is going to be enormous capacity able to go out but it'll never make straight level or change over time how do i what we do is we have something called a unity sizer right so say when you're going into you say I want to buy 10 drives will tell you upfront or our algorithm how it calculates and will tell you how much capacity or expect just get from that okay and also we guide you to buy the right set of drive but if we try to enforce best practice also through the buying capability so we don't want we already tell you what is the optimal configuration right so yeah so the other advantage is that in the in the previous maybe used to have a hot spare sitting outside the spare is not built into the thing right so the legibility was saying so from a rebuild perspective it's now becomes faster right so drive fails now you're basically leveraging the power of multiple drives by then one no spare management so we talked about unified so let me talk a little bit of the file side switching gears a little bit so we have new I mean not new it's more than a year now but relatively new file system right case of 256 terabytes right so for the use cases that unity goes after transactional as Enterprise nails at the mid-range lower end through physics talking most customers more than adequate right if you need caleb multi petabytes that is one that's an X presentation right inline compression for available so take advantage of that native building policy-based positive shrink file system growth so you can set the policy and say hey my file system should always be 80% full as an example right anytime you delete files Y system automatically goes shrinks and goes to that call whatever you set and then again when you add more files it automatically grows right so automatic shrink automatic growing fully integrated with VMware we were one of the first vendors to support vai for NFS as an example so that's continues right so if you are using the via for example VDI horizon right or isin uses vai for NFS to go and the clones of VMs so we can do offload the cloning of the VMS on that on the file system by a multi-tenant mass servers so so these math server think of them as containers right so each of these containers have their own file systems logs maps or applications everything right so if you have a multi-tenant environment you could say hey after a 10 and 8 you this is dispatch servers yours can be dissin answer is yours right the G has presenters on separate IP addresses that's actually direct active record range so we can we can also have a same IP that's right and we can ever so this is a real multi-tenancy at the network stack level but these are other things that we did the new file system but the other advantage of unity that we move from we leverage also Lee notes for for example for the network we contribute in the Linux community for the network we use it advantage to have a completely separate network and doing the and then on top of that we use our containerized technology to leverage the network stack that leanness provides was they can you actually like multiple VLANs and have each of those running in a different villain we can try and be it yes we can try and different particularly as a as the encryption key and those sorts of things use there right which is yes we are yes so the encryption key so there is a difference so there is on at the disk level right a system like when I when you created will we generate a key for each is Drive we have an internal command but also we should portage there on humanity right when you remove the drive that creep is going to be destroyed so you cannot use it at that multi-tenant level for an as system virtual data system can have its own encryption key different than another tenant name system so in that case with the smb3 you can have increased SMB encryption from - oh right and then you exchange the encryption key over there so there is no concept of one key for a tenant so there does not have separate ease but if you're using smb3 as protocol for example system III it comes I'll encrypt it is at the protocol level in this case because miss Winkley works fireplace that's right so also three or four point yes we have separate independent routing tables right eight separate Active Directory domains right DNS or less what was the limit Allah holiday can you go to 96 or some from Nigeria based upon that so depends on the system the higher the model you go the bigger it is I think it's like up to 256 at the higher and yeah not server that is beautiful datum which I don't evening the most exciting the max I've seen is like 25 or 30 yes I got on the higher end right which limit is more than I or customer see and then the last piece of it which is becoming more popular is we have this capability to transparently clear files from the file system into into a public cloud or a private club like some customers don't want to move their data to public like a government customer is that integrators that using from iCloud gateway or something like that you have a cloud gateway service type of solution or cloud gate where we have our own solution we call it CTX all cheering appliance right in the virtual so that's what we use but it's policy based we set up the policy and you say hey any file that they not touched for the last six months or last three days right with these set of extensions over to the car lot like you have to do you bring a very good question I think it's integrating is not integrate the reason reason and I give you the reason why we think this solution in this case is a superior if you look at our tremendous install base in the DNS right we have an underdog causal system we have people use a CTA today with the ECS target or other target or cloud we actually maintain the same format of the stealth so everything is in the cloud you don't need to read or take to my grave so basically you save a lot of time just to do the migration so I can swap avian acts with unity like this again I mean I guess that that does make sense so that when n CPA as is a technology that is very mature so in support encryption ricci and support in line compression to the cloud so it's a very complex but very easy to use very very good cheering so it's not just a simple identity and up and again it's one basically quality bait and the second part of it is that eclipse transfer right so so when your application is going to read the file we leave a stuff behind and then stop automatically and then goes and go fetches the data from the cloud if needed right and we move to particular spot any any any cloud with SP protocol talking about cloud so let me talk about one another new capability building on that we call it forever snapshots right so so we have we have a great portfolio of backup products right great for customers the higher enterprise higher end of the mid-range we want to use that you can use that but pay for customers who are at our lower end of the mid-range buying it ready $5,000 box if they don't want to buy a exclusive data domain solution as an example they could potentially use this as their backup the solution so so that's three parts to the solutions one you can use regular snapshots right you familiar with this on your local you can replicate the snapshots to a remote system write anything and the remote doesn't have to be all slash it could be a hybrid it could be obviously potentially depending on what your use cases if you're just planning to archive the snapshots there could be a smaller system if you have many use it as a dr then then of course we would recommend that it's a similar sized system right and then and then you can say hey let's keep it for seven days let's keep it here for seven weeks and weeks whatever you want and then you can take the snapshots and then also move it to a slab and three gloves right and there you can keep it for even long now I know some of our computations also have these type of similar solutions but there's some important differences I want to call them so one basically we use s3 Amazon s3 right versus some of the other solutions which uses ec2 EBS now you understand when you are using easy to EBS those are much more expensive because you have a paper to compute the VBS is more expensive compared to x3 right plus the other problem with those solutions is that there is a limit how much data you can move right typically there is a VM running on it so the VM pay has a tender of item here it is potentially there is no limit right whenever I'm moving the snap out there I do not I'm no longer governed by the limits of my system so that's why I call it forever snaps because with this step of the solution you can you can never have to delete a snap that any s3 compatible system not just Amazon so that's correct so we support this do we support Amazon we we are also looking to support put we also support work Ustream and we are also looking to support ECF it is our internal don't do with us yet the easiest one is coming for this one we support it for the file we support it just for the file this is a block photos it was not Stanley we extend the hearing capability we insert our API in the block stack so we can also snap okay so that's the do we imagine to do any other object stores like Swift or maybe blob storage start right now all right we are compatible with all GSP basically the snaps are fully rewritable fully redress yes you can have a snap reader right you can have a snap and you can also have sync own that we do not speak about this so the naps are read writable he local snaps rewritable out here you can bail over here you can restore it for the cloud of course you have to regenerate before you use it you cannot directly go if there is no computer out there now so this is one use case for snapshots like for backup which is typically the use case of another to use it this isn't a comparatively new uses ten clones right so what is the thin clone so things own is basically a snap but it actually behaves like a lot right it's a it's a first-class citizen so so so it's sharing the same data set but because it's a long you can set up its own snapshot schedule you can set up its own replication schedule and so on so you could say hey my primary production line I'm editing to a my thin clone I might say not to replicate it I might have to get it to be why is that important because typically the use case for they say you are doing some tests that you're doing some analytics right and you're going to write data to it you're doing some testing so you can modify this one is being backed up you may want not want to back it up or you may want to back it up to another place another location etc with a different schedule not at the same schedule as this one is right so that gives you greater flexibility so the QoS or the snapshot level really applies to swing clones that you can do it for snapshot as well as in clones oh really yes right you see independence the pink ponies an elevated snap with that when you allow you to have a simply present himself as a one right so they thank us understand only orchard sold for it for the us doesn't matter is an object so we can apply that there and we applied to revolt as well so yeah not only just last right so anyway so we have independent here where that's very important now you might and like us important because we are the snap Anglo's are sharing the same data center and you're running two different applications on incorrect we want to make sure eggs my desk guy is not impacted my production sequel database as an example that's why yes now up if that's still a problem what I show with my nap shot replication you could do that as well right so if it's still a problem putting it on the same system and you want to do some testing heavy duty testing you could take that naps are replicated to another system and then run the testing test they have analytics on that remote system as well so you have both options to bring it local with the us taking it your remote and running it there as well okay ah we you can manage all of this from unit sphere which is our the management interface or you could use appsync if you care about application consistent snapshots right pin clones say if you want to do a testing of the sequel database you could use that thing to do an application consistent snapshots in clone whatever you want to create that all available through some sort of API you all everything is available in REST API so we have a robust REST API so everything that I'm showing you here you can run it through REST API so actually the recipe is the core of so when we create unity we decide to create the REST API as our way to manage the things also internally so everything that you see in the UI is the call to the recipe I so everything you can do in the UI you can only see a lie and you can do investopedia all right so and then a couple of high-level usability features and an outward of the demo so we have built in integrated migration now yet Delhi FC provides whole bunch of migration tools each one of them solves a different problem right this particular the use case for this is basically simplicity right it's built in into the product you pull up unit sphere and iterate discovers the arrays discover the launch disk of the file systems you right-click insects move them over and does it point three easy simple right you can do file it can do block right now it only supports vnx and we are looking to extend it to other arrays as well right so so really to a question unity has been on the market for about a year right yeah so what is the adoption rates with customers which have beginnings before are they moving massively to unity or are they moving some other billions of technologies so that's great question so yeah so that's the reason why we sort of prioritized vnx migration right by Andrea next we can replicate with the vnx we can migrate data from the vnx so yes if the customer is running sort of at the end vnx tools are still have a life like gotta be nx-1 for example which are sort of the end of their support contract yes we are they are sort of moving as I said there is a massive migration obviously from where bnx one unity right because as you see the numbers from at the beginning of the presentation the install base of the unit is already growing but these four days or bnx one it says three or four times that so customer day wants an easy way to migrate and by the way import file this is a completely transparent we create user impact and technology that can be as a caution was explained in the future we look to stand the speed also third-party and then we can do that because we use a proxy so to just expand on that so for the file migration views this concept of Nan service which i showed you so this container right so VNX as a concept of this containers we call that vdm you can take the container as it is with the pulses and with the snaps with the IP addresses and then move the container to the same thing so that's why transferring because I can move the file handle I can move the pipe I like the addresses keep everything online right there's a host ever remember this is not an easy things because we change the file system so we're going from an old file system to a new file system so we create and patent the technology that allow us to do proxy and we keep the mirror so you can always go back something goes wrong you can always automatically go back to the old one and then one of the sort of new addition to our energy drain covers we launched it end of last year it's loud right this is uh this is really cool because it allows you to manage multiple arrays from a single pane of glass right now you could say we used to do that earlier with unity essentials CCC and SRM can't do that the difference is this one is sitting in the club so you don't have to set it up on a server you don't have to set up a server you don't have to set a VPN access all that stuff is sitting in the cloud you just log into our Delhi MC portal while your username and we know we we know you are part of this customers customer we know where your data centers are where the unity is located you could have some in Europe you could have some in out here will show all of those arrays together single pane of glass right in in counting where where is that data stored and can you regionalize that data yes or say German customers kind of thing yeah I mean we are a we are a these tool the other cool things about this that's completely a software service so we have a very very quickly iteration of release up there so one of the things we are working now is to regionalize so we can partition now like for a big customer we can say ok you look only at a pj you can looking at all the at these things there is no real dinner smally metadata so but for a particular region we are looking to obviously have the locating the same region ok energy timescales oven huh any time scales on that yes I think you guys here we are working on them right no so the other thing I just want to highlight one basically we move metadata the second part is that we have been doing columns for many years even be an expert before that so that areas were always sending back metadata back to the EMC before right but before even this theme all we did was just take the database and provide a window to that but we were anyways collecting the data now it just gives you access now you can go back and see what data we're collecting right and Ray is going to go go demo of this to get our key and just quickly summarizing the cloud like you the key part of it but we do have multiple ways through which you can manage it right we have the html5 Unisphere as I said we sub-site integrations with bigger VII for NFS or block REST API that's the base of our unit sphere right everything that you can do from the GUI you can do it for the rest api as Nemec support SMI support right and then very robust of in tech support as well so that's all building it

Channel: Tech Field Day

Views: 25,260

Rating: 4.5229359 out of 5

Keywords: Tech Field Day, TFD, Storage Field Day, SFD, Storage Field Day 13, SFD13, Dell EMC, Dell EMC Unity, Wei Chen

Id: oJ9T1lJf538

Length: 56min 18sec (3378 seconds)

Published: Mon Jun 19 2017