>> Announcer: Live from Gillette Stadium in Foxboro, Massachusetts, it's theCUBE, covering #VTUG's New England Winner Warmer 2017. Now, your host, Stu Miniman. >> And we're back, with SiliconANGLE Media's presentation of theCUBE, we're the worldwide leader in enterprise tech coverage, happy to welcome back to the program, Steve Pao, who's the CMO of Igneous Systems, Steve flew out from Seattle here to, welcome to the home of the New England Patriots. >> Oh my gosh. Number 12, number 12! >> The 12 man representing here, you've got, I have to say, I almost canceled my season tickets when Pete Carroll was our coach, so, luckily he's worked out better for you than he did for us. My wife's a Brown fan, she says the same thing about Bill Belichick. So, it's the coaching fraternity is kind of like the tech world, it's a small group, you all kind of get to know each other and move around, so, thanks for joining us. >> Yes, well thanks for having me! >> Alright, so Steve, we've been talking to you guys since you were coming out of Stealth, why don't you give our audience, what's the update on Igneous? >> Okay, well for those of you who don't know us, what Igneous really does is we offer an onsite private cloud storage service, and that's our first offering, it's part of our greater mission of providing true cloud for local data, and what we basically offer today is an unstructured data store that's completely delivered as a service, we take our own equipment, we install it, we monitor it, we manage it, we even refresh it when necessary, and all the customer has to do is really subscribe, and that's it. It's all pay-as-you-go, and it's all zero touch for the customer. We launched back in October, as you recall, and one of the things that I think that it's been really great since launching is that we've really started to see how customers that didn't know us are actually really evaluating, really, I think, the convergence of two trends, one is there's this data growth, data trend that goes on, and pretty much everybody we talk to, citing data growth rates on the order of doubling every three years, where IT budgets are growing less than 5% a year, so there's this mismatch where, basically everybody's hitting this juncture that what they used to do can't work because the data's growing faster than the budget. And at the same time, there's this data growth that's actually happening, and the data growth is not from relational databases and structured data, but rather, a lot of new applications that are logging sensor data that are supporting machine learning, AI, really, it's machine-generated data being analyzed by machines, with humans really just training the AI and the machine learning. >> Yes, Steve, I want to unpack that a little bit, let's talk it, because many of us that watch storage, it's been like, well the storage industry, it needs to change, it's not about selling boxes, it's not about capacity, and even on unstructured data, it was kind of like, okay well, what's creating data and what's actually valuable? How much is it just, do I stick it on a cheap tier, what do I actually do with it, what's interesting you guys do, some of those use cases, throughout machine learning, machine data, things like sensors, every time I hear that word, that IoT buzzword kind of pops into our head, but maybe you could talk to some of those, what's bringing customers, what's that driving challenge that they have, that you're helping to solve, that's different from the way storage has been done for many years? >> Yeah, I think, that's a great question, and I think that there's just been a real transition, and I think the transition has been largely created by the kinds of data that we want to manage and that we want to curate, and as we're seeing these sort of large unstructured data sets, it starts with the data, so as an example, you take equipment that used to exist in the past, like let's say in scientific computing. You used to have flow cytometers, which were just time-series data, and then what's now happened is is associate with ever flow cytometers, now a real-time video feed. When you look at the old world of microscopy, what you used to do is you used to flash freeze a sample, and basically take a picture of it, and now what you can do with lattice light-sheet microscopy is you can actually look at cells in vivo, while they're alive, and you can, I've personally gotten to watch a T cell move through a collagen matrix, and that's all microscopy, but generating orders of magnitude more data. That is, we're looking at these very, very different data sets, we're looking at very, very different kinds of computing, and what that requires is very, very different kind of infrastructure. And so, the infrastructure has just had to get a lot more intelligent, and the architecture has had to get a lot different, and what we've noticed is is that, that a lot of the patterns that are actually being built in the public cloud as they've taken kind of a fresh look at the computing models, have really become appropriate for this new kind of computing, and we don't see that on the premises, and that's really what we set out to go do. >> Yeah, it's interesting, it's probably the wrong term, but it sounds like we're describing kind of object storage 2.0. 1.0, I remember this healthcare use cases, everybody, when I was doing radiology, when you're doing certain healthcare and sciences, I need metadata, I need to understand that, but now there's just orders of magnitude more data, and technologies are making, it's denser, prices have come down, so the idea has been around for a little bit while, but it sounds like the technology's matured to allow kind of an explosion-- >> Well, and it's just a computing model, it's like one of these things where we're really, because of the emergence of microservices, one of the things that we've seen is applications want a restful interaction with the storage layer, and so, so it turns out that that tends to be very, very perfect for a cloud-like implementation where you can actually implement high volume, unstructured data really, really well via a restful API, where in the old world of POSIX semantics and that kind of transactional model, you just lost scalability. Either you had a lot of proprietary hardware, with that VRAM, you had proprietary interconnects data with things like InfiniBand, and nowadays, being able to loosely couple distributed systems is really the name of the game, and that's ultimately what we aim to build at Igneous, and that's all the technology, in terms of our commercial offering, the customer doesn't care what's behind it, but fundamentally, what you're looking for is the scalability and resilience that the cloud offers by doing that on premises. >> Yeah, so Steve, we had a really interesting crowd chat about a month or so ago talking about hybrid cloud, and the thing I've been saying for the last, probably year, is, as customers try to figure out what goes where in the cloud environment, you know, I've got SaaS, I've got public out of, I've got my private cloud, it's follow the data and follow the applications. In the cloud, things like mobile and even some video streaming, I think we understand how to do that, but why does on-premises make sense for your customers, your workloads, and your solution? >> Yeah, absolutely, and so, first, a little bit on hybrid cloud, there are kind of two different definitions of hybrid cloud, one is kind of the AWS VMWare scheme where what you're really looking to do is run your old stuff that you were running on-prem, in the public cloud, and you call that hybrid. But there's another way to look at it, which is to say, hey, let's take a look at the computing patterns that are being run in the public cloud, how do I bring that down to the premises? And the reason that you might want to do that is, it's really twofold, one is the gravity of the data, so it might just be that the datasets are too big to move back and forth over very thin internet pipes, and so you want to actually keep the data close to its source. The other is something that we've seen, which is really more of a preference, which is that while I think that cloud technologies actually have a lot of capability for security, there are a lot more hoops for folks to run through to ensure that they're compliant with their own internal policies, and where they've already set out a set of policies for how they run the stuff behind the firewall, sometimes it's just simpler for them to actually keep all of the data on the premises, and not actually have to worry about some of the issues in tracking, and compliance issues associated with how you move the data around. >> Yeah. One of the things we've heard from users is when they use public cloud, one of the things they really like is, sometimes the CFO's not fully onboard, but buying things as a service, so, they want to understand predictability, but they want to buy it as a service, understand, how does your solution fit into that kind of paradigm? >> That's great, I think our solution fits into both trends really, really well, because what we're really offering, we talked a little bit about technology, but really fundamentally, we're offering a service, and so when Igneous goes to a customer, our interaction is as a service. Customers interact with our service via APIs, and they get a bill for a subscription, and so it's an as-a-service model, you don't buy hardware, you don't install software, you don't have systems to manage. At the same time, there is a predictability that's a little bit of the downside of the public cloud, because there's a fee, generally, to access your data at a storage, and often, when people don't actually understand their data access and their data movement patterns, the costs of running applications in public cloud become quite unpredictable, and you actually don't run into that unpredictability with a solution like Igneous, because our data is on your local area network, and we don't charge you to access the data that's on your own network. >> So, I've come to an event like this, if I'm thinking about my storage today, the conversation in the marketplace has been, well, the new choices out there is, there's, the HCI, the hyper-convergence infrastructure, and there's flash, the AFA devices out there. And of course, even the lines between those are blurring, because I can have an all-flash configuration of hyper-converged, and some of the all-flasher a things are getting converged and put into more things, how do you help customers as the, what's the bullet point as to, well, this is for this kind of application, this is for this solution, and hey, there's this whole new category that you need to be thinking about. >> Yeah, I think that's perfect, and I think the real trick here is is that there's a difference between your hot tier and your flash tier, and your capacity tier, and fundamentally, the flash tier is really good when Time To First Byte is very important, so that might be for your relational database applications and things of that sort, where there tends to be a lot of searching through an index, and so you've got a lot of low-latency requirements. And then on the other hand, what you have is a capacity tier, they may be your video surveillance, they may be your large, unstructured documents, they may be your censor data, and in those contexts, you don't necessarily need the Time To First Byte, what you really need is capacity throughput, and so the overhead of setting up, for example, a restful connection is not significant when compared to the amount of data that actually needs to go through the system, and that's actually where restful semantics actually gets superior to positive semantics, when you have very, very large, unstructured data sets. Hyper-converge is actually a little bit of a different world, and I think that while hyper-converge has worked out pretty well, I think, for virtualization workloads, we've really found that when it comes to these very, very large unstructured data sets, hyper-converge isn't necessarily always the way to go, you tend to find a utilization issue between your compute and your storage layers, where you have to actually think about how you're balancing all this stuff, and so, really, the world that we've really seen emerge as new applications come forward, is there's really a trend to write microservices that are stateless, and to have them talk to a stateful layer, that's why in the public cloud, there's a pattern of having things like elastic container services talking to an S3, and we definitely see on premises that same type of things that's going to emerge. There's going to be some time to get there, admittedly, as I was mentioning kind of at the beginning, we've seen this really interesting set of interest patterns, one is from the folks who are developing these new applications that are utilizing unstructured data, there's a lot of interest we're getting right now from IT folks that are just getting started with object storage to do secondary workloads, to do backups, to do archives, and it's been interesting that we've been getting a lot of interest in our service as a new way to approach some of these data protection workflows. >> Alright, so Steve, last question I've got for you, came out of Stealth Q4 last year, what do we look for in 2017 from Igneous? >> Yeah, so I think that you'll see it on both of those fronts, I think that one thing that's going to be seen in 2017 is a lot more development on our side around building up a tool chain for folks to use for a data protection tier, and so, we've got a new offering coming online, we're calling it Igneous Insights, which provides information about what's currently on your primary storage tiers, we've got a whole set of replication services, they're coming up to do backup, archive, things like replication to the cloud, but what we're also really moving forward with is a lot of what's needed in the tool chain to really support hybrid and multi-clouds, so how you facilitate the data movement in and out of the cloud, as well how you do the auditing and management of the data, no matter where it lives. >> Alright, Steve Pao, really appreciate you catching up, and if you want to find out more about this category, check out cube365.net/trueprivatecloud, that's C-U-B-E, number 365.net/trueprivatecloud, which has resources from the whole industry, including from Igneous, including from Wikibon and theCUBE, as to what's happening kind of this true private cloud, hybrid cloud environment. We'll be back with lots more coverage here, thanks for watching theCUBE. (electronic music) >> Announcer: Since the dawn of the cloud, theCUBE has been there.

vmworld 2015 brought to you by VMware and its ecosystem sponsors and now your host dave vellante welcome back to San Francisco everybody this is the cue the cube is SiliconANGLE wiki bonds continuous coverage of vmworld 2015 this is our sixth year at vmworld we go out to the events we extract the signal from the noise our friend vish mulchen is here with HP storage fish it's always good to see you you know in your hometown and of in the backyard it's great to be in moscone thanks for coming back on the cube thanks Dave great great to be here as always again so we have seen you know I go back to 2010 and at the time you know 3par was a separate company and then we watched the the acquisition occur I'm really badly needed that acquisition was we were at vmworld when the the bidding war was occurring between Dell and yeah that's right and we predicted HP he's going to win that war and of course that changed change the course of the storage at HP you know permanently yes so it's been an amazing run three pars become the crown jewel of the of the portfolio but the most amazing thing is how you've evolved that platform into play into the all flash world very very competitive product so so we've been sort of documenting that that traction but give us the update let's talk about sort of where you've come from and you know where we are today sure sure Dave so I mean if we look back in june 2013 when we first announced the AFA right and since jun 2013 we've had a fire series of announcements in in december we announce something called adaptive sparing which you know was actually very unique flash innovation treating the flash separately giving customers twenty percent more capacity in jun 2014 we brought two dollars per gig deduplication in december 2014 we brought the constable we call the converge flash array right flasher a flash focus design but hey you can add spinning this to it if you want right and several of our customers are actually doing that because they have a need for that and then in june of 2015 we double down right and we announced the 20000 series we brought the affordability even better to a dollar fifty a gig and that was in June so the other amazing thing is the pace of the cadence of announcements I mean I had to say I mean remember for years you know HP the announcements were very slow to come up maybe have one a year maybe you know maybe a name change but now it's like bang bang bang I presume it to the architecture that allows you to do that but a lot of skeptics when you came out yes with the all-flash right yeah it's going to be a bolt on you said no you know NASA died we'll see but now you're proving it why help the people who sort of don't understand the nuances how was it that you were able to do that and what are the proof points that it's not just a bolt on right so you know I think the it all comes down to the architecture right you have to have an architecture that's modular that extensible and you know as we looked at the three Power Architecture all the attributes that we put in place early on we're very applicable to flash now flash did have some differences and we did account for some of the differences in the architecture but the architecture proved to be able to be extensible and a lot of the tenants around scalable controllers for performance the ASIC to offload the fine-grained virtualized operating system with a very small page allocation size all of those fundamentals were perfectly suited for flash right and and you can almost probably say they were there were too much for spinning disk right why was to say was that just was that luck because a bit but of a lot of what the original designers a three-part did were trying to most of it was trying to offset the deficiencies of spinning disk yeah you know did they just have like amazing vision or was it just I give I give the founders a lot of credit for their foresight and in fact if you look at the founders and I spoke to them they were they had a server background and they started right and they said its own server guess I'm sorry guys say they said to me wish when we did a server benchmark it would take us six months four and a half of those six months was getting the storage right and they said they really don't understand why it had to be so hard right and I think they've brought a very different approach to storage to how sort of the industry was handling storage right it was it was very different it actually turned it on his head and they are actually architected some very interesting capabilities which you know I'm very confident as we go to flash 2 point 0 as we talk about other newer non-volatile memory technologies if nan something other than man comes about you know I'm very confident that the architecture will be able to gehen to isolate the media from the customer Martin fake hope said that's member stare but we'll see we'll see what whatever gentleman sorry you know we'll go about the industry members has a big element there but we'll go what the industry wants to look at of course so let's talk about vmworld 2015 what you guys are doing here you know sure emphasize the announcements that you're making talk about that a little bit sure so in vmworld we had several announcements we made what i'll focus in is on the flash announcements and you know if you look at the approach we've taken with flash we've had three vectors right affordability performance and data services and some companies have done one or two but i think it's rare to see all three vectors being attack of the same time and that's been our approach from the start and all the announcements we talked about and in this announcement that we made this week same approach so let's let's maybe go down those three vectors Dave if you allow me to yeah please okay so so let's start with affordability and we announced a new 8000 series which is a refresh to the 7000 line right a very successful 7000 line of which is 7450 flash arrays one of them now the starting point for the 8200 the old flash 8200 now is down to nineteen thousand four hundred ninety seven dollars two controllers six drives six terabyte usable capacity 19,000 999 we're under 20 grand by a lot we make sure you got that 497 right so that's great we also announced then a lower entry price point to the 20000 series that we announced in earlier in June those were as your call aight controller systems we announced a lower price point 2450 a 4 controller capable system as well again on the theme of bringing affordability right driving the price down okay so you have dollar fifty per gig if you want to buy that way if low entry price point with 19 k if you want to buy that way or if you want a scalable system that you can grow to the extreme you can buy affordable price point that way as well right so in my mind the the adoption the success we've had in the marketplace has been a function of a couple of things affordability is a key one right it's economics that's what drives adoption okay now your performance everything's okay let's flash over he's got the same performance is high performance now it's somewhat true because relative to spinning disk it's gonna be you know better performance but there's it's nuanced so talk about your performance yeah so performance is very important we announced a couple of interesting performance first we talked about some some improvements in bandwidth now let's take a look at sort of why that matters right Dave so if you were doing a million iOS and there were small 4k blocks do the math it's four gigabytes per second now if you're doing large block iOS like if you're doing a sequel database query analytical query those are typically large block ayos right we do a million of those and there Sarah megan size then that bandwidth becomes a choke point to the array so we've announced with with the 8000 series you know twenty four gigs the second of bandwidth which is two and a half times more than so but this is ever saw it started erupting but this is why a lot of the existing arrays that bolted on flash failed what yeah so one of the reasons why they fail is their controllers are not able to handle the IO load and once even if they do can they handle the bandwidth requirements and then you know here's the other thing that matters is the latency right so the other thing we announced at the this week was a forty-four percent improvement in latency soumillon I ops 387 microseconds latency Adam denials that's just low latency so you're setting up this little latency storage versus capacity storage right and you got you playing both but we're obviously talking about the latency piece here okay correct so that's the performance piece and then there's there's there's actually two more there's the availability which answer this well free part is known for high availability and it's the new tier one yeah yeah so and it but there's data services associated with that yeah so the resiliency is a big factor there and you know there's single system resiliency pull out a drive pull out a controller fail a cache board how do you react right in fact the reason why we succeed in the marketplace that our customers tell us is that reliability factor and they go and they have these tests where they pull things in and out right and they watch how the other arrays operate right and you know consistently we've come back really operating well in the area of single system resiliency now there's also a multi-system resiliency which is what do I do with replication what I do with snapshots can I move my snapshots to addy duplicating backup device all right how quickly can I move how much do I move so I think there are all of these elements that you look at resiliency that I think important that's another piece and resiliency that's coming up as well emerging Dave and that's around protecting the access to your data security do you encrypt the data so now if you encrypt the data and you have a snapshot and you move that snapshot to a duplicating device what happens to that snapshot and the key do you have to a multiple key so your keys get compromised so that resiliency topic is a big one lots of different areas to go off go after and whether it's replications snapshots backup devices encryption key managers we have all those elevators well how about so again one of the we always talked about this one of the big advantages of an architecture that's been around for a decade is is you've got the stack it's hardened you know that sets the storage services so that's that's a big differentiator from what you see in a lot of the startups yes and and or the bolt ons which everybody thought you're going to be both on baby architected the whole thing so that's cool what about quality of service what about the ability to sort of address quality of service to pin application performance and to actually change that programmatically yeah so quality of service is a very very big big attribute of ours in fact week the product for full HP three parts called priority optimization and in this week's announcement we announce further enhancements first of all we have latency goals on our queue as product which i think is unique nobody else offers latency goals and this week we announced the latency goals going down to half a millisecond I mean if that array is operating at you know three to four hundred microseconds you want to be able to control your priorities with that granularity right and so qos granularity is exactly what we brought and you know Dave Lee if you remember when we did the last cube we talked to the cloud and they they had taken a gold silver bronze tier hardware-based and then put her on a flash array and put priority optimization to implement in software the gold silver bronze right yeah the cloud is a company music louder company yeah so that's right and that was interesting to see that they did that with with flash right you know yeah exactly yeah what do you think is going to happen there right is a worship we're hearing increasingly it shows like this and others that that you're starting to see more tearing and flash you're hearing it now in in the hadoop world and big data world the example that you just gave a lot of people initially and maybe still think you're going to have flashed here in the latency tier and you're going to have the capacity to air the bit bucket what's yours what you're thinking now on how that shakes shakes out and how practitioners should be thinking about their storage architectures going forward and I think you were gonna see the variety of that I think that's one very possible use case which says hey I have a applications that are critical service optimized service level optimized right that got to be on flash and then I may have either a backing store for time or I might have another set of applications that are not service level optimized more cost optimized may be right so and maybe that changes over time maybe it changes by quarter what is cost optimized today needs a spike and come back so this notion of data mobility I think it's very key right and sort of the fourth data service pillar I want to talk about because we announced for wave Federation which is the ability to take for arrays and operate as a single logical hole and you can federated Atta among those arrays now but if you extend the ideas can you federated to a backup device can you fed rate it to the generic cloud right can you federated to an archived here I think these are the kinds of things that our customers are asked that's right they want a first of all federal rate to another array to work load balance for example Oh asset refresh right but all of the other use cases federated a cloud federated to archived here those are all coming up alright so I suspect we're gonna see more of those as I said can I and let's stay tuned state-owned you know I mean as we as we look to to raise the bar once again these are some of the things what we're thinking about all right so so I know you can't give details but give us high level road map what should we be thinking about watching you know HP generally 3par and all flesh specifically yeah so I think we'll continue to drive a affordability right 3d and 3d Nance available as well now there's other flash technologies and you know we want to isolate our customers from whether it's CML CML see 3d 9 i'm gonna say to them look what's your price point what's your capacity point what's your availability point ok and we'll meet that let us worry about that technology problem out there how we get there so that continues to work us on you know the media faster controllers again to drive up the drive of the performance hosting the connects you know there's a lot of talk around the role of 25 gig Ethernet 32 gig fibre channel the RDMA technologies right I sir are I war rocky so there's all these things here nvm e to the backplane nvme to the host so you know flash j-bot so look at yeah it's shit we're shifting the bottleneck are we are you going to look at the bottlenecks across all areas into n and make sure that you're looking at this holistically right as you drive as you drive forward doesn't get less complicated but at least for the for the for the guys who are building this stuff hopefully for I we who are using it it does but fish motion thanks very much baby greater pleasure always pleasure sir I keep right there everybody will be back with our next guest this is the cube we're live from vmworld 2015 in moscone we'll be right back you