Dell EMC Midrange Portfolio Review with Kaushik Ghosh

so this afternoon so good afternoon by the way shall some caution coach up in Pierre Lucas team like like a spear Lucas Adil I'm responsible for the unity and the SC the roadmap the strategy and so we're going to talk today but we're going to talk today only about the SC like belugas head wet at stroyed field 813 we did a review of unity in depth so today we're going to talk about the SC now for those of you who were here 2010 you will see some of the feature functionality we talk about today is sort of similar that's not a bad thing because they'd be that basically tells you that some of these features are pretty mature unlike unlike what our competition is offering right now I mean these products have been their tried tested and and basically we have basically worked on top of it providing additional new capabilities right so so so that's on without like you would go there so with with that so this is basically what our mid-range portfolio has now there are other products that we have on an N minus one products that we still sell but these are our lead horses in the in the in the mid-range space right so that's basically the unity line and then the SC which is the Compellent the formerly the component line right I don't know if you saw the announcement yesterday we launched a brand new SC all-flash right not a surprise given where the industry is going with all flash pretty much all of the rest of the daily AMC portfolio already had all flash flavors right unity launched when we launched we brought unity all flash Isilon has all flash v-max is all flash extrema is already all flash so it was just a matter of time when we had to come out with the SEL flash right so that sort of completes our all flash portfolio now everything is has an all flash flavor and and adds that flash any different to the SC that you've been selling with flash in it before or is it just a gardener tick box that you can't that it's been in this kit oh good somebody else complained about so so yes right now from a hardware perspective it is something similar Robert from a packaging perspective and I'll you all talk about it is different right so for example one of the things that all flash customers have been asking about is we want everything included I don't want to buy separate software separately so you'll see that included you'll also see from a software capability promotion perspective like for example when you buy and unity all flash you get a free cloud with with it right and we're going to offer the similar capabilities on the SEL flash when you buy it you get a certain amount of capacity free on our Dell EMC cloud operated by virtustream right so so yeah so so that sort of it's a good segue because of what I wanted to tell you is that even though we have two products in the mid-range there are some one basically there are some unique capabilities in each one of them I'll talk about a little bit about what's unique among each of these capabilities but so we're not going to get you're not going to replace one with the other right so that's that's something I wouldn't want to make sure everyone understands we each one has their unique capabilities so we can continue to offer continue to invest in both products right for how long or till the till we distill we end up service like these products right end up like these products so so till the life of these products we are going to continue to invest continue to offer these two products but where we are going to bring things together is is in the ecosystem right so for example from a sizing perspective earlier when when when the companies merge we are two separate sizing tools now it's all merged into one right when last year on the Delian see you on the unity side we are introduced cloud IQ or cloud-based monitoring now that's cloud IQ is now supported on the SE so you have one single pane of glass through which you can monitor both mid-range platforms right we have recover point based replication between between unity and SC right we have power path support reflex support right so on the ecosystem app sync support is coming soon right to be on the ecosystem perspective you see all of these things coming together okay in the cover point built into both platforms or external appliance so the splitter is built in the the the appliance is still external that you do need yes right we offer a recover point for VM also yes yeah that is correct yes yeah once you no one's more impressive than the other yeah yeah so your opinion it's good well my opinion about which one yes another thing to Kaushik so we build this ecosystem but now we're gonna change also the experience on the products so someone was asking we're gonna give you a preview of for example the next Unisphere so we bring in unit sphere also on the SC line so we're gonna give you a preview of that that's very important because the look and feel will be the difference but obviously the product are different so we don't want to replace with the same unit sphere it's a look and feel that the customer can experience and the other part that cost was pointing out clouding you really broader this ecosystem with the proactive alt so we can have a proactive alt in the cloud and I automatically tell you which action you need to take who we have an alt score to make that score better on the SE or on the unity so that's allowed the customer to actually use as we want this system in the right use case right so we we have a bunch of different customers that they used both right so that's very important actually yeah thanks so quickly should just give you a quick review of the business before I go and deep dive into the technology this is just from May last year that's the time when we launched the unity platform as you can see we have done really well if you look at just unity and SC it's 2.3 billion dollars 35,000 systems have been shipped since that time frame right with 20,000 plus customers so so a pretty solid solid portfolio that we have right now right as you can see a lot of customers lot of enterprise level customers also not just any right so it's a solid portfolio that we have right now with these two products right wait okay so both products that mid-range products so so what that means is from an architecture perspective at a very high level 10,000 feet level it looks similar like it they are both dual controllers both both servers block i/o extremely well very reliable products rich data services and so on but if you'd go a little deeper there are some unique capabilities in each of these products right so for example on the unity line it's all in line efficiency right inline compression zero detect deduplication coming soon right so everything is in line se on the other hand we use we call it intelligent efficiency where we first acknowledge the i/o and I'll walk you through how it works we acknowledge the i/o first and then our process kicks in where we compress and dedupe it right now the advantage of this is that one basically this technology works not only with all flash but you can also work with hybrid and also the fact because it does after the fact we can get you a much much deeper much better efficiency then I would get saved typically with an inline equation right so so that's a that's a key differentiation advantage on the on the SC side the IC is also federated so even though it's a dual controller machine you can start with one and then you can add multiple machines and manage it all together with the Dell Storage Manager as one one yeah one unified cluster and then be able to move the Lunt's around which is what we are going to do demo today one one lives on one array right that is correct that is correct absolutely so so that's the Federation side of it unity on the other hand is unified right so you can have five you can have law you can have evolves everything together there's also a software-defined option so you can buy it as a appliance you can buy that the software-defined or you could buy it as a converge to a Vav block that we offer and on that comment right it's one land yes but again this is mid-range right so oh no no we want to use the portfolio right you you listen about v-max extrema yo and so that's why it's rid of it I know that I want to modify I understand just just remember we're just trying to clarify we're not about yes so yes when we are making value judgments we'll let you know absolutely and do it didn't know me well enough man yeah I know and then moving on the the the Union has one of the key things that we have when we met last time we talked a lot about is this hybrid cloud integration right customers going with cloud decide trying to decide should I go with cloud should I go with on-prem or what we are saying is no you don't have to make that decision you can do both right and we offered the that not only the technology and what I'm going to show you later is we also offer you that cloud part of it free today with virtual stream to customer so you get our so you get the cloud you get the on Prem we have the tools that integrate so you don't have to have make a decision between this or that it's an end and conversation ok are we going to come back to the Federation yes we will talk a lot about the Federation we gonna have a also live demo yeah and then there's the demo that we're going to do right and then on the SC side one of the value prop is on the economic side so so so just from a pricing perspective we have a very attractive price the d loop and if compression gives you on top of it lot more efficiency as I said the deep end compression works not only for all flash right which is what most of our competitors does but it can also work for hybrid because of this technology right many of our competitors only off for data compression on all flash there are issues with reading deduplicated data from spinning discs because there is IAP amplification inherent right well there you sure what happens is when you're doing in line right see we'll have near lines as behind it right Iowa's coming in you have to quickly process it send it back down right if your disk is slow then that automatically creates about that's right that's less of a problem because I can address that with a log in a right and free space file system it's when I try and sequentially read data that's been stored on deduplicated spinning disks yeah but then I make one one megabyte IO request but it's not one I up to hit the desk right and it's fine for colder data and all right and a lot of vendors are just ignore and that's why how we position a right so you don't want to put and that's where when we're gonna go in the deep dive of the architecture you're gonna see how Compellent understand that skewing of the data yeah they decay most is that selective can I apply it to some ones and not yes absolutely you can't create defy then I like it yes and most other competitors avoid dupe to spinning discs because the occasional customer will do stupid fingers and get lousy performance that is right and we also offer you flexibility in the sense like one basically you can turn it off if you do turn it on you can decide do I want only compression you could decide like this is not be duplicate workload so I'll just use compression which lot of our customers do and then you could turn on both compression and you right so they you get that flexibility all right so before I go into it there's one one last thing I just want to highlight so this is also something that came out as part of the announcement so apart from the awesome product that we have we also came out with our with our awesome program to go around with it right offering some capabilities that's pretty unique today in the industry or for the maybe first time that we are offering it so the first class to quickly go over these so the first one is the eyes a three hour satisfaction guarantee so the important thing I want to highlight is a three-year part that's no one in the industry to do does a three-year satisfaction guarantee right so so it's three years if the product doesn't meet the specs that we say it it should then we will either fix it we will repair it or if none of those things work we'll give you money back the one is easy all flash storage efficiency guarantee this has been there but we sort of rolled it into the into the all flash portfolio now so it's a forty one all flash storage efficiency guarantee no assessments needed so this is basically will give you guarantee pretty much any customer who who wants this guarantee right and of course our what we have seen in house you can actually depending on your workloads we can actually do much more than than this four to one so and this is just the baseline game yes there are so obvious now set out what are they so what of it so it that's include the Finn snap and and and also obviously D the cavities the workload need to be compressible and de doop OH yes so remember it's the Guerra I mean let me know I know I know I know I know why you asked this so we we are actually looking to because we are above for the fourth one actually guarantee is across different products is not only on the mid-range product right so that's why we taking a step back and we say it's four to one for everybody right so we need to be sure that it's four to one across the board I don't blame you for doing it I would want to give no waste but but we are we are looking to actually look into how we can make this specific to the technology that we use right yeah now finally think about that is that there's a difference between an Porter one from an efficiency guarantee perspective and another for when you're actually quoting systems out and sizing them for customers yeah so we're not talking apples to apples and those things but so you're kind of seeing two different things the customers need no actually for for the use case where the customers say you know especially we have large customer that they know their data set we run actually a tool to estimate the things like all other ones they doing and then we stay in behind that guarantee that is specific for that work yeah yeah with banged other vendors on the same exact thing just as hard so moving on equal opportunity moving on one of the other things that we also make a guarantee on or make a commitment on is is this something called never worry data migration right so you see me talk about it in this presentation today on our as C line as well as on our unity line we have built in tools to migrate data in some case online from our previous generation so if you are about vnx your customer e of vnx array you can move the data transparently in some cases into the unity likewise if you have a ecologic you can take data from there and move it into into the into the compelling and what we are saying is that moving forward when we come up with our next generation product line we offer the similar capabilities and everything online they want so if you have a SC series product or if you have a unity series product our next generation you will be able to do an online migrate from from those two products to this next generation products so that's basically the commitments it's not just n minus one to today but even into the future we will be offering you these online migrations how do you do the past SC migration we'd offer PS to SC migration I'll talk about that in in the in the presentation today and it's not just PS we also offer MD we also for a whole bunch of other products as well so I'll talk about that when the the fourth fourth pieces are hardware investment protection so this is just saying that hey if you have an existing WMC product you can trade it back in we will get a credit when you create it back in and and and buy buy a new unity or a new SC SC product so and so not only we will be able to migrate the data will also give you a credit for that hardware that you already have if you trade it back into us yeah all-inclusive so that's up on the unity side we are we were all-inclusive now with the SC now it's every is also becomes no all-inclusive so when you buy the product all software that I that you will show you is all included plus not only that in future also when we release any software those will also be all included right so you never have to worry about buying a piece of software in future and all flash versions of Si yeah so right now the all-inclusive on SC is only for all flash on unity we do offer it for hybrid and all flash right so it's a and we've so we'll definitely take a look into on the hybrid side as well and then the land yes on the hybrid side are you continuing what Compellent used to do and not making me rebuy the software when I upgrade absolutely so the thing that you're talking about is called persistent license persistent license so so yes on the hybrid that's another word that's actually one of the reasons boy why we haven't offered the all-inclusive yet because if someone has bought persistent license you might say hey why is there's some challenges there let's put it that way it ends up raising the cost for the existing estimate that that is correct right so that's the reason I didn't go into it into it previously that's the reason we don't have it on the hybrids yet but we're looking to figure out how yeah but that from Compellent that was really the first of the the trend that we're now in with things like Pure's evergreen where the buying cycle is just completely different absolutely so and then the last piece of it not least is what we introduced it recently with the unity is we we have Delhi MCS one of the only companies which offers both and in in on-prem as well as a public cloud solution right through watch history so so we are now if you buy any unity all flash as an example we will give you 20% of that capacity for free on virtustream cloud right right now this is only available on the unity all flat we will be offering this on the SCO flash of beginning of next year ok an inverter stream that's storage to run VMs on that object storage right no this is the object storage this is the a storage so the use cases basically you run your primary workloads on the oil flash and then any data which is like snapshots which are old you don't want to keep it around you will just move it into the s3 and that's your backup tier right if anything goes wrong with the there is a dr there you get a new array you get all the data back no that is correct yes that is correct so we we use a technology that it's a virtual appliance that is a CTA so we can tear file but we can also tear the block snap and you can actually one of the the the most interesting use case for our customer today some big customer that for example the CTA also tear to ECS our local object and also with the IBM cloud or safe one of the most interesting use case we have customer the using cts since a long time what a city a stand for cloud ceiling appliance so what is happening in that case we use this top technology so you can basically retire your old storage and bring in the new storage without read rotating the data but that's you just move this stuff that they are four K blocks so that's that's an advantage in in that case which technology is that that's easier where I come from it's come from old EMC is called D HSM is a protocol that we invented the 20 years ago file mover the things here that we find out looking at the at the technology that we can extend that to block CTA original was only file so we are extending that to the SC 4 block and we extended for unity so that is free because it's included you don't need to support H a and you can go in this case as Kashi was explained via to stream but it support Amazon and all the other cloud right so the technology supports all clouds what we are going to offer you as part of the promotion is our own what you stream it was 20% of what this facet about is that was great but our 20% of the raw capacity so you bite hundred terabyte of unity all flash you get 20 terabyte of virtuous stream cloud unity all flash only that is currently on the unity all flash DSC all flash we are going to offer it like beginning next year and you to connect this qualified did you say that you if you it's only 20% of the capacity of the box but you said you can re instance your array if you lose it from that copy that is correct yeah yeah you need to buy you need to buy those free water with the spring bars the razor handle certainly comes with blade that is that forever or so the cloud reuse let me clarify so yes so 20% is what we offer the 20% is also include it is the lowest year of virtustream sorry that's the one we offer for free so of course you there if you want to pay if you want to buy more capacity you say picking 20% of the slowest tier as a free giveaway is just perfectly giving away the first lollipop to make customers hungry to buy more thank you one year sorry well it's only one year at you buy this one year so the day you buy the array you get a click from that day one year you get it 20% capacity for free if you want to keep it for more than after that one year you have to pay for it if you want more capacity have to pay for it and and if you want to go for a higher tier Varane capacity then that's extra as well I would have preferred to see it the term of the support agreement but whatever yeah but but if I've got a 20 terabyte box yeah and I want to use that as my dr solution because that's what you implied i could do if the box died yeah you're going to give me 20% of 20 terabytes I still got to buy a lot more than that in order to reinstate box because I need at least a single copy of that entire box in order to reinstate it I can't do it from snapshots did that is correct so there a couple of things one basically you are not going to fill up your capacity day one right so as use of the cloud you can basically keep it behind because you can tree in stance a box that I haven't got all the data fro in the first place I can't do it from the snow and they want the whole thing you have to go so for example agree explain the situation so you you bought a trendy terabyte box in your example right off which say for example on Safeway for example five terabyte you have filled up the remaining is not filled up right so you only need to protect that 5 terabyte right that 5 terabyte data you need to you need to just make sure your box right sorry you're only supposed to use a fifth of your box right there is another so the idea is that if I've got an empty box when I first buy it but yeah then you're protecting or if I only got 5% of my box filled when I buy it then leave the lab any virus on it that you get three three months of any of that so it's a little bit different actually sorry because also we have a compression CT a support compression right in line so there is more space there obviously we not say that you can protect the entire thing now at the beginning you're gonna protect the entire thing I mean it's obviously that we we need also to make money at some point on the on what we offering the envy there's nothing wrong with that at the time I just wanted to make sure it was clear that yeah I'm not getting a free backup dr service or effectively yeah I'm gonna be paying for it's in some form yes and the other thing is we backup that HTC this with backup the snapshots and LAN data so it's not like every every date on the box is backed up so it's not an image mirror image of that thing so so so this situation if there's a flood and whatever reason you lose the box you do have to get a new box you have to set it up everything so those things you still have to do which is in the lawn in the snapshot there's the data you can just restore basically from the from the from the cloud I have no problem with the offering right um I would say of all the six things are put up on that board that one is a little misleading it's cuz it's it did when I look at that it looks like I'm getting that as part of my box and that means I'm gonna get it I'm gonna have it and hold and love as long as I have the box that's not the case it's a one-year promotional you know sample of your dreams would you get it or not really getting it and so you know if I if I start using it and it becomes part of my process I'm gonna have to start paying for more later of which I have no idea what that's gonna cost me because I don't know what it is neither do you and it's fair because there's a cost to it you guys and I'm not saying guys if you give it anything away right I just need to be a little bit more more clear yes III think I think is a good feedback so we will bring back no problem with the offering yeah sure so anyway so that's that moving on so again like as I said we introduce the SEL flash right great performance right this is this is of course from a singular eight now this is the best city scenario with read-only right now if you Fed rate that of course you can get much more but as long as you've distributed workloads that's not not one one being shared so as long as he distributed you can say allow federated and not scale out exactly absolutely and then similarly from a capacity perspective you can do that and I'll show you how we do load balancing rebalancing we call something called volume advisor which does intelligent rebalancing across this Federation right so so we're going to talk about it and then we're going to do a density it's going to do and that's one of the part where it's very different from the competition we're gonna show you the volume advisor you can do different things is not only performance or capacity you can decide what you're doing and there is an intelligent that spread across the multiple system right so if a particular area is getting filled up basic say for example you say based on capacity it fills up 80% you can set the threshold automatically you'll get an alert and say okay this is getting filled up here is another refrigeration that is not used as much I would recommend you move the LAN out there so if you say yep you agree with it you just click the yes and then it automatically moves that learnt transparent your application to that particular array in the Federation this is fabric analyzed kazi correct this is yes by which analyzing everything you support that we're gonna show that yeah dynamic so that's basically on there's a capability of this ability to move lawns around for different use cases right whether it's whether it's for load balancing whether it's for migration whether you're adding a numeric array right or even 4dr right so different use cases we can solve the disc or technology the this this this mobility use case and we could talk about it and then the efficiency part of it right so like as I said the guarantee when we're talking about is just a baseline depending on your workload you can get anything from a 3 to 10 X data reduction from just from d druvan compression right cloud based analytics we talked about cloud IQ and then have also frank related we talked about it earlier so when we talked about all softer I just wanted to give you a flavor of the different types of effort this is not a comprehensive list just to show you some of the key software that's included now with the box when you buy the box all of this is included so D Dubin compression or that's the included instant replay that's what we call our snap technology snap and replication technology that's included right we include data at rest encryption you still have to buy the key manager but the encryption part of it is included we also offer application protection application consistency we call it the replay manager right so with that you can take application consistent backups cloud IQ analytics cloud IQ that'll also be included with the account live volume with auto failover so that's our dr technology where you can move from A to B automatically failover into a little bit 9 back to that right yes we are going to talk about each some of these capabilities in depth right we're going to talk about this but this and we can we talk about the Federation yeah yes q is not we can talk about it but we do I didn't have a slide specifically but if there's a specific question we can definitely address it or answer it demo is a different part is it a throttle or is it more than that the minimum max so this is basically minimum right now right in the sense like the remote is basically I don't know what it's not a guarantee that's what you that's what you're asking no no so the the entry stakes for QoS yeah is to say don't allow this LUN to exceed X IAP startle that is right why does that hey so we can do i ops we can do by in wit latency right most systems go to q SRO and so if you are throttle that's a step right there are more sophisticated methods beyond that but right just we're actually we're actually gonna come on we're actually gonna leverage QoS to inject latency into one of the demos so that volume adviser sees that hey we've got latency that just escalated for one of the one of the applications that we have and then volume advisor is going to make a recommendation in the Federation move that application seamlessly so you'll actually get to see QoS as part of that a lot of them that's right and it absolutely is that I ops limit that I define yeah that's gonna drive while you can see okay okay cool uh yes one recommendation for the next Rev mm-hm have the throttle only apply when the system is stressed oh I understand so if you reach certain when the system is only running at 15% of capability yeah if I'm if I'm a solution provider I always want to say that guy gets a thousand I ops because he's paying for a thousand items yeah if I'm enterprise guy I want him to have ten thousand I ops as long as there isn't anybody else he's stealing them from yeah he's in noisy neighborhood and he can be as noise whatnot but give them both I know I don't want to ever see ten thousand because then when he gets his one thousand is gonna complain depends it depends very much on the application but in the interface yeah as you know it's here's the threshold you know here's the throttle absolutely and then say you know only when overall latency exceeds this yeah should those throttles be applied alright it's good that we have disagreement the option whether it depends who the customer isn't yeah no absolutely I mean Asian is you know sometimes it's yeah like I have a low priority application I want to run as long as it as fast as it can as long as it doesn't bother anybody else you're gonna seem that further in the in the federation one of the other things is look at the CPU consumption right that's another way to do it yeah it's and there is that logic already in the in the volume advisor right and you have to hook the two together absolutely alright and then live migrate we're going to talk about it volume advisor we talked about it will show you again as a demo and we will talk enough about it and then federated management so this is basically tell this is basically our Dell storage manager today which we are going to be Brent Unisphere in shortly and that's basically ability to manage multiple arrays from a single pane of glass right and that's where you will see this capabilities like volume advisor is sitting today right you see all these arrays you monitor all these arrays and then decide how you want to rebalance your work all right great reviews we just had a we had a we had a storage review did an article big just beginning of this year very very impressed talk about SC but or this is just one article you can see 20,000 but customers already bought it so I that's more valuable than this right so but but anyway we like I would like this story about SC product how many of those 20,000 customers have deduplication to known I think about so the last time I look I think was about 30 to 40 percent mmm obviously we have a lot of customer with you need to be certain level of code we announced that in 7.0 I thing was so is about one year enough so was about 30 percent 30 to 35 percent I can't give you the so so this is a showing you the full lineup of the SE the full family right so these are the two new children like there are 50 20 F the 70 20 F that's the whole flash family we just announced it yesterday yesterday and then this is basically our hybrid family the v3 is the newest addition to the family but what you see is that basically since the time we merged pretty much everything at cross this family was refreshed right the 70 20 came out immediately at the time of the merger that the 20 came the v3 came right and then now the 50 50 50 20 f and the 7020 F right so pretty much the entire component lineup got refreshed right and that's the hardware just quick highlight some of the capabilities so one we talked about so now across the family pretty much all of these areas of the same capabilities right in the past if you remember the SAV two did not have all of these capabilities right so that's sort of we we sort of was causing issues so we fixed that so apart from deduplication which is not there on the v3 just because of the memory footprint and all that stuff pretty much everything else is basically available across the board so if you want a v3 put it in a branch office and you want to put a dual 50 20 F in your main office and then a 9000 as a backup you can do all of that and and replicate between all of them and you can federate across all of them and so on all right I may have missed the point before but is there any architectural difference between your hybrid and all flash products are you using the same technology to address the way flash memory is being addressed so from a hardware perspective yes yes so from hardware perspective there are there is no change between the two products right this is the same hardware from the software perspective exit from the software perspective most of the capabilities are same because we do want it to be able to replicate we won't want it to federate among the whole slash hybrid so most of the capabilities I said where you see changes coming in future not right now is that we might start putting in more optimization right so if you are not doing enough like hearing to the two hybrid which is what we have done on the empty side also you'll see the code starts starts diverging at some point in time in the field so what's the difference between a 70 20 and a 70 20 F if I decide to go BER 70 20 which is filled with flash drives so one right now from a hardware perspective it is the same in the software perspective right now that being functionality is same but only thing is like as I said from a packaging in perspective the 70 20 you have to buy all the software separately with the 70 20 F all the software ok last then you'll see like for example from a cloud perspective when we offered the free cloud right we will be offering that free crowd only on the F family this one will offer the capability but the promotion is available only on the S so right now as of today with your flash units versus from a software perspective is that from the hybrid that's been around for a long time and the oil flash there's been no specific data path code optimizations to this point well that's where I'm heading on so that you ask the right question right it's if I have flash the system do we treat flash in a different way that's I think one what you guys obviously there's been a ton of evolution in that place the code that we ran is the same but we we already have people that they bind these full flags right so for example the page sites that we're gonna speak about we optimize the page size for flash okay it's 4k the other things on the eye brain if you use an eye breed you remember was very hot at some point the fast track capability putting the things on the on the edge of the other store this clatter that is there only for ivory we have a technology now that we can expand and shrink a red group and we try to do also when is all flash we try to optimize the page sides and the caching for that thing so there is changing since 6.34 autumn eyes the system for flash now if you fill this this system with cool flash obviously you get that optimization you are in an ideal situation for example we suggest two magic pages the caching has been modified to be more more compliant with the with the flash right the flash needs a lot of I don't need to spend time in cash sometimes with with with with the flash I'm doing for example stripe right and all these things right and that's the work that we went in the code now if you look at the to your point right there is evolution in the code obviously is not and also we can the other things that we did we can differentiate from writing test if a reading test drive and we can use them in the system to automate our cost yeah well I mean I go back in my memory to an article I think you wrote a long long time ago or a conversation maybe about you know built from the ground up flash versus changed to flash later yeah kind of stuff and we've had these conversations over and over again with lots of different vendors right and we know which vendors we talk about Compellent Compellent first of all started yeah in the hybrid in a better position you know and frankly it's you know the difference between a 70 20 and a 70 20 F is packaging and making Joe Unsworth happy because because if you buy a 70 20 full of SSDs Gartner doesn't count that as an all flash' right and if you buy us a 27 27 to 20 F they do can't do it no I understand that and it's not just gardener I just want to address it there we have had RFPs which would say that if this is not certified as an all flash I'm not going to well that's the only latency perspective right yeah and that that's my mind leaper survey like we can do as many I absolutely want but at what Layton sees are we looking at yeah yeah so when we get to the technology and the the architecture you're gonna see why the latency is so good on these balls that's what we gonna add up what what pellagra was saying about the all the flash optimization that we have done so so from the days of the hybrid we put in all these optimization choice we had to make that do I make this optimization available only on the F or not doing it on the F and that decision we did was like let's not put artificial restrictions so the code so that's what we like was saying that if you take the 70 20 and fill it up with also action you will get that same optimal that optimization that you have on the F side it's also it was a decision we were we had to make should I restrict it or should I not restrict it and and we sort of went with the flexibility at this point in time well we thank you for that it's also important that we've reached the point where having come from a hybrid has advantages yeah because some of the guys who started as all flash arrays with one tier of storage yeah don't have the logic for right intensive disks and Reed intensive disks which will reduce the cost and next year don't have the logic for nvme yeah yes salutely right and so and that's a very good point right and I mean we don't have a time to discuss this we can discuss later but I think you touch a point that is very important using multiple media and multiple media type in the near future will become very violent who is gonna win he's gonna win who knows how to build this thing I've started I started talking about solid-state hybrids yes where there's multiple media type but there but not I use I believe my presentation like hybrid and they say no you cannot say that right but we did the intelligent technology of place the data in the right place is gonna become pretty another big things in the near future yeah so it's always a circle coming back yeah what we're already seeing from at least one competitor a system that has two tiers of storage and no logic for assigning data to which tier yeah alright so let more point on previous I have a point for VM is that you're now included with the old flash so record point for VM is we have two bundles basically that is that is included right so for one for the small and one for the large yes okay so just showing you so this is again like you can see the hardware here I mean so there are two it's two controllers so this is basically one of the controllers taken out right to controller bombed right is it heavy what if we're tough all right here so two controllers depending on that this particular one is the SCV three so it doesn't have these additional aisles so it has only one extra additional i/o modules on the 7020 you will have three of them right eye modules we also have built-in i/o modules right so so built-in 10 gig I immortals also all 12 gigs ass back and then I'm gig based T no no yeah a sleeper board both basically risible optical right and then these are the expansion pci slots low on the 7020 you have three of them for SP on the 3000 and the 5000 you just have one then you can see it out there now that's on the hardware side now from a software perspective actually that to cue you capabilities all sort of to architecture that we already talked about was it so one is basically this what nice page pool right this entire abstraction of how I lay out my data set and some of the capabilities that that is enabled by this technology and then this concept of this virtualized front end right which is basically the core capability behind behind things like live migrate the live volume and so on right so the virtualized front end and the virtualized page pool those are the key architectural core basically of this to this product enabling multiple favors and we're going to talk about each of these in depth so let's start with the virtualized page pool so the virtualized page pull offers these technologies so one basically this is basically the keyring technology right ability to move the data around right efficient placement of data at the right level at the right ear right now you might be wondering if it's all flash how does that matter it still does matter in the old flash as well and I'll walk you through that right one app one of my athletes about Compellent historically has been the schedule on which you do progression used to be relatively coarse grained how frequently are you moving data now it is there is a by default it's once a day but you can move as much as you want really its demand you can do it on demand it can do it automatically so I could tell it to do it continuously right it is on demand it is automatic and it is scheduled so you can do all three and we take advantage of all three right snapshots is built into the product and I will talk about it and it's it gives you a lot of flexibility and I'll show you like you can move snapshots between A to B to C you can move o everything around whenever you move a long arrow moving a lot from one place to another snapshots automatically go with it right so so a lot of flexibility gives you many use cases opens up many use cases with this technology and lastly but not least is basically this is also the core of our D depend compression logic as well and the the D dependent compression it is post process that is correct which brings me back to the same and and what are my scheduling options for that because so as I said you can you can get yourself and troubling with post-process dee doop on primary storage and capacity planning because you rights be boom you right you you told all that bought one second where we're going in the architecture you're gonna understand there is an elegant way that s/he use to define what is that even not active so we have at page and read only page so we're gonna explain that okay okay I can wait I'm almost yes it's two more slides and I will walk you through that yeah thing all right so let me walk you through the AMIA in fact we can actually show you exactly how it works so this is basically an all slash system right I'm not even talking about hybrid right this is an all slash right these are my all flash drives right we create this virtualized page pool so one first thing different you'll notice is this page pool it does not map everything on the drives right it basically is very it that's why we call it dynamic right it it allocates only what it needs right now and then as as you write as you need it expands and it also shrinks right so that's one thing the other thing you'll notice is that we stripe across all the drives right we try to maximize it so so this part is raid 10 this is raid 5 we we will stripe it across all platforms so that we sort of maximize the eye ops that I can get out of this thing so so this is basically your initial state you said hey let me create the screen who will create enough through for now right rest of it is unallocated at this point new IO comes in writes comment they immediately go to this raid tent here right so that gives you the fastest response time right like the fastest operation basically the best response time coming right written to rate 10 it is there no longer an nvm right buffer in the system so there is a right buffer but we recommend like in this type of a situation if it's an all flash you don't need it you can just go directly to rate and you might get faster because you you you miss you don't have to do this two levels of writes right so then this concept of data progression happens to your point there are three reads three ways when your data progression happens one basically you can set it based on a schedule you can be based on on-demand every time you take a snap immediately data is frozen and moved moved out here and then the third time when it happens is when it's our automatic right way if it sees that this thing is getting filled up it will automatically make sure that you take a progress you do a data progression move it out there so it's automatic scheduled on demand and you get all three right so every time you take a snap or whatever the reason is we move this data from the 8th 10th to rate 5 rate 6 during this move when you're moving it if you have deduplication and compression turned on it this is the time we will deal up it will compress it and I'm going to show you how that what what really happens during that stage right so it's not waiting till the end of the day to do the droop and compression it's as you moved at that time immediately it does the different compression and and the amount the relative amount of space for the raid 10 landing zone and the raid 5 called state data is dynamic yes absolutely and I'm going to show you that as well right do you know do your policy that you can then keep in sudden rise and keep them per minute that is correct so you could say that p3 is very important I don't want it you to move it to writing though so no matter what we'll always keep it in rate and then ok how do I identify that what kind of algorithms we used to actually identify those you know blocks that keep them in the cache oh and you think that would in so we use basically we look at the temperature and the access rate of the data ok so we can freeze the database on the on the pattern that we see a LAN the logical and then a the line is basically collection of LBAs right so that LBAs they are attached to this slice and this lights are attached so you see ok freeze this lives here and we freeze this lights over there but but to your point I think before you said about the how how we we make sure that we not have input in the in the in the performance is because everything that you do in here a grade one zero any read-modify-write happen here and then I freeze the things again and goes down this is shrink it is expand this referring this is a spam and that's a very elegant way to do this it is coming from the years of SC data progression so it's a very elegant way to do that but that's why again there are certain customers they want fool in line we have a box here there are certain customer that they want the best efficiency right because inline means doesn't mean that you have the best efficiency always if you look also at our some of our competitors they use background stuff also if they are in line because they cannot catch everything here is very elegant because I can actually divide the activity from inactive and if you look at the data data decay model right if you have 30 days when it's the time when you write more is when you have the new storage after 30 days that thinks so your raid 1 0 is basically expanding but then after that time you get more efficient every day mm-hm so that's absolutely one of the key things here now now obviously raid10 give you the best performance right so there is a there is a penalty there for that amount of time but the system is so intelligent that can shrink and expand if every clear and the another things that you may not notice you can have different rate type on saying this right so this the physical disk can have different rate right so I can have a raid 1 0 for this slice and I can have the raid 5 for the other side so in the data movement can data progression can happen this is very good especially when you have flash drives you may have a lot of capacity there you buy larger flash drives but you can have these things working together now you guys are staging writes in memory and energy staging right so you guys are collecting writes and putting them down right you're not all just going straight down to the right great time with every write that comes in the cell we have also cache yeah yeah yeah so what I'm trying to figure out is what the benefit of the raid 10 has to buy you that will only ask the question what is your measure difference between writing to an all flash RAID 10 stripe versus a all flat all flash raid 5 or 6 stripe in turn do you what's the difference so there is obviously man a a a performance heat when you go from a raid 1 0 5 well if you're calculating it in memory and as you're coming in is there well there is there is a difference you know in the in the way you when you do the XOR and then calculate the parity to make sure and you have a very you are very sensitive to the latency right and you read a write everything from that we want to use raid 1 0 now I mean cash now that's what I understand what well the cash yeah but it but the raid 5 first if the right is smaller than the stripe side it's still a read modify write overwrite so that's why I pointed out the read modify write this is the keys when you have a lot of metadata and you have a lot of incoming data right you do read modified right then you need to compute it you do backfill in the cache right but that's a very time-consuming operation if you're going down with the raid 1 0 then you don't need to do that and the fact that they can expand and contract every time because raid 1 0 is its rate and it's basically mirror yeah right there's no every time motion and the fact that I can have the other ones that give me the efficiency on the other pay space all that all the things here is calculate how much raid10 I want right and it needs to be minimal or Dean kami right so we have a tool for doing that we have a sizing tool and we say basically minimum we say about 10% to 20% need to be in that space ok that say one needs to be tens 20% yes your support get really angry if you haven't got very well if you if you call a custard if you call out with a problem and it is around flash if you haven't got 30% your support will we speak about an old flash memory this is an old flash there is only clear one there is not right this is all so we didn't what we are saying is that we we split it up so think of this like like rate that part of tier one is rate ten part over tier one is raid 5 raid 6 the secondary question right now you're saying it's all tier 1 so we're exactly the same flash profile yeah we're all the storage in these base I'm not getting like 12 NV I can't do tier 0t r minus 1 with an nvme instead of disks for my raid 10 slot where I really want that performance with that right so what you're bringing up a good place yes so right now we don't support nvme so but when we do yeah see that that becomes me will be part of these so we'll be getting our tears back shortly once we because right now we're doing we have one tier and it's just a logical dispersion with the type of great protection correct you have one tier with multiple raid protection rain with the ingest of inactive page that they do in the different compression exactly you not you not sacrifice your incoming rights or read-modify-write for performance right that's one of the advantage to use a background but the real benefit is gonna come back when we actually do have multiple tiers with different to your point regarding the memory right why we turn it off because if you write it in a memory because we're an active active operation versus other guys who are active passive we have to mirror the data right on the other side right with this when you turn it off you don't have to do any mirroring the less cash you have the less cash you have to keep coherent right yeah so we have to keep Korins you know let's have you don't worry about any of that this is all if you write as acknowledged the power you someone close the plug you don't worry about anything yeah it's also important to note we're talking about chancla trade so it's not some SSDs in the raid 10 and some SSDs all right so let's move on so so so anything which is moved out here you can do all the reads from here but anytime say if you want to do a write that even if you do order to do a read modify write it comes in as a new page I don't want to modify it it comes in as a new page the progression will take care of it right so if it's even if it's like deduplicated you really you will uncompress an under loop to read it but when you write it I won't go back and try to fit it in I will go immediately out here write it there and my next progression take care of the deduplication again right so becomes it out here now as you can see this particular page is being now replaced with this page right so so the snap is pointing to these blocks the line is now pointing to p5 which is no longer pointing to p3 because that's the newest version right so if you delete this in half automatically that gets freed up right and and then we can give this face back right and then like as I said these pool is sort of growing and expanding and shrinking right so so if it needs more space it'll just go and automatically allocate a more space and then if it sees that it has too much space you're not using as much it will free it up and give it back basically ok so so moving to the other part of it so snapshots as I said as I showed you with is sort of in green that's part of that technology built in out there so our snapshots are all readable snapshots right so if you want to write it you mount it as a volume we call it a view volume right you mount it as a volume you write it it now becomes like almost like a separate Lund even though it's sharing the same space you write to it you can have its own snapshot so basically what behaves like a almost like a separate LUN but sharing the same dataset right you can then sing set up a sync I'm showing you all or multiple use cases you don't have be implemented them just showing you the possibilities you could have a think replication setup where you have these snapshots are going to a sinc site if you so choose that you want to have now branches branch offices where you want to send it out and you can then take that volume and send it out along with the snapshots to that site right so in any in case there is an issue on the branch office they could just quickly recover from that snapshot you could also have another backup site where you can create keep snapshots for a longer duration so he could say think of this like an all flash where you keep like seven days worth of snaps you can have a remote site dr site where you would have like probably 100 days and we have the cloud for example you can think of it this as another tier behind this where you could play like a 5 years worth of snaps right so five days five weeks five months five whatever you wanted even right so as you can see and so so what is showing you basically is that it's a very flexible solution right I'm just showing you a lot of use cases you don't have to implement all of them based on what you need you can pick and choose the configuration that works best on the snapshot pulse retention policies where does that get to finds that goal was that the volume has that obviously it's specific to where it's living at a moment in time right do I have to specify on this point on this Bay or this storage system this is the my snapshot retention policy or do I go to the volume and say here's all the places I will find it and here's the snapshot policy right building that in UI I think is more I'm trying to figure where do we the snapshot policies are defined on a per volume basis on a birth system basis okay so you see both and we can we can certainly it's not uncommon to see many many different types of snapshot policies based on the different workloads that would be on a shared storage system so separate policy for each there's separate steps snapshot policy for each volume and any given volume can have more than one snapshot policy so we can start to stack you know the not only the times that they're snapped but the retention as well so the moment I declare a replication policy it will give me associated snapshot retention policy along with it can you restate your question so I'm starting from my primary volume I declare a synchronization policy somewhere a sinker synchronous and once I've done declared that replication policy the next option will be and once you've got it over here what's the retention policy associated with it so the return the retention policy for the destination replica will follow the the snapshot policy from the source no we won't we're not going to snapshot the replicas because the the source is the source of my snapshot and as the snapshots are taken those frozen snapshots are replicated over and then on the desk both on the source and the destination they will roll off as the policy states okay I think I think to answer the question very basically I think if I understand remove the policy the scheduler it's gonna be propagate to the other one all the Polish then you can modify it on the other side so we replicated will carried forward and then you modify it and say ok I change it here and you think I'll be more interested in seeing it up on the UI when we show in the demo and then the last piece of this of this virtualized speech pool is a deep compression so this is what I was showing you when the progression data progression happens we move it from raid 1 to raid 6 while we are moving it during this process while it is in flight there's the time we do the deed append compression right so so what really goes on in the behind the scenes just a very high level our paid sizes by the way if we remember was is 2 megabytes 512 KB for 2 megabytes right so from a d-do perspective that's too big right so what we do is we immediately slice it into 4k blocks so that gives you very granular deduplication you slice it out we fingerprint each block we look up in the dictionary yep it matches we verified immediately take care of remove those 4k block so it's a very granular 4k block deduplication right so take care of that so so then once we have taken out those 4k blocks then we do compression our compression works better with the bigger blocks right so do you Duke works better with smaller box compression works better with bigger blocks so what we do then is they we take this 4k blocks then unique blocks we pack them into 64k blocks and then we compress those 64k blocks we compress the 64k blocks and then we drain try to correctly create a page because everything's stored as a page right we read two pages here we did compression and deduplication dat's the page we write down right so so so just during this process all of these things happen while this is in flight being read from read and written to rate sex raid 5 right say whatever okay so that's our deduplication compression logic so couple of benefits one very deep right as I said right and give you depending on what what your workload is 3 to 10x efficiency as I said you can make it work for all flash and week work for hybrid right doesn't matter right if you're doing reads as I said reads happen from here so the deduplicated page will get read and then the original page will get reconstructed in memory and sent back but when you're doing a write it gets immediately written out here right we don't try to do that in memory and then like as I said minimal impact over performance with this technology today so that was the virtualized page pool now I'm changing gears a little bit and going to talk about your virtualized front end capability right with the virtualized front end we get three three three capabilities that we get with this technology one is online data migration we call it live migrate right our online intelligent rebalancing that's volume advisor and then online disaster recovery that's live volume that's all enabled by this technology the virtualized front end so let's walk you through each one of these so this is online data migration this is our life knife migrate how it works so here's a volume right map to a host right you choose to do a migrate now there are two options to do the migrate there is an automatic and there is a manual so I'm showing you the manual one first so you say hey I want to do a manually migrate so this is a migration use case right you had an SC 8,000 you know about your SC 5020 flash you want to migrate this line from this array to this new array right so we will say hey I want to do a manual migrate from A to B then you say that we will automatically set up create a lon we'll start the replication will give you a choice if you want to use an existing line if you there is an existing but if not we will create a new lon same size set up the replication this is all happening under the covers you select and we'll show you as part of the demo you just select yes and we'll take care of all of these things then then you can sorry then you can map this one to the host so the host basically sees two paths but it actually to the host it thinks that as if there was a second path surface for the same lunch so from the host perspective it sees one virtual LAN it's not seeing two lands it sings one LUN with two paths right so it can do round-robin IO across both right and will take care of making sure but they still a primary so all comets are happening at the primary right and then once these once all the data is out here are in sync will give you an option saying hey the two were ones are now instinct do you want to do a cut over so when you say yes I want to do the cut over that's the time we flip the roles so this becomes source this becomes target host doesn't see anything it still continues to use the two paths right no changes right and then when you so choose you can then cut the cord so everything is online and then Jason is going to do the demo so everything is online you can do it from any SD theories to NES series except the v2 right when you do the move the LUN as I you can move the snaps we can move the replication session so there's no impact from a dr perspective saying if if this was protected to a third site that all moves over when you move the lotto this is only you get this when you buy the all flash here is right so this technology is available on the hybrid and the old flash yeah I know there but for for one yourself for hybrid it's a separate license that you have to buy yeah or all flash its included and if I want to migrate from an hybrid to an old flash you can you can I use it yeah yes okay right you can use this technology yes absolutely okay simple easy to use all right so that's migration in that example of wanting to go from hybrid to flash yes so hybrids the old legacy system immune to an all flash do we have to go back and license the hybrid system or the migration to allow us then to get to it or is just because we've got it on the flash system that's enough no you need a license on each area needs a party so you don't have the license you'll have to buy it that's good so could I buy an off flash and then buy a shelf of disks or my licensing situation would be there yes that is correct you can so though so I think what you're asking is is the is the license based on capacity the answer is no I can buy as much capacity as you earned and the license is applicable we're saying that your flash came with the full licenses so if I need if I look at the old flash and I go this great yeah but I'm starting to accumulate lots of cold data yeah I don't want to buy a full flash right shelf at the back end of this that would be cost-prohibitive right and not really useful oh yeah yeah you can absolutely I can take that and throw a 30 disk shelf of big ten gate ten terabyte drives and go that's my cold data it lives down there right you could you can add it here if you didn't buy an old flash every tool add you there and you can do data progression or you could add it to another array and move the lunch of that array basically so you can do that as well okay right oh yeah just the because it's the biggest problem I'm seeing and I think we're all sharing that question in a null flash array the value proposition of data progression has not disappeared but it has been severely limited right so the interest is I want to buy this to front all of my heart data yeah but I'm still going to keep accumulating cold data and all the value-add would really be the ability to keep I mean most people we usually end up with fairly static workloads by the way that's where the volume advisor in the Federation come in place as well yes as we said you can monitor that by capacity as well or performance so you can move stuff out right when you need to this cold cold data storage cold I'm still looking at your value prop the beginning was always yeah I don't want to have to move it if I've got the Ray that can handle my volume yeah it's so much more interesting just I have my Dewar you know whatever I got for my disaster recovery but this guy I just keep adding cold data at the back yards of the accumulation because I've never been able to find a way to build an archive policy on my file servers that anyone will follow so I just keep having cold data sticking around forever but I only have a certain amount of truly hot data and it fits inside this flash shell so this is perfect yeah and I just want to keep stacking more disks until I felt a thing up and you guys give me your next generation right oh good today yeah what sort of latency can I expect that this when you quote ions but you don't quote latency so your system not on live migration sorry I'm on the system not on live migration yeah we're not talking about the slide we're talking about yes latest during the migration you mean no no no general general now $299,000 didn't all-flash basically that's basically the standard nowadays right so sub millisecond response time so that's basically what we offer you with not at 399 thousand I'll have to look it up but yes with sub millisecond response time you do get we did some tests with the data progression and everything you are most of the time under one millisecond right and and obviously not with the three hundred ninety nine thousand I don't remember a lot of of my head how much was but the goal is to stay under one millisecond when you have data progression he goes to probably 1.3 and then it goes down depends how many how much data you you need to move okay all right no no this is I mean the things that as I said before right the the real advantages is the read-modify-write so everything that you do in read-modify-write you need to pay a penny a penalty there so the fact that you're not doing that that's give us actually very good latency on that here one so that's I'm not too worried about you know how you doing it all I'm bothered about is what latency am I getting for the price I'm paying oh yeah if you're giving me a better value yeah by doing that under the covers no no I got angry yeah we ever did understand the cost and and and what I'm getting for it alright so so that was basically a migration use case this is basically the same technology but now applied in a federation use case right so basically I have no in this example I have six arrays right even though the picture showed identical but they can go all different and in the actually their demo that we're going to show you it'll be all different right so so it could be a V 3000 it could be all flash it could be a hybrid it could be an 8000 doesn't matter right so you take multiple a series put them under one Federation basically right volume advisor is on the del storage manager right so it's monitoring all the arrays these are some of the areas that we are monitoring so you can you can be based on CPU ttle ization it could be based on respond time it could be based on system capacity you get you decide what you want so you could say for example in the demo date we are going to do we will do it based on response time is in response to massive utilization one of them latency yeah so from a latency perspective will say hey if if the volume shows you can set the threshold if it shows above this recent response time give me an alert right so so so it will not only give you an alert it will also give you a recommendation that's where the intelligence come in because giving me a lot is groove great now what do I do with it right so it will also give you a recommendation and that's where the intelligence part comes in it will give you a recommendation and say hey it's it's crossing the thresholds that you want it now I I would recommend that you move this line from this particular area this particular a right now if you most customers make that recommendations I kept I like it click OK we'll move the lon if you want to change it or do something different then that's also optional to you basically it's storage DRS at the array level okay yes exactly so so that's basically so you get a thing it gets an alert right you'll get the recommendation you take the recommendation will move use like migrate which I showed you earlier that same technology move the stuff and then hopefully you're back your situation is normalized right and this is the part where we are going to do the demo so so we'll give you a live demo of how this thing is working is it fully automated what's the policy on the migrations yeah so you can set the threshold or yourself right we will give you the recommendation but will not move it for you right we'll just give you the recommendation and then you decide to both you click the button anything yes I like the recommendation move it what if you don't like the recommendation you say ok it may be recommending a but you say no no a has some very critical applications I don't want to move it to a I want to move it to be you can in over all available through the REST API I wanted an AI at the back so it was an important point that you brought up I was just gonna talk about the API so we've got a public API which supports both rushed and rest and PowerShell and although it is manual like will provide volume advisors will provide the recommendation at the end of the day the administrator needs to evaluate that and and carry forward what you bring up is interesting because the way I think about the API if it's in the public API that we offered you know have some build a script whether its rest or PowerShell do at any given time on you know one scheduled interval evaluate recommendations made by a volume advisor and automatically and maybe you put some conditions in there but automatically care you through you know for audit fully automated storage DRS as you brought up I can get my own script that will push the button for me I say it's interesting and I'm assuming we actually API yeah so hold your thought on the brain is going to address that situation so if you want fully automated right if that's what you want right I I don't want to press a button I want the thing to automatically move it right say for example typically this would be a situation where it's a mission critical application right say something goes wrong in the middle of the night you're sleeping you don't want to be you are not there to press the button and say move it right so this is basically our live volume technology comes in so if there's a mission critical application and you trust the system to do the right job for you right because it's absolutely critical that the thing is up all all the time even when you're not there at your desk and that's the technology so in this case again if we have time we'll do this demo also so in this case basically again the same underlying core so you have the volume you it shows up as a virtual volume to the to the launch so it sees both paths you have a sink replication so we if you want to do an automatic failover we recommend the I mean it has to be a fiber channel link with a 10 millisecond response time right you could potentially have all the hyper channel MPI skocy if you want the auto trail over if you want that automatic failover then it has to be a fibre channel we support ice Kazi also for the link you are talking about right okay so the replication link has to be fiber channel but the presentation layer could be ice fiber channel or ice cuz your bonus okay right okay great right so so this is a response time that we need then there is a third site we potentially if you want like so this could be like within a metro distance and then you could have a third site where you can replicate which could be an async replication happening right in case both of them go down then you can just store from there right so so in the middle of the night if a disaster strike this array goes down if you have set up the auto failover then automatically we will move the lawn we will move the snapshots everything automatically fail it over to this to this to this other area while you are sleeping right the other thing that happens right if you've got a ten note you know if going like eight or nine milliseconds latency between those two sites you might as well get a hybrid or right because there's no point having a flash at that point yeah so does the max of what I'm saying right you don't need to keep it that way you could it could be laying that's the max all right if you don't but if you're going in that environment that's God yeah save yourself some money buy yourself a hybrid flash never say that two guys trying to sell stuff that's duh yeah so that's your maximum you don't have to go there so so we do the automatic failover next morning you come you you you fix the system up when you bring the system up you because the line if the LUN wasn't destroyed you can do an incremental attach so you don't have to send all the data back only the data was well that was written during the night you only do an incremental attach so it gets recovered pretty quickly right and then it's restored when it gets restored this comes up as your secondary that still stays as your primary so you can operate in this mode you don't have to fail back right but if you do want to fail back it's an optional step you can again do a fail back and bring it back if you want this to be the primary always you can do that and that's optional entirely okay so when I'm in this Metro this is basically Metro clustering that just standard rails exactly when I'm in this configuration my host is seeing two paths as being live yes I - I see they're there they're all more than a few miles yes but they're all gonna be they're all gonna be active okay David and if I send IO to the passive for the non live copy does the bay route that IO back through or does it say don't talk to me I'm not really here on a vs. don't talk to me so we want you know her to pull off stretch storage and also you know past the vSphere Mitchell storage cluster certification which we did yeah the you know what you call the passive we call it the secondary live volume okay in the architecture that is a proxy device okay a storage host museum or any any storage host we'll support live volume it appears as a readable and writeable volume so feel free to send it redial right IO okay um but if it's going to the proxy that IO is forwarded through the replication link whether it's fiber channel rice because you've back to the primary live volume if it's a read the data payload as fashion sent back if it's a right it's committed the acknowledgement is sent back through the proxy to the store JAL's and then in turn replicated synchronously back to the second or so that we have synchronous flow so I do want to pay attention to my site preference assignments in those cases so that I'm not in a case where I'm trying to use the pass for my secondary as yeah yeah you don't want to redo that to all the yeah because I'll be doing a lot of back and forth the path selection policy with respect to this architecture brings up a lot of interesting scenarios on what you might want to do what path selection policy you might want to use fix them are you round-robin support Ulua comes into the discussion yeah whether your uniform or non-uniform storage presentation comes into the discussion yep you're absolutely right you identify that if there are optimal paths in a round-robin uniform and there are available non optimal paths though I have to in terms of performance so from a design perspective I have to pay some attention pay some attention i I can say that in the future it's gonna become you know Louis support obviously that'll be really easy so we are introducing the Aloha support so at that point we can automatically I sleep at the producer yeah so if you are in uniform and you want to use round robin and levy eight the management overhead of you know defining all your paths and on your fabric you would then be able to use round robin in a uniform configuration will and then the secondary live volume the the you know the passive volume will advertise its path to the storage hose as available non-optimal in other words they'll have a higher cost right round robin VMware's and all further down those will ignore those paths until such time all optimal paths are gone or dead or not not available okay that's good to know that's coming cool alright so transparent between any essays all the same things I believe in applicable life again move to that will show you all these things and I think one of the things that we need to remember that's what allow us to support stretch class no it's just wondering exactly how it's done a summer active active this is an active passive with the parking notion practice so so you can use like migrate to move between the SCR A's transparently we also have a technological thin import that allows you to migrate from other arrays beyond the SC family so it could be a ecologic it could be MD series potentially it could be any air it doesn't have to be these two right these two are what we qualified but it could be pretty much any array right so and that's all built in into the product right and as you can see it can do move all these Lunz into this array and then lastly but not least we touched upon this a little bit cloud IQ so cloud IQ is basically our cloud based monitoring monitoring analytics we have built in intelligence there that actually looks at the array and can provide you with recommendations right machine learning type of thing right so so and that's also again included as part of the base base array license right you just have to sign up and and that use right we offer it on the unity extending it now to turn now to SC the s hybrid are no flash hybrid in all that's correct that's right right so that's cloud IQ single pane of glass for monitoring they the other cool advantage with cloud I key was that you don't need a VPN you don't have to install anything you are traveling wherever you are as long as you have an internet connection you have a laptop a tablet whatever it is you just log into the Delhi MC portal and you are able to see all your arrays across all your data center sorry all you made range arrays in all your data centers on this right it'll give you alerts that he'll give you performance that'll give you utilize capacity utilization if there's an error it will give you those alerts everything you get in one place right and then we will bring in the html5-based Unisphere very shortly on the SC similar to what we have on unity and these are some of the other integrations that's already built in into the product today okay so if you don't have any questions I want to now flip to the demo I wonder whether we have enough time yes please data a drastic caption is it it was included in the license yes is it enabled by default so data at rest encryption right now on the Union on the SCC's uses external key manager right so so the license is included and if you have to turn it on there is a you have to buy these special drives right the sed okay yeah drives right and then you have to also by the external key manager so it's not internal key manager so you have to turn it on you can set it up to use that so what's the performance impact if you enable it there's no performance impact no okay that's a drive yeah actually more so than right industry now they are enabled for yeah cuz the v-max enabled it by default and didn't even get differences on on the unity side we have there inside of the controller yeah so we not need to use a drives you is a difference there but same same results right you don't need you you don't have a performance thing but one require external kena manager the other one does so depends what is the use

Channel: Tech Field Day

Views: 9,102

Rating: 5 out of 5

Keywords: Tech Field Day, TFD, Storage Field Day, SFD, Storage Field Day 14, SFD14, Dell EMC, Kaushik Ghosh, SC Series, Unity

Id: PDmq-0Dph64

Length: 85min 36sec (5136 seconds)

Published: Mon Nov 13 2017