Inside NetApp's Cloud Services with Anthony Lye

I've been in a nap not five months yet and I'm not a storage guy never have been I can never will be never what my background is really exclusively in software engineering application development middleware I date myself back to some very interesting startups and my role in my career tends to be you know somebody that the polite term is probably assists in transformations to Gabe's message I've broken a lot of glass at various organizations and sometimes it's worked and sometimes it hasn't and I always sort of maintained that you learn a lot more from your failures than you do from your successes and my first job out of university was actually writing a relational database for engineers and scientists on unix from there I sort of bit the distributed computing bug and my accent probably sort of gives it away I was grew up in the UK but at 24 I moved to Austin Texas which is an experience and I joined a very small company called Tivoli systems I was Employee 12 at tivoli and then after tivoli i moved here and joined another small company called remedy and built the product organization at remedy corporation and then like all impetuous kids i started my own company for a people that we ran up one side of the bubble and down the other i then sort of tried to take a year off and within 3 weeks i went to work for Tom Siebel running the Siebel organization and building not just the main Stiebel on Prem product but really I was put there to sort of build a set of a competitive product Salesforce and then was acquired by Oracle and spent you know six and a half years trying to leave but having some really fun interesting projects with very Ellison and then as I said I sort of ran a company called hot schedules as I see almost four years and got called from NetApp and I guess part of my interest was having worked for George's twin-brother I almost wanted to see how alike they were so for anyone who doesn't know George Korean the CEO of NetApp is the twin brother of Thomas curry and president of Oracle and as far as I know you're the only person to actually have worked for both of them I should write a book they have they share some some similar mannerisms it is true and they are both equally as fun to work for so my remit here really is to run a business and I'm given a complete autonomy to do that I run an independent P&L we run where we have two different systems and where we get leverage from the existing systems we leverage them but my business is really about you know a cloud first methodology and how can we monetize existing assets where it makes sense and how can we build or potentially acquire new assets where it makes sense so as I said my lens is really you know how what can I build on AWS on azure on GCP and bluemix and I think that will probably extend to Alibaba and Oracle I think we're probably in a wait-and-see mode the question you asked mark earlier the clouds I think present two very distinct opportunities and I sort of liken the opportunity sort of metaphorically to like a dumbbell you know something that's in the middle with two big weights at either end and why people are I think the cloud really represents I think a sort of an infrastructure play on one end a SAS play on the other and then the middle bit which is where I think the cloud is more of an enabler is the sort of the platform as a service so the infrastructures of service is very appealing to the migration of existing workloads and the other end of the dumbbell is very very appealing to the new application developers and the services that present themselves and then the platform sort of tries to find middle ground between - but in terms of the business opportunities it's very very squarely on one end on the re-platforming or existing workloads and on the net new application development and the new services the streaming analytical services machine learning micro services containers so what I thought I'd do is just sort of take you through a little bit about the portfolio like with mark just stop me and ask questions it's much more interactive that way and we can focus on some things I've got some interesting data in here some benchmarks and things that you might find quite interesting I hope anyway a picture of me I run five miles every single day I try and do seven days a week I think I average for the last seven years six days a week you don't want to meet Ben all right one of our delegates so unfortunately had to leave her is a particularly ever drun er yeah it's my I get up at 4:00 in the morning and that's my alone time I love those - he does - yeah then on Tuesday yeah and he did not get - he says he's one is battalion was mine sorry he says he says he's one is better than yours - remarkable no Martin why did that sorry no one there we go I love the talks and I would say even my jobs that people in Oracle just like here are startups well if you don't mind me just picking that apart party because you been brought in as the cloud fix-it guy to break all the glass ceilings at meta but going through your history of Siebel and remedy and Tivoli and Oracle you know that you know the diehards of the legacy sort of legacy talk about why they work I'm not at the time no the time not ago I joined I joined Iliad 12 I joined remedy when it was 22 people so your we were breaking glass I was at IBM when we acquired Tivoli because it was such a radical company and we had all bailed by them none of us were typically your your sort of use cases is that sort of small company disruption that then grows into the future legacy and I hope you can only yeah I mean just answer it specifically I could for you guys if you wanted to I follow a very simple strategy for investment and that investment strategy is both my money and my time and it's based on something called discontinuity theory and if you ever want to get bored by me for an hour it's a really interesting read it basically suggests that markets are rarely if ever made they are almost always resegmented and you can basically watch disruption and resegmented occur around one of four lines of opportunity a technology disruption a regulatory disruption a distribution disruption and the emotions of standards and so you know when we started Tivoli systems management existed we didn't create a category called systems management but we bet on one simple thing which was UNIX would crush the mainframes and machines would proliferate so we bet on a very simple distribution discontinuity but the capability of user administration of systems monitoring that was all done there wasn't anything new in that and remedy was the same way I mean people had help desks people had call centers that just again there wasn't the opportunity to leverage the kind of distributed nature of the machines and the idea of building applications that could change so quickly on a dime so everything gets old it's just a fact of life but the things I've done have been at the time when they were new and disruptive and my career is about leaving them when they're kind of getting old so everything I've done you know when I joined Oracle and Mary's to do the cloud I was like have you seen your press you know you keep saying it's a piece of and I shouldn't really say that on a public web site but I and he's like I know I know but you know just go do it and I would have to stand up you know with Andy Mendelsohn on the database side and that time us on the middleware side and they were looking at me going who are you like you know this is Oracle you know we have a 12 billion dollar database business and we have a multi-billion dollar middleware business and you're doing this little when swear again but this little tiny thing and without Larry's protection I would have been assassinated there's no doubt about it and so I think back in that day that there were probably more naysayers than there are today about the cloud I mean there would probably I would go I could tell you funny stories I'll just share with you one I went to see CIO of a very big bank in New York and he said you know pitch me the cloud and I did and you know he kind of stopped me and just said look you know I have to laugh because that's just the biggest bunch of crap I've ever heard he was like we're a bank he was like do you really think I'm going to put my data my critical data somewhere else you know we're back like we we collect things you know we have vaults with money in them and I said yeah yeah you will and he was like you seem pretty determined but yeah I mean why wouldn't you like there's nowhere you'll be able to do it as good as in as they will that they will get so much scale and and the whole idea of sort of just commoditizing the stuff I mean I will tell you that as markets commoditized so as the technology industry right so what used to make people different was they could afford a computer and the big guys could afford mainframes and mini-computers and the small guys couldn't so there was a massive sort of step improvement in productivity for those that could afford computers and those that didn't and what the cloud has done I think more than anything else is I always say is democratized innovation and what I liked about it is in many ways the smaller companies that couldn't afford the computers are now in at an advantage over the big guys who have stuck with them and so the cloud sort of a spread itself I think an created a framework where anybody can do something and to that end there are sort of more and more new buyers so you know I remember back in the day I'm old enough to remember when IBM told everybody what to do and people would listen and then as more and more computer companies came up there was this person called the CIO who started to show up who would tell the vendors hey we know what to do we'll manage you and I was remembered with the cloud the cloud kind of were you know on the application side we found the IT people to be just painfully slow and very controlling so one of the wonderful things with the cloud was we could sell it to the line of business and we could basically go into the line of business and pray our names would pray because ice be called a departmental renegade because we could pray on the cost and time issues that IT had and we could serve applications to line of business people that they put on their credit cards and so all of a sudden decision-making went from a few small vendors to IT to line a business and now with the cloud it's gone even further I think anybody inside an organization can buy a service and I think increasingly if that service demonstrates you know alignment to the CEO objective of business growth IT has a choice to either stand in front of it and get shot or build an infrastructure that enables it and so what we try and do here is we try and build applications that serve the needs of our existing users but but more so probably than ever we're building services that we think appeal to the non traditional IT people so the kind of cloud people the application is this the kind of products you're going to talk about the actual cloud products that you have it okay I told you I can talk for a long time I do this I mean this I always think is a funny slide and didn't format itself lightly as we copied it over but you know I mean if you're not in the cloud you're sort of weird I mean I almost think that when people say to me we don't have a cloud strategy I just say to them you know can I speak to the guy in the organization that does because everybody does whether they know it or not every single organization on planet Earth is doing something with the cloud and so rather than align ourselves to the 1% I think NetApp strategy is is a lot better served by aligning itself to the 99% and so we love the cloud things I mean I think the clouds are going to expand the market they're not going to cannibalize to 100% there are still mainframes that run around IBM just announced a new one and they're still there and so so the markets continue to expand and so the net opportunity for us is what we've got that's growing plus this new thing that's growing very very quickly and there's no reason we can't make as much money on the clouds arguably more than we make on print so today this is old now I told you we're up to just under 2,000 customers who buy from us on cloud so this is not a sort of an incubator this is not a sort of an idea lab this is a real business the business is growing about 300 percent year-over-year and will continue to grow 300% through at least the next two if not three years and I believe that this business in the next 18 to 24 months will be meaningful enough that it will start to show up in our earnings it will start to be a core asset of the business because we've got some really good things we're at where is it at now is it we don't know close it I wish we could because it's a really good number but then apps 6 billion right Mexico financially I can't do it until it gets to a certain percentage but it's a really good number and I keep arguing with our PR people but it's a good number and it's growing really fast and I'll talk about some of our I can talk about some of our customers I mean we've got Odin have the storage number up here where we're met with managing more than two hundred hundred and forty petabytes of cloud data now and that's growing triple digits and we love that we love you know data and wherever it is we'll try and put some value out on top of it and try and monetize the services and we're going to monetize the services in the traditional way in terms of applications but we're now really starting to monetize the services through a POS and not just the functional api's but increasingly orchestration 2vp is a POS that we believe will appeal to application developers and you'll see some of that stuff as we go through the slides but you know you're going to start to see a fairly regular kind of drumbeat of interesting I think announcements from us around API s and around orchestrations and around the combination of storage not just in the traditional sense with compute but we're doing a lot of work now with the modern containers and we think we have a ton of value that we can bring to the people who are building the new micro service based application stacks these are these are the products that I have and I'll take you through a couple of them and then we can I've got some benchmarks after this which you may find interesting so on tap cloud we take on tap as it comes off the production line today every six months we basically have wrapped it in a set of web services we've built a brand new sort of on tap management system that allows us to deploy on tap cloud to a hyper scaler or to multiple hyper scalars and then to connect those endpoints to traditional endpoints just as if everything was on-premise and we are finding just the the basic benefits of on tap obviously appeal to an existing storage administrator but are becoming very appealing to non NEP administrators now showing you one of many other benchmarks in a little bit Netta private storage is an interesting concept we we have been sort of architecting a more traditional hardware architecture where we put box very very close to the hyper scalars we find that a lot of customers are reluctant in some geographies in some industries and through financial considerations to just copy large amounts of data into a particular hyper scaler to get access to compute and so we sort of pioneered an architecture where we can put storage very very close to the hyper scalar and using the Express route you know kind of Direct Connect whatever the terminology is from each hyper scalar we can actually put storage close enough to the cloud where we have latency that is as good and sometimes better than storage in the cloud and so we're able to separate storage quite nicely from compute which actually allows as Mark said the customers not just to consume compute from one hyper scalar but to consume compute from any hyper scalar and as an application developer look I sort of drank the AWS kool-aid and it's it's you know it's a developer you're like a kid in a candy store it's great until you get the bill but even so you know what all of a sudden you're like this is really good and then someone gave me a demo of like you know Google's machine learning you're like ah how do I get that and you're like oh no I can't get that well my data's right there and the only option I thought I had was to copy it and then all of a sudden IBM gives you a cool demo of Watson you're like okay this is getting annoying why do all these guys start to one-up themselves and I've sort of you know largely because of history and ignorance I've sort of put myself in one place and so with NPS you can basically swing your data and mount it on any of the hyper scalars and so one of the things that we're doing is look I mean it's really good at selling boxes and people who love boxes understand how to buy the box and go to a Colo and put it in the Colo and connect it and establish that stuff but it's still I think for a lot of non storage people too hard so one of the things that we do in my group is we take really good idea we wrap them in software and service and we express them as cloud services so you know we'll be we'll be putting up NPS as a cloud service we net up will be putting some boxes into kolos will be doing using the Express route Direct Connect there will be a website at nano comm where people can basically move data to our infrastructure and use compute from any hyper scaler they want and they can take us on a you know consumption model or a subscription model they're just simple things that that start to create very rich and capable cloud services that we monetize actually in many ways more profitably than we would our machines by themselves because the thing with the cloud I'm sure you guys have all done the math it isn't cheaper I mean Amazon charges your load money and with you know with up X versus capex we on the vendor side you know continue to say you know that subscriptions are the gifts that keep on giving because you keep paying it's not like you bought something and you capitalized it you know you get the bill every single month every single quarter every single year and so if we can take our infrastructure and share it amongst many customers who pay us continuously we actually can make very sizable margin contributions to the company's bottom line all Travolta's are is our Cloud Gateway product it can be purchased as an appliance or a software and it basically just sort of sucks up data from locations and basically moves through compression and D to structure data off to the lowest cloud place it can find it connects to about 23 different cloud infrastructures today so it will tear off not just to s3 but to glacier it will find the cheapest place it can put the storage and as well as compressing it anywhere from five to one to ten to one it's basically starting to sort of replace tape as a low cost medium for data with the added advantage of it being available if you need it as opposed to calling a mountain and having them ship to take back to you cloud sync is a really nice product it's a very simple sort of source destination mapping tool that allows businesses and I'll give you something a second to move data from one format in one location to another format in another location we can take a file system data and we can transform that into s3 and we can basically your customers can take clouds and can do it once or they can sign up for a subscription and cloud sync will continue to sync between both ends so you were taking a file format tagging it into an object format so it's presentable for things like EMR and other kind of services but we support now our own storage grid infrastructure we support various NFS infrastructures we support sips and a lot of our customers there are non IT it's kind of one of those things where if you do a good job it's like fashion it kind of spreads and when I joined I noticed that there was a high collection of life science customers that were using cloud sync and I said there must be a life science use case that we kind of called a couple of the customers and the first thing they were like we don't want to talk to you I know your user of our partner like yeah but we don't talk to you and waste of wine they said because your tell IT what we're doing and I said oh and I said who are you and they said really you know we're a couple are like data scientists in the labs and we don't really use ITV they're just kind of they do the corporate stuff and we do all the you know sequencing stuff and they said we have these hugely complex modeling systems that take days if not weeks to spit out models and these models are like gold and if they become in any way you know polluted we have to go all the way back and regenerate them so we have been buying cloud sync and just ripping off copies of our models into s3 and then using s3 compute but knowing think we'll move any changes back into our system and they said it works great and we just pay you out of our R&D budget and you know we've been telling all of our friends you know life science events this is the coolest thing ever and we've sort of started to sort of become a little bit of a cult product in life scientists but line of business people know IT involvement at all and just to pause for a second as I said all of these products have application user interfaces but all of them not 100% today but soon we'll all have ap is that any of the services that you can do through the UI you can also orchestrate through an API and so application developers can come in and take advantage of our services in their code and so instead of so for example in on tap plow going through the setup of a storage volume and on tap there's a million things you can do as application developers we don't understand 90% of them we just want a simple share and we just want to be able to populate it with files and I can do all of that now to a single restful api and so with application developers we are sort of NetApp is creating that same democratization that same sense of expanded innovation but doing so for the application developers directly on the hyper scalars so do you see a3 becoming a default protocol for the cloud because you talking about uploading and downloading files I mean that's out in the way we talk you still talking NFS FMB no no I mean I think s3 has done an amazing job in terms of object storage I would say by far and away the leader I think a lot of people are starting to sort of try and base their own services on an s3 compatible you know kind of model because there's so much in s3 but it's like anything I think standards are hard to maintain and I think every time someone emerges close to being a standard the technology industry likes to go fight it and try and create something else I think you know we're in a very early you know to use a baseball analogy we haven't even been through the first innings yet and and I as close as we are and ask the hyper scalars I'm just amazed by how much each of them are now doing and how much differentiation they are putting into their platforms but in a way you try to abstract that differentiation but there needs to be some degree in some use cases yen's and others no in others if we think that you know moving the data directly into a hyper scalar and making that data available to other parts of the same hyper scalar we think that's a good business I mean there's no reason our business a lot of our business is moving data from one part of a hyper scalar to another part of the same hyper scalar whether that's tearing to lower cost or whether that's doing H a there's a whole bunch of reasons that that the customers themselves need to move the data they need to move it to analytics from the transactional systems so there's a really big problem just within a single hyper scalar now I think everybody is going to have to do business with at least two if for no other reason than to keep the first one honest so everybody I think is going to have a multi cloud strategy certainly on the infrastructure side on the application side it will could be tens of cloud companies and so you know there there are lots and lots of interesting problems that I think we can solve natively on the single hyper scalar in a multi cloud environment and then in an extended hybrid cloud environment where that's our gear or somebody else's cloud control is a brand-new product it only runs on today AWS soon to be as you it has no understanding of hardware or on-premise at all it is a micro services based platform that focuses on what we call SAS data protection so I don't know if you guys know this you probably do but you know all of the SAS come he's back up the data at Oracle I backed up the data for CRM and HCM I had the data there and I had it because if I experienced a disaster I could recover to an agreed and contracted RTO and RPO but if a customer called me and said hey Anthony you know can I access the backup for my account from two and a half weeks ago and can you restore the opportunities from John Smith who left and deleted them unlike no well you've got the data yeah but I can't give it to you and in the way you want it it's me I backed it up for the whole site and so when you delete things in Salesforce Salesforce recognizes that delete and happily deletes it for you when you're in office 365 and you start deleting mail after you've been through the recycle bin process Microsoft does a good job of deleting it and so a lot of customers are realizing that while the SAS applications are running somewhere else and giving people the functionality that the granularity of the protection doesn't exist in storage terms we would say they've got a disaster recovery backup but they don't have archival backup and so we basically have built a framework that allows you to sort of orchestrate protection policies that in turn create sort of processes that basically call the api's of the SAS applications and basically what they do is based on what they're told to do they go into the API and pull out the data they're supposed to pull out and then we persist it in an object store we say we take Microsoft p55 data emails SharePoint sites onedrive accounts and based on what the system is told to do by the administrator we go in and pull out the necessary information and persist it we can actually persist it in an s3 bucket we can participate in storage grids our own product so you can actually back up office 365 to on-prem if you wanted to and then from the application you can recover it and you can recover a certain mailboxes and all all subject all things with a certain subject all things in a certain date range and you can restore them to where they were or to a different place and so a lot of organizations are trying to realize for legal reasons and for other reasons that they need a granular protection strategy that the SAS vendors don't give them but you're saying this is both backup and archive both well not even again I don't want to debate he's probably a better explanation say what I can tell you what it does what it does is it basically protects data any data that you want the whole thing or certain things but the granularity is more like archives than it is like er okay and the nice thing about that control yeah you can restore it grain and that's that's what I was trying to differentiate just to restore a service is talks more to your point about dr vs. archive which says your back recommend trusting 365 those guys will probably beat you for just restoring the service yeah but when you go back to your example of legal like doing legal holds and whatnot that's not a dr function correctly exactly no I'll at all I'm the email that Dave sent me that I accidentally deleted last week I just make a comment about the trend disease because I think it's so cool when we first started giving back to the what are you doing and when did you start the first thing we did was let's take the software we already have a know and throw it in the cloud right so we threw it in Amazon and then we threw it in Azure and it's like okay it's a start or let's take the boxes we already sell and connect them really close to the cloud and we extend it a little bit with like alt of all let's have an appliance that sends data in the cloud so that's a little more cloud centric for a long time cloud sync was my favorite feature because they didn't necessarily involve any net app hardware it is a cloud service I mean you you go to the Amazon Mart and you buy that cloud service and it reaches out into your data center and slurps the data in independent of whether it's our boxes or not our boxes it's still an on-prem to cloud connection now cloud controls my favorite because you've got the thing moving from one cloud data source to a different cloud data source and one's a sass I mean we're not there yet but you can imagine well what if I wanted my Salesforce data to be connected to Watson well whoa now you start to get I'm not announcing a product or anything I'm just saying look at what we're doing in the infrastructure we're building and having a data architecture that spans all this stuff now you're starting to get a sense of oh those guys could be very interesting in a multi cloud multi cloud service environment with data as the central characteristic of their about the other did anyone i rolling no no I think it's about the other dimension I would add to that sequencing is you know when we first started with ontap cloud we basically created on tap as the same on tap and threw it in the VM and we said the customer here you go start it you stop it you maintain it now what you're seeing from net up is we're actually starting to build cloud services that we are responsible for the uptime we are responsible for the performance customers come in and subscribe to a pure service from us and they expect that service and the API to that service to be those available and performance and so we are increasingly putting our resources our engineers not just into throwing software up back to that Bob that dumbbell analogy we like VMs and we'll put our software in VMs and people can start them and stop them and fail them over and do all those kinds of things but at the other end we are building application services data services for which we maintain the availability and performance of them just cannot pick that a particle you said we're building application services and data services a lot of these to me is keva bit shifting either on Prem to public cloud public are back in between public clouds but where's the where the application I'm thinking of things like analytics or things like sure you can Moo you can do a copy of your Salesforce data say to as you have to doing some machine learning all to Google machine learning all that kind of thing are you maybe moving up the stack to do some of that analytics or Dell data stuff always it's still good shifting around no I think I think well so I think it's I think we start with the bits as I'm what I'm trying to sort of present it I think we are recognizing that the data itself still has a very tight coupling with compute and the way in which application engineers are now building applications there is there are an ever-increasing number I think of connections between what we do with the bits and how those bits get consumed by compute and how those combine packages are then distributed across the public clouds so what do you say to companies who for example are building a new app and they're using Amazon or as you're not the blob storage or the s3 but they're using DynamoDB or they're using Kinesis so they're using whatever on the Google side among the Amazon Yard those are sort of slightly isolated data lakes now I'm sort of if it was in an s3 layer I can see how you can replicate somewhere else but now someone's got a whole bunch of data array DynamoDB how would they possibly connect their Google service to that and how would you be able to move data around to help it well so I mean I think some of the great things that on tap is done is it it extended the concept of sort of bits right into the databases and it provides snapshots and it can basically support DevOps flows or analytics flows and so you know just let me give you a very simple this is a really interesting benchmark so what this is is this is over three various different storage requirements on AWS this is if you bought EBS just EBS and then you bought you took that same EBS storage and you added on tap cloud onto it so the cost of EBS compared to the cost of EBS plus on tap cloud and what's really interesting is if you go up here or even up here you're you get on tap cloud and your bill goes down single-mode and h a so if nothing else everybody with an AWS account who runs on tap cloud gets on tap tap for free and money back and they get all the services of on tap to every single Amazon storage administrator love's on tap out as you're saying that on tap cloud can manage the data more efficiently we do effective deduplication compression a variety of stuff now will amazon add that someday will Azure sure we're in a future war as we've always been and right but we're how a petition yeah and that don't love it let's be fair there are certain things that Alan Tech cloud can't do in in AWS right if I've got it on tap cloud and H a in a single V PC I can't access that data in it from a different V PC without not NAT not yet yeah I mean it's like anything I'm saying is the stuff that took 25 years is pretty good and you know as good and as fast as Amazon is some things just take time and some things are harder to do and I think some of the challenges that amazon has with its its sheer speed have I think given it some architectural considerations that it will be punished for and we see those as great opportunities and just in the same way as we always did but it's just interesting to me that just forget forget a hybrid cloud environment forget swingin workloads anywhere like I could run a whole company just on this graph I could literally go and raise money from the venture capitalists and say by me on Amazon Asia GCP and SoftLayer and just run me on top of your existing storage and I'll save you money and give you extra features so I'm curious companies to do just that have you have you done any benchmarking of this against like Dropbox Box onedrive yeah how does it over there so we're going through sequentially benchmarking our capability yeah but I'm asking about the results well so I've got one you're a great great straight man so this is interesting we ran an NPS Ben mark so this is a box outside of AWS the storage isn't even in there and we basically used the Express rail and we've matched it up to AWS and we consumed AWS compute with data bricks and spark and so when you tell people hey let's keep the data outside the hyper scale everyone goes how can it perform it why would I ever do that well actually data bricks has been working on s3 for eight years optimizing spark on s3 for eight years little old net app comes along with the box connects it and runs benchmarks on one case stream 100k streams and the million streams and these are the NPS boxes with various workloads and these are the AWS s3 native workloads so we actually perform as well and sometimes better than s3 okay so my question wasn't about s3 my question was about box and drop but this is this is this is a spark data bricks use case so the application is data bricks okay I'm just saying this is just an example of an application a specific application use case where we are benchmarking our data services against the hyper skills so I don't have the box one but we are talking to box but of course we're doing a whole hello to voters so you don't have the results for the question that I asked not to box no okay thank you can I just clarify on what that's actually doing I'm not because we have a picture I'll I'm not sure so s3 n AWS that's sitting within absorbs clouds so the NPS is the net up infrastructure this connects via Express route to a degress okay so it goes across that where the data resides which is where the data resides don't judge me so the s3 comparison it's not residing in the same place it's sitting inside AWS all right good yeah okay we've got the same data set yep put 1 and s 3 and 1 on our MDF service and ran them both through data bricks this example shows how we tend to think about app stuff right someone was saying well do you need to get into there historically NetApp is and labeled all sorts of environments we got big into oil and gas and we made good buddies with landmark graphics who does that app and then we partnered with them to say well what's the best way to configure net app and landmark graphics to make the whole thing faster and better for the end user does the end user come to net app in order to do aisle and gas seismic processing know they go to landmark graphics who then says you know if you want to build the best environment to make this good let's go in together with net up and figure out how to make it and then and then you see here Anthony's doing exactly the same thing with this particular spark example it's like it doesn't mean that we have to get in a competitive fight with spark we just want to learn how with sign off the big ones right start working our way through work with that up and we'll make it better so nice you have an interview by this is so when I put the data in s3 and I moved it to GCP I had to pay to get this stuff out all the data on the MPS box I didn't pay anything know and that's kind of where my question was going because doing that processing where is that are you sending data into AWS to be processed no is it coming back the data continues to live in the box next to the cloud right and it's accessed real time over the express route or Direct Connect and because of that I mean we gave we have seen people literally do database failover cases where the database is running an azure they shoot that one reconnect from the other side bring it up and Amazon and independent of the size of the database the failover is whatever the database failover is so five or 15 seconds right so that's completely shifting what's possible in the cloud arbitrage your compute costs change your mind once a week change your mind once an hour yeah right I get a different two into the weeds on that but it's like it's a CPU which has access to memory stuff which is not physically in the same location as where you are well they may be able for you like max when exco open it a little is often Amazon in fact is sharing great but they say oh I but the connection is overexpressed throughout so if bytes need to transfer across a network and then they move in both directions someone is paying money for that bandwidth you [Music] which is yep which is why I wanted to understand what the workload is here is it is it all going in or is it coming out or sorry hyouta nodes are are basically connected directly to the net up storage yeah so there's no copy so that whole job trans the whole the whole idea of being able to you know copy data into nodes and then processing that that's that's that step gets skipped and then on top of that the ability to process that data much faster you can see it at when you increase the number of worker streams that we get better yeah I'm just it may not understand the work experience I think so the strains they're streaming into AWS to do the processing yeah it's streaming type of thing so the data bricks guys they have that with their spark they have that stream streaming process with regards to the work flow so extreme to kind of working memory on the yeah no I'm not paying billions of dollars a year to stood process this using this thing okay and you know the question you're giving at the cost structure and stuff if you have a model which is read lots of data and then answer give loan don't give loan that will be very effective if you've got a model which is fundamentally data transformation read lots of data do something and write just as much out then you'll have a write bill right I found out that's where I was getting to it but you know if you just think about the adoption of spark and then the uptake of machine learning from Google or machine industry or from Asia I bet you that's going to be like an arms race those guys are going to fight it out every week they'll be you know showing some better feature and I reckon that every company in the world will want to swing their data to the best service they can find and instead of having to copy it in and then pay as it gets copied out and the problem of duplicating our data we have a fantastic service a fantastic service that I think everybody on planet earth would consume sass providers all excited about the NPS solution architecture here is that I'm data bricks all of a sudden like my cost of Sur my cost of storage goes down alright I don't have to pay s3 all the time and pass that on to my customers so I can be more cost competitive and if I you know if I stand up a data bricks cluster and let's say Asher which you don't have today but if if they did you know they may be able to arbitrage their compute costs so you know there's a there's a SAS play work we could sell this architecture and these this NPS services to SAS providers and give them the ability to pursue a much stronger multi cloud architecture versus just like I've got a big giant data center full of stuff at AWS and a big giant data center full of stuff at Google or whatever you you don't have to they can do think we can enable things that that their current architecture approaches not I have a said you know that amount of data that I purchased with capex it's sitting my Center and then I burst with CPU because that's cheapening yeah so we do this and maybe Mark you can elaborate but you know this is SPARC we have a really good business now with Hana where Microsoft Accenture and us are basically running Hana workloads through this architecture that is correct yep so Microsoft's SOP harness that service is is based upon our or equipment completely or liking it at a lab that cannot be part of it no it's this is this Appy Hana service that they're rolling out globally so you know these are services that we can run and we can basically provide performance levels we can provide uptime all these great things that I think are wonderful opportunities for us in the Hyksos gala next to the hyper scalar across by for scalars we figured out a lot of really really good businesses that are exhibiting very very rapid growth I think what's really interesting about that statement is and maybe I'm just dead just flat-out dead wrong right I'm sitting here I'm not standing up there but NetApp has made of business I mean you started this net up has made their business on just core enterprise problems I mean solving really big and wide problems and what you're talking about is a very narrow very specific aspect of that okay and my question is when do you come back to taking those people that you solved all of those really core problems for and help them move forward to the next thing to whatever the next stage is right as they start to embrace cloud I think this is not just is that other end of the barbell right remember I said the bottom analogy there's like transferring big complex existing work loads from on-prem to the cloud that's all Britain but what I was trying to do and maybe I find it too much the other end of the barbell is I wanted to present to you guys that we're doing cool stuff for the young kids right yeah but the punks that I guess that's my point is that you're you're talking you're talking apples and oranges I mean we took apples to people that like apples and oranges to be like ok but no we put it a different way you interchange those freely and those are very different migrant requirements I was given a brief amount of time and I was trying to but what I said oh god no way I'm sensing a concern or a question of when it changes or I blow well like I guess my point is that there there's a big need I would think in the net app base I've been a net app customer and a number of times and I'll say there's there's a fundamental need if I'm sitting here listening to this and going ok so I'm a current net app customer is that up going to be able to help me in the future right I'm thinking about this from a customer standpoint right I look at this and go ok well if that is my specific requirement then sure but I've got a I use net app for broad purposes today across a number of different use case notes how are you helping me across all of those use cases move things forward and that's the piece I'm missing well so maybe I didn't give an credit to on tap I mean on tap cloud is on tap and so anything that you're doing with on tap on premise you can move you on tap flap okay but color me skeptical because that sounds a lot easier than what I believe it truly is I defer to Arthur mark and Dave but well I don't know collaborative gentle in the cloud right because that's not in there of course it's a lot ones they have to be ice cozy if you want the protocols will support the NFS and the this is an NFS whatever version on tap runs it's on yeah but I'm thinking Martha is one of the guys doing that work but the David with what I'm thinking is how you start to move those applications forward it's not just the fiber channel connections it's not just the LUN architecture it's its how do you move those applications now what we're calling workloads but how do you move those forward as I start to embrace cloud and that's a piece that I think changes you from being a storage company to being more of a solution company that it's helping them along that Jerry don't know I'll tell you we've got an Innovation Award contest and I was just going through to judging yeah and and one of the one of the candidates was a university that had been just a hundred percent NetApp on pram customer and they came up with a strategy that they wanted to go into the cloud partly because they they just wanted that different infrastructural model and partly because they were going global they were expanding and that was an easier model for them to be able to support me what they had multiple campuses and students and they basically did a lift and shift of traditional platform to style of applications the cloud into the cloud using on tap in the cloud snapmirror I mean I don't even know what set of tools that people do lots of tools to move and on-prem into a different place infrastructural workload but but having on tap be the foundation though saved a money the way Anthony showed but also it meant all of the tools they were used to using things like Snap mirrors to get data to another place and also the the enterprise-wide data looking around and what's there and who's using what for what tools that we have is there orphan stranded storage we've enhanced that to be both on-prem and connecting into on tap in the cloud right so everything they were used to from NetApp going from one data center to another say they could do that going to the other data center by the way that one's a sure that ones Amazon so I think in their case the particular customer was Amazon but so we don't do it all ourselves but we do partner and then we work with customers figure out what are the resources we need to get in place to help that happen now if we don't a whole bunch of those no not yet that's why it's in our Innovation Award one of the big reasons we do these Innovation Awards is go find clever customers doing cool new stuff and then promote the hell out of the cool ATM I would say that we're moving away from you know the people who we want to I mean we always love our storage administrators is this is historically the business that we have done and we've built our products to make it easy for them to consume it and with what market talked out mark and Anthony I talked about earlier is that the whole idea that we need to broaden our ability for people to consume these storage services in a way that that immediately allows them to have value so like when you were running your net up kit it was in or you were supporting some line of business and they had and they hired you because people didn't know how to how to consume it so the world outside of this large world outside of the net up administrator the storage administrators huge you know the data scientists who don't know anything about net app but they loved they loved being able to have that problem solved with them and the whole idea is that the approach from a cloud perspective is to make our services because there's so many so much incredible technology that Arthur and all the people in gave that have built into on top that we need to be able to tease out in a way that's easily consumable because I want to monetize the heck out of that that that stuff and make it make it a you know I want people to to be basically people we've never even talked to able to consume our yeah become in command lines into an on top administrative interface is not the way any cloud first person wants to forget about it we've done some interesting hackathons where we've built api's for some of our capabilities like data replication and and data cloning and giving it to the hackathon folks and said we what would you do with that right there oh I could use this to get it into a separate data center that would be cool for distribution and or you know the cloning I can do sand boxes for people to play with so we're doing experiments with how to take all of these years of value add and put them in you know a DevOps API quick that model but then getting back to the sort of difference between we sort of talking about a legacy lift and shift of data which is one thing and I as this is Cal field day we we very much want to focus on the cloud native stuff and I'm sort of seeing a disconnect of I don't see people necessarily running ec2 instances with EBS back storage which this NPS is is doing in a better way well actually that's an interesting thing you you not comparing the NPS with with EBS but with the s3 which is also different variety cloud native applications being built not even with containers but looking further forward with server lists where there isn't a date a data store that can be shifted around between applications and I'm just I love the whole story in the beginning about the cloud kind of thing and I'm sort of now sort of worrying that we're not quite getting there and I'm missing the install trend we going back into the legacy thing and I'm that worries me think about EVs is POSIX applications so anything that's a file service like like safes I scuzzy anti-fascist POSIX and what you're talking about is event-driven micro service architectures or even 83 even in OBS threes you know it's not POSIX yeah alright so the idea is that I think with regard to storage gap web-scale some of the areas around what what those guys are doing is just I you know I was on a WebEx yesterday and our day before yesterday and there was some stuff that just blew my mind with regard to that that the idea can we can we talk about and well anyway so anyway so there's some stuff that they're doing that would we wouldn't allow to be up our surge good web-scale engineering team okay okay because we we offer an s3 compliant object store that would allow like you know like we regard to lambda functions being able to be used against that kind of stuff all right this is the area that we want to go in that's that it's highly exciting can you imagine like an s3 compatible data store that that is at the edge of let's say an azure cloud and people want to use as your functions against s3 I mean there's there's things or that's that we would do we're doing really blind blowing things like offering s3 inside of a sure I mean these are the these are the kind of things that that kind of can because we're independent like what mark talked about where the Switzerland we can do kind of things that structurally these other companies can I want to audit overflow check there's fifteen minutes before launch it will have more free form at lunch you may have little more stuff and I had some stuff I wanted to share no I mean my knee my knee I guess my closing remarks probably to as well as building services ourselves that will be representative met up services we are increasingly building services that will be first party services on the hyper scalars now I can't tell you any more than that you want what that means I didn't so I just use the third-party first party metaphor so we NetApp in the Amazon Marketplace or the azure marketplace are a third party we layer our services on top of the same services that everybody else can layer on as a first party you are going to see services largely branded by the hyper scalars themselves that are NetApp services directly accessible from the consoles but net up in finally up underneath there okay and we will be providing the availability and the service of those services expressed directly through the consoles but that won't be native your the the cookies wouldn't have any idea that that correct meta well correct it's not secret but it's not it yeah it's not well know the choice building well now yeah I need it serviced well the nice thing is they will go and they will purchase a particular service and then your and I think it's fair but when I joined this meeting and Marc was up here you guys were like you guys don't get the cloud you guys you guys are going to how can you even survive in the cloud and thank you now we're talking about how do we survive in serverless environments which is the next step which is like I mean trust me you know containers it's a huge opportunity so to me you know if nothing else I hope we've given you some sense of our credibility in the hyper scalars with meaningful solutions that we're building that's all are probably a relatively small number of specific use cases today but you came here probably not knowing we did anything three setting expectations here and so we went from zero to low awesome but I will tell you you know in light is good how can you that there's so many things that you're going to see pumped out of this my organization over the next weeks and months that the business is starting to you know operate on on a velocity and acceleration curve that are very cloud like so the message is wake up this is a new net up the pieces aren't necessarily in place yet but there's a whole bunch of stuff that you're probably unaware of that we're doing already it's an a whole bunch of so it's like a jigsaw puzzle and some of the pieces we think I've found you know that we found the corners and we found the edges and I think the nice thing is we kind of know what needs to go in this gaps usually at least have a framework you showed the 2000 customer number you didn't share any anything about the number of petabytes we've know a little more than 40 yeah well I'm just saying so so I think like anything else it's still an emerging category the cloud is still there are sort of some very well-defined areas there are some lesser defined grayer's and there's still white space and so I think you can think about our portfolio like that we have some very well defined repeatable value propositions that we can sell the crap out of we have some we had a constant stream of new use cases and new services cloud control only came out months ago I mean I mean we're releasing very very badly 365 and time I've dealt or wherever else I mean that one I'll get lots blood coming here there's a lot coming and I think like I said you're going to see more and more announcements that I think we present to us not only with credibility in the cloud but improved credibility through the relationships with the hyper scalars and credibility with the new buyers

Channel: Tech Field Day

Views: 3,982

Rating: 4.6923075 out of 5

Keywords: Tech Field Day, Cloud Field Day, Cloud Field Day 2, CFD2, NetApp, Dave Hitz, Anthony Lye

Id: AOs6Q-nHuu0

Length: 62min 45sec (3765 seconds)

Published: Sun Jul 30 2017