Grey Beards on Systems - GreyBeards talk with Pivot3 and NexGen Storage about their recent acquisition announcement

Episode Date: February 9, 2016

In our 29th episode, we talk with John Spiers (@lefthandsan), Co-founder & CEO of NexGen Storage and Ron Nash (@hronaldnash), Chairman & CEO of Pivot3, a hyper converged infrastructure provider.  We have talked with John before (see last June’s podcast episode) about NexGen Storage technology. Recently, Pivot3 announced they were going to acquire NexGen Storage and Howard … Continue reading "GreyBeards talk with Pivot3 and NexGen Storage about their recent acquisition announcement"

Hey everybody, Ray Lucchese here. And Howard Marks here. Welcome to the next episode of Greybeards on Storage, a monthly podcast show where we get Greybeards storage and system bloggers to talk with storage and system vendors to discuss upcoming products, technologies, and trends affecting the data center today. This is our 29th episode of Graybridge on Storage, which was recorded on January 29, 2016. We have with us here today Ron Nash, CEO of Pivot3, and John Spires, CEO of NextGen Storage.

Why don't you tell us a little bit about what's going on between your companies, John and Ron? Hi, this is Ron, and John and I are very excited in that we have just announced that we're going to combine our two companies. We each had what we thought was a pretty exciting company in the hot market, you know, which is what you live for in our industry, to have something that's moving fast and doing well. But when we met each other and talked a few months ago, we sat down and looked and thought more broadly about all the things that you're going to need in the software-defined data center to take it forward. And we mapped our

two companies into strengths and weaknesses, and there was an incredible combination of the two, and very supportive. And we're going to be able to bring out a lot of exciting products just by combining what the two of us have, as well as some additional innovation on top of that. So we've decided to combine and we're excited to announce that today. Gosh, that's great news. Ron, can you tell us a little bit about what Pivot3 does? Pivot3 is in the hyper-converged infrastructure space. And so one of the big waves that is crossing through and running through the IT world right now is this wave of software-defined things. Because if you think about it, all the functions of a data center used to be pretty well siloed. You had processing and storage and networking and security, and they all had

their own software. They all had their own hardware that they ran on, and they all had their own chips that they ran on. And starting a few years ago, some customers started doing innovation, actually, instead of vendors, which is kind of unusual in our industry. And Google was one of the leaders of that. Instead of having all these disparate parts of hardware, software, and semiconductors, the wave came to say, let's standardize on x86 commodity price servers as a base hardware platform and chip platform. And then let's put software on top of that that enables it to act like a processing unit sometime, a storage unit, a networking unit, da-da-da-da-da-da. And that's the wave that's crossing our industry. And so Pivot3 is in the hyperconverged infrastructure space, taking that kind of stuff, delivering it into enterprises to help them change their data center around to a software-defined. Now, the reason that you change that is the cost is dramatically different. Let's say half is a reasonable estimate of what you can achieve.

And secondly, it's much simpler to operate and easier to operate for people. So we took out complexity and we lowered cost. And so that's a big thing for people. Hey, John, you've been on the podcast before, and we've talked a little bit about next-gen storage, but if you want to just give a highlight of what next-gen was doing before the acquisition, that would be great. Yeah, you guys know the history.

We started the company in 2010. Kelly and. Yeah. You guys know the history. We started the company in 2010. Kelly and I were both former founders of Left Hand and pulled the band back together to start NextGen. And we raised some venture capital back in early 2011, got our first product to market and generating first revenues when Fusion.io came and swept us off our feet. And Fusion was a flash components company looking to broaden their horizons and build solutions, software solutions with their flash and take those into the channel.

You know, fast forward 18 months and SanDisk was looking to add more flash technology to their portfolio and acquired Fusion.io. They quickly realized that the next-gen product line, which was a hybrid flash system at the time, competed with SanDisk's OEMs. So we worked on a very friendly spin-out agreement and SanDisk actually helped finance the spin-out

and we got us back on our feet again and we've been running independently for a year. We've been out raising a venture capital and we received our first term sheet. What was it? Probably four months ago now from Brian Smith at S3 Ventures. And we were about to close on our financing round when Brian and I were talking. He says, hey, I've got this other portfolio company over here called Pivot3. And the more I learned about you guys, there's a lot of synergies here. You're looking to expand your sales force and, you know, they've got, you know, 26 sales guys that you don't, you know, expand your marketing and they've got a

broader marketing team. You know, they have 10 people in Europe. You have one, they have people, you know, in the Middle East and Asia and Mexico City and UFZero and looking at their channel too, their 100% channel and at a high level, you know, and then, so I said, it sounds appealing. Let's, you know, I'll get together with Ron and we'll kind of go through everything and see where there's overlap, see where there's synergies. And, you know, when Ron and I got together and looked at everything, there were synergies everywhere. I mean, I could probably talk about it for an hour, but on the technology side, as you know,

NextGen has a hybrid array with Flash and Disk. We have an all-Flash array with different types of Flash, and we leverage our quality of service to dynamically tier data across those different tiers. And then Pivot3 is in the hyper-converged market with a scale-out hyper-converged system. And as you know, hyper-converged is servers and storage kind of combined in a scale-out architecture. If you look at NextGen's roadmap, as you guys know, we have scale-out on our roadmap. It's one of the things Howard asked about last time he was in Colorado.

You know, what's your plan for scale-out? Well, now Pivot3 is our plan. Oh, okay. The good news is, you know, what's your plan for scale out? Well, now Pivot 3 is our plan. Oh, okay. The good news is, you know, Kelly's done this before. Yeah. And Pivot 3 has erasure encoding scale out. You know, if you remember from the left-hand days, one of the things we used to get penalized on is we used to replicate. We had one, two, three-way replication. And each time you cranked up the resiliency with replication, you cut your storage capacity in half. And so that you cranked up the resiliency with replication, you cut your storage capacity in half. And so that capacity utilization was a big thing. And we started working on what

was called parity-based network rate at left-hand, and we eventually got that to market. But Pivot3 has had that and it's been shipping that for quite a while, and they have it in a hyper-converged configuration. So you're running your applications on the storage and VMs, and you have it in a hyper-converged configuration. So you're running your applications on the storage and VMs and you have the scale-out storage architecture underneath it. And it's very exciting. And so we're looking to add that capability and features to our storage system offerings. We have quality of service, which dynamically tiers data

across multiple storage media types and delivers service levels to customers. That's something that's very appealing on a hyper-converged system, because if you look at Nutanix and others, they don't have anything like that. And it's a big problem in that market where performance inconsistency is still a huge issue in being able to deliver service levels, especially when you're running all those apps and doing storage functionality on the same system. Being able to isolate workloads and dynamically tier data is a huge thing. And so when we looked under the covers of how to implement QoS on Pivot 3, it was a fairly straightforward thing to do.

And so we got really excited about that. So at a 50,000- foot level, we're in three fast growing market segments, not a lot of overlap. And from a technology standpoint, there's a lot of cross-pollination things we can do to make all of our products more competitive in the marketplace. And then the whole go-to-market sales engine is quadrupled for both of us. And we're very excited about that. From a hyper-converged perspective, there's servers, storage, and the networking aspect. Ron, is there something you guys do specifically for networking? We provide a platform for networking. What we found a lot of times is a

lot of people, rather than wanting networking to come with a package, they want to use their other networking. So it's one of those cases when they bring hyper-converged in, they've already got the networking established. So a lot of times we just end up working carefully with other people's networking. In small cases, we can bring software-defined networking in or something like that. But when I look at the majority of cases, people already have established a network. They're unwilling to change. I guess it's a philosophy thing. And that's one of the reasons we like this quality of service was a lot of inventors think they're going to revolutionize the world and everybody's going to use everything of theirs and everybody ought to immediately throw away everything they have and start using this new

stuff. Well, we try to have a more mature and long-term view about that in that we recognize we're going to be in a data center. We've got to coexist and we've got to give people a path to transition from the data center of today to the data center of the future from an architecture and technology standpoint. And that's one of the reasons we saw the quality of service that NextGen had. if you could extend that, not only over NextGen's storage, but over our hyper-converged infrastructure, also extend it over cloud and extend it over legacy storage. So if we could do those type of things, we could provide a much better way for somebody in an orderly manner to transition to the software-defined data center. But it's just less of an ego thing of

change everything to my stuff and more of a pragmatic business view of we'll give you a way to get to the future. You start to use our stuff and you'll buy more of it over a period of time as your old stuff ages out. Leaving aside the difficulties of applying QoS to somebody else's storage or to the cloud, my first reaction was that hyperconverged systems really beg for QoS because the sheer number and types of resources that become constrained become so large. By sharing the same CPU with customer VMs and the storage stack, the ability for one to impact the other just becomes enormous. Correct. That's absolutely the case.

And, you know, another thing we did that some of the hyperconverged other ones don't do is our software runs on a line of servers, but allows you to manipulate the configuration of the server. So if you have an application that is more processor intensive and want to put more memory in, you can do that. If you want flash drives, you put it. If you want larger hard drives, you can do that. Forever in IT, people have manipulated the hardware configuration such that software will perform for different applications. And that's one of the founding tenets.

Isn't that our whole job? It is. It's absolutely our whole job. We build the infrastructure to make the applications run better? That's correct. And that's why we run over pizza boxes, one use, two use. We run on blades. We run on various other kinds of dense configurations. Don't get me started about hyper-converged on blades, Rob. Well, not many people can run on blades because they don't have the storage efficiency. And that's one of the things that kills the performance for a lot of people. It's just a place they can't go to because unless you have something like our erasure coating,

you're not going to have the storage efficiency to run applications on blades. At this point, I have to come clean and tell the world that, you know, both Pivot3 and NextGen are clients of deep storage. So I do have some inside information. Yeah, that's right. And, you know, and the world just has to know that I do also have some ulterior financial interests. Yeah. Hey, John, there's a question that I do also have some ulterior financial interests. Yeah. Hey, John, there's a question. I've always assumed NextGen was a hardware-based storage product. Although, coming from left-hand, you certainly understand the software-defined stuff. Is there a software-defined version of NextGen that I'm not aware of? Well, we run on standard off-the-shelf super micro gear, somewhat specialized because it's a storage bridge bay. And there's not a lot of white box storage bridge bay systems out there.

But if you look at the storage bridge bay, it's just PCIe or 10 gig interconnects between CPU domains, which can be internal in the same box or external. So yeah, our software runs on standard x86 servers too. So if you took two servers with storage in them, you could make a next-gen system. Our software would just run on that. Coming back from our left-hand days, and if you look at what Equalogic and others are doing in the space, and even Nimble today. So Nimble uses the exact same platform, Supermicro platform we do as well. And it's standard off-the-shelf stuff, but what makes it unique is it's just all in a single box and it's plug and play and it's fully redundant. There's nothing

to say that you can't have two independent boxes that are plug and play and fully redundant as well. So we can run our software in a virtual machine today. In fact, a lot of our software is tested in a virtual machine and not actually on the hardware. So it is very adaptable, and most of it runs in user space and those types of things that make it very flexible. Yeah, there are good reasons to be software-defined but not software-delivered.

Exactly. Not the least of which is the support issues. Right, right. You don't want the customers to be afraid. I bought these servers on eBay, and now they crashed, and I lost all my data, and it's your fault. That's right. Yeah, we were thinking about installing a shrink-rect machine, but decided we wouldn't go that way. Hey, Ron, one of the key things about HyperCoverge is the virtualization solution that's in use.

You didn't mention whether you guys are using, you know, which one you're using in your solution. Is there a particular virtualization software that you guys use? Well, we have used Zen, and now mostly we're using ESX from VMware because our customer base are enterprises. And that, you know, they have the dominant market share in enterprises at this point in time. So we just want to be friendly for that. There's only one capability that a hypervisor needs differentially, and I think most have, but there are a couple that don't, and that's we need the ability to pass through the hypervisor to run the storage. Otherwise, we will pay the hypervisor performance tax of 30% to 40%.

And so we avoid that. And VMware gives you an opportunity to do that. And so that's the one thing we need that is crucial from a performance standpoint. So we get NextGen's QoS and Pivot3's erasure coding. Right. Running on a series of servers over a PCI fabric so that the network latency of the erasure coding doesn't matter. Uh-huh.

Oh, this is starting to look very interesting. So now you're at the same point that john and i were when we first had lunch about 15 minutes into the conversation we got to the same spot it just the light bulb went on and we said wow this can work the pci fabric here is somewhere is it's between the servers is that somewhere in the future ray oh in the future it's a future solution okay but the problem with erasure coding in this in this kind of hyper converged environment is you either need to have a really serious log structure so that you're never doing in-place updates or the latency of collecting all the

data to recalculate the parity to do an in-place update starts to get hairy but if we start saying there's an rdma fabric even if it's rocky underlying it then a lot of that latency goes away and now the erasure coding for low latency type reconvergence scale out starts making sense and you make it you know pcie switched kind of latencies and almost all of those you know disadvantages of being disaggregated go away yeah and we'd love to have you uh do an nda whiteboard session with bill and kelly to talk about yeah that's going to be fun because Bill and Kelly individually are both fun. And being in the room with both of them could be very interesting. For the audience, Bill here is Bill Galloway, the technical genius and founder of Pivot3,

and Kelly Long, who holds that role on the next-gen side. So yesterday we had a really fun session in that we announced in Colorado. And then John and I and Kelly flew to Houston, where Bill was, with a lot of our developers. And we had a whiteboard session at which I spoke for a minute or two, and John spoke for a couple of minutes. And then Kelly got up, and he and Bill were standing there, and all the developers were there. They spoke for a minute or two and John spoke for a couple of minutes and then Kelly got up and he and Bill were standing there and all the developers were there and they spoke for three hours. So it sounds about right. It was a love fest down there. Good. Howard, you mentioned the word log structure. Does NextGen support a log structure in the back end? Yep. And then Pivot3,

did you guys always have a log structuredstructured sort of backend as well? I'm not sure I can answer that. Okay. Well, that's all right. It's not as obvious in the Pivot3. It's very complementary from a technology standpoint. We're both Linux platforms. And serendipitously, a lot of the way we've developed software is very synergistic and works well together.

It's not true in all cases, but in a lot of cases, you know, some of the stuff we're going to combine are fairly easy to do. Yeah, we obviously work very, very hard to minimize the crosstalk. Because we were doing a distributed, and Howard's absolutely right. If you get too much crosstalk on the side going out of thing, it could just kill it. And so to get the performance we need, we worked long and hard on that. But the PCIe stuff will help. That'll be the next step forward. Yeah.

And, you know, no offense, John, but the left-hand network RAID was not a big performer on small IOs. Not at all. All right. Well, we won't go there. Gosh. Well, you know, now there's the really important questions, you know. It's like, so is this going to be, you know, pivot four? We actually talked about that. Lee Caswell once explained to me what the three pivots were, but, you know, I'm sure you could add a fourth one.

There are a lot of claims as to what the three means and then, you know, a lot of different ways of it. So we'll see. But for right now, we're going to leave it at three. You know, three seems to be a good number for us. It's worked so far. We'll hang with that for a little longer. Okay. What's the timing on the acquisition? Is this going to close pretty quickly or? Yeah, I think it'll close pretty quickly. It is uh we have signed a definitive agreement to merge both board of directors have approved it so we're past all that so now we're just in the final legalese so now you just got to wait for the antitrust clearance yeah well i don't worry about that we're not yeah we're not quite we're not quite there we're aspirationally waiting for that, let's say.

John, you don't have to wait for the SEC, but, you know, the antitrust stuff could be private companies. Yeah, that's right. But I think it'll close in probably sometime in the month of February. You know, that's what we anticipate. So it's pretty quick. And then if I remember right from the release, there's a funding round that comes shortly thereafter. We have the bulk of the funding lined up already. We did not announce that when we announced that the companies were coming together. And so we'll have a subsequent announcement about that funding. But then we've also got some other groups that have heard about this or had known about us previously and expressed maybe an interest and wanted to come in. And so we're talking with them. So people are lined up outside your offices with wheelbarrows full of money.

Good for you. Yes. Good problem they have, actually. That was a dream I had last night. We'll see. They are lined up asking, so we're that far along. The wheelbarrow cash is a windy day here. I don't know that that would work. As far as like hyperconverged, it seemed like the market space for that is either

at the low end or the extreme high end. Is there a mid-tier or kind of market for hyperconverged? Do you see that, Ron? I think there's, you're right, there's two different pieces. One piece of hyperconverged, forget about the operational, the technology, the technology in the box. It's so simple to operate. So a lot of the early buys have gravitated toward regional offices and, you know, small pieces, parts of a corporation or something where they needed some technology that was highly resilient, very easy to install, very easy to operate. And so you're right.

A lot of that happened. Now, in data centers, people are starting to learn that hyperconverge works quite nicely for a whole lot of other applications other than, you know, say VDI or something that they started with. And they'll buy it for one of those things and then start loading more things on it. And the limits are going wider. So you're right. It's kind of a barbell if you looked at the distribution right now. But I think that middle ground will probably fill in over time. There's nothing in the technology that doesn't say it'll do that. I think that's just the initial market driven and the initial uptake is what's

causing that pattern that you see. And John, what was the market spaces for NextGen? You seem like you guys were going after VDI kinds of things. Yeah, I mean, our performance led the market versus other hybrids and even against many of the all-flash race that are out there. And so we kind of had a broad landscape of use cases, but I would say we're predominantly in databases. So transaction-oriented environments where they're running SQL, Oracle, SAP, or whatever it might be. So financial services, we do a lot of business there.

Even VDI is very popular. And one of the things that drives Salesforce too is the QoS capability so like we've talked about you know being able to handle that bootstorm or those spiky workloads and not impact other applications we fit very well in environments where there's multiple applications running on the same, you know, using the same storage system. And there's critical applications that are transaction-oriented or customer record-oriented that were quality services needed. You know, for example, Pivot3 and NextGen both sell into VDI, but, you know, we would

tend to win those environments where there's already servers that are looking to bring storage in and run multiple applications and really need the QoS stuff, whereas Pivot 3 would fit better in situations where it's a very large VI implementation and they want to set it up as a separate system and have it very scalable. And so those are kind of at a high level of how we're thinking of things. So next-gen cells from the low mid-range all the way into the low high-end right now, but adding the scale-out capability to our all-flash will definitely move the sub-market. And I think Howard was right a minute ago when he talked about the power of QoS in the hyper-converged market, because most of the uptake being hyper-converged is people using it for a particular application or a particular solution.

And then people are starting to load things on it. And we saw that trend. And that's why the QoS was so appealing to us. Because once you start to load disparate applications on with different performance criteria and different criticality measures, you've got to have something other than just let these things interact on the same platform and all. And so it's going to allow us to extend the hyper-converged platform far broader if we have that QoS layer on top. Yeah, it also seemed to me that a lot of people are buying all-flash systems because they want not the raw performance, but the lack of doubt about performance. And that QoS is as good a solution to that, to say, well, here are your 12 really important applications. They'll always run as fast as you need them to.

But you don't have to spend all the money to buy all Flash. That means you can buy three new servers as well. Even with all Flash systems, though, you end up with quality service issues, depending on how their data path is designed. But there's, you know, we've proven there's huge latency spikes on peer storage, for example. And so how do you smooth that out? How do you deliver performance and share, you know, as you're sharing that same data path, same IO controllers, same everything, there still needs to be quality of service stack in there,

allocating workloads and dealing with the idiosyncrasies of different types of flash. So we see that as a huge benefit. And different types of flash. Imagine everything from NVDIMM as tier zero down to the standard SSD bolted to a legacy disk wire and different types of performance and reliability there and being able to dynamically tier data across those to hit a service level. The whole definition of hybrid is up in the air this year. Enrico actually did a good post on the register about how hybrid next year is going to be some next flash memory like 3D cross point. Right, right.

As the performance tier and then flash as the capacity tier. And we might have four and five tiers where we're now or two has become reasonable. Right. And when you combine those you can deliver different levels of reliability and and then cost right yeah you know you can lower your cost by having a cheaper less reliable tier in there along with a high performance tier well you know and just if you manage it well you know if you make it last longer. Well, if your capacity tier is doing 0.5 drive rights a day, then you can use a substantially larger amount of the flash in that SSD as addressable as opposed to over-provisioned.

No, exactly right. That's exactly how we're thinking of it. You know, one of the questions I have from a hyper-converged solution perspective, most of the solutions out there started out with DAS, disk mostly, and they've slowly but surely been moved to using more flash. I would say next-gen is a bit of an overkill for hyper-converged. Do you guys see that as an issue? It depends upon the application you're running on the thing. thing, you know. So, you know, if you want to get into high performance applications like a database with big demands and such like that, sometimes, you know, you're not going to be able to assume

you're on one machine. I mean, that's how I think. And also, I think that's a limit of some of the hyperconverged people by assuming that your application and your data are going to be on one appliance or one server. I think that is really limiting you because, you know, this whole thing of, you know, server virtualization, the way that VMware led and such, that type of way said, you know, it's much better to have a big pool and use things in it and flow across the pool and all for a lot of reasons that we all know. Yeah, I call that the Arlo Guthrie theory you know one big one big pile is better than two little piles that's right that's right yeah you can sing that song and you know please don't howard and you can imagine you can

imagine a a cloud multi large scale multi-tenant cloud implementation. Yeah, and I could argue today that, you know, PCIe, Flash, and hyperconvergence start getting into a point where there isn't enough CPU to run enough VMs and that Flash really well. But, you know, we also have to think about, you know, what are things going to look like three years from now when we don't have 18 cores,

we have 62 cores? And with all that much more CPU available, it makes more sense. That's right. We know what the roadmap looks like. Well, you do. They don't tell us anything. John, we talked, oh gosh, a couple years ago about FastSCSI versus NVMe and stuff like that. And I've noticed a lot more activity on NVMe these days.

Does FastSCSI still exist? No, I think NVMe is the standard that everybody's going to go to. And Intel will tell you that. SanDisk will tell you that. Micron, a lot of the guys. There's still some work to do on NVMe, and it's, you know, making it hot pluggable, you know, and all that stuff, and work with the BIOSes. And, you know, everybody wants to drag along the disk drive baggage, the SCSI disk drive baggage forever, but I think people are finally starting to be smart about it and realize everything we built for the disk drive

industry doesn't necessarily apply to Flash. And so... I think 35 years was a good run for SCSI. Yeah, exactly. Some protocols never die, I should say that. Although some of the first NVMe devices, in an active-active

configuration, there's only two lanes of PCIe. And so someone could argue, well, is that enough bandwidth for a given device? I guess if it's Gen 3, you could argue it is. If it's not, then it's really not. But, and so we think there'll be different form factors. The fact that flash has to be on a two and a half inch form factor is kind of a destroyed thing as well. I mean, fusion kind of debunked that and did cards on PCI buses. So I foresee different form factors of flash.

Sure. It has to be serviceable, hot, all that good stuff, but it doesn't necessarily have to be in a two and a half inch form factor. I've always been surprised that there haven't been a substantial number of three and a half inch SSDs just to get the capacity. Yeah, we had that conversation at Fusion IO.

And I guess one of the problems is power and cooling. And when you put that much flash in that much in that dense, smaller package, the standard drive power won't power it. I was reading the other day NVMe has there are vendors out there doing half height half length nvme cards today yeah so the cards themselves will have the same nvme interface as the drives and so you look at this the fusion stuff which is now sandisk they will have an nvme compatible card and it's basically the same card with the different drivers look at it oh so you're not going to use 8 or 16 lanes because that would require another controller so nvme is limited to four lanes right well no the express bay connector is limited to four lanes right exactly so the

cards won't be so but if you're idt or somebody who's making an SSD controller that's NVMe, most of those are going to sell in the four-lane version because they're going to go into the two-and-a-half-inch package. And it just might not be worth the R&D to make an eight- or a 16-lane chip. Yeah, it's really a physical connector limitation. And, of course course it has to be 12-gig SAS compatible as well. That connector's doing a lot of things and there just weren't enough pins on it.

Right. And again, that begs the question of why does it have to be two and a half again? But anyway. Right. Right. If it was two and three-quarters, we'd have that extra three-quarter of an inch we could put four more pins on. Right. Right. If it was two and three quarters, we'd have that extra three quarter of an inch. We could put four more pins on. Exactly.

Yeah, but it's all about backwards compatibility. That's always been our curse. Yeah, we're captive of that. We are. Okay. So, Ron, is there a limit to the number of servers in your configuration or that sort of thing? No. I mean, once you get into software, you scale.

I mean, that's the simplest way to make something scalable is make it software rather than hardware. And so you can get up, you know, to really large sizes. Our largest currently installed single site has 275 appliances and running seven petabytes of storage. And so that's getting on up there, you know, and it runs quite nice. It's been running for several years, very dependably. And so you can scale a lot with this stuff once you get over this.

That's not all one cluster though, right? That's got to be multiple clusters, right? We have for the erasure coding, we have logical groups that are smaller numbers like 16. So you're not striping across the whole 275. It's operational on a single pane of glass

and operates logically and all. It's just for the erasure coding that you have the subgroup because you're right. It striping it across 275 would be a little much. And isn't there like a vSphere cluster limitation too, somewhere like 64 nodes or something like that? It rolls, that rolls okay.

You know, we can operate and such. So it's just, it's pretty amazing. Yeah. Both of you guys talked about cloud as a potential solution adjunct to what you're doing. Can you explain how HyperConverge works with a cloud solution? Well, what we're trying to say with the QoS is the way we were talking about it. With a QoS over, we'd like to be able to allow a data center to use the right processing and the right storage for the right application to get the right performance.

And if that is your objective, then you have to economically consider cloud or cloud storage for the economic reasons for some applications and some use cases and such. And so that's why theoretically, we immediately, when we started talking about this, said we need to extend this to the cloud so that it can use that. If it makes sense economically and performance-wide for a particular application, with our QoS layer, we can enable and help that. I thought it was just because the IT Executive Statement Act of 2012 said that you couldn't talk for more than 20 minutes without saying cloud. Also true. And the other thing we haven't said is Internet of Things. you know, these gigantic rooms, E-Trades, data center, other, you know, gigantic, you know,

with tens of thousands of these commodity nodes running, you know, software stack, highly redundant caching, load balancers up front and all that stuff. And I can imagine a massive consolidation of that type of infrastructure, you know, from 10,000 nodes down to a couple hundred nodes that delivers more performance, quality of service, better reliability, better scalability, better performance, all those things with our technology. And so that's kind of what I'm thinking long term of, is there an architecture that can replace that cloud architecture at a similar cost point that's much more efficient with a much smaller footprint and has multi-tenant quality of service.

You can do chargebacks with and all that good stuff. God, you're going after the whole world here, John. Well, if you're going to think, you might as well think big, right? No sense trying to take over Lithuania. Actually, one thing I learned earlier, I talked with LJ7 of the Seven Rosen companies and we were chatting about, because he was the first investor

in Compaq Computer years ago. And somebody said, well, why would you want to take on IBM, about a little startup and all? And LJ said, you know, building a startup is really, really hard to do. So if you don't take on somebody big and you do it successfully, you don't get the right reward. So he said, you might

as well take on somebody big, dream big and go for something big, because then if you do it successfully, you're going to get the big reward. Yeah. Although, frankly, hyper-converged vendors arguing that, you know, what they do is like what Facebook does. One of my pet peeves. Yes, they have 4 million x86 servers and some of them are for storage, but compute and storage don't run on the same machine. Yeah. Well, in some cases they do.

In some cases they don't. In Google, they tend to more so. I guess we'll leave that to offline here. Yeah. Everybody's laid down their cards. We know what the solution is here. We just need to get them in the same room and have them hack it out,

so to speak. Bring out the globs and do it. What sort of things are on you guys' roadmaps that are going to have to be merged together? Oh, come on, Ray. They've been at this three days they don't have a roadmap yet but they had six hours of kelly and bill together imagine that would have solved everything well kelly and bill did start trying to categorize john and i were throwing ideas out of about we'll have this soon and we'll have this

soon and we'll have this soon and they were have this soon and we'll have this soon. And they were starting to put buckets of short-term, medium-term, and long-term and trying to put all these ideas in them at one time. John and I, so far, you know, trying to resist that, but they're already starting to categorize them that way. So there's some work to be done here. A number of things will pop out pretty early, but some of the things are pretty, we've each got pretty sophisticated code and software and platforms and it's tedious and it's some, you know, innovative work to do it,

but I'm very confident. Hopefully it will go faster than merging Spinnaker into ONTAP. Ooh, yes. You know, I'd just like to say that, you know, we're going to continue to support the existing NextGen product line. You know, we're broadening the product landscape for all of our channel partners and customers. And then there's some really exciting stuff coming on the roadmap and that's targeted towards the channel partners and customers out there that love your stuff. One of the things that just kind of fell out that we didn't really expect was using pivot three, you know, as you know, the next gen system right now you can add jbods to the back end up to three jbods to increase your capacity well we can easily without any engineering work

add pivot three to our back end and have that as our scalable storage jbods to our head unit and scale to petabytes versus just you know three or four or four JBODs. And so now we have... Have the head unit mount the Pivot3 via iSCSI? Yep, yep. We're just an iSCSI target behind a head-end unit. That comes out pretty quickly. And that offloads the data protection from the head-end as well? Yes, it does. You get the erasure coding data protection, absolutely.

That would make a really nice Tier 3. Yeah, it took it to the... It sure would. You know, because we had the Pivot3 system in the lab full of 7200 RPM drives. And, you know, that system works real well for big IOs and sequential kind of stuff. And our QoS, of course, can dynamically manage those tiers based on their capabilities. Right. When you say dynamically, you're saying what those capabilities are as revealed by their

normal I.O., John? Yeah. Okay. Okay. I got you. I like that. I hadn't thought of that. It did take Bill and Kelly about 45 minutes when they first met to invent that.

Right. Well, I mean, it's an interesting play to go back to customers who started off with the storage system to say, well, you have all these drive bays in your servers you're not using. Make that the archive tier. Okay, now you got me thinking. You know how dangerous that is. That's why we need to get you in a room with kelly and bill and just watch the sparks fly

you guys know where to find me i wanted to ask about 3d tlc and and 3dx but i think we'll save that for another time we touched on post flash yeah and 3d tlc is kind of just the future, isn't it? Yeah, we're looking at all that technology and we've had Intel, Micron and Samsung and Toshiba all present their roadmaps and give us kind of a technical overview of what's here now and what's coming. So we're anticipating that technology and looking at how we can take advantage of it. Very good. All right. Well, g can take advantage of it. Very good. All right. Well, gents, best of luck with your marriage. Thank you. Well, this has been great.

It's been a pleasure to have Ron and John with us on our podcast. Next month, we will talk to another startup storage technology person. Any questions you want to ask, please let us know. That's it for now. Bye, Howard. Bye, Ray. Until next time, thanks again, Ron and John any questions you want to ask, please let us know. That's it for now. Bye, Howard. Bye, Ray. Until next time, thanks again, Ron and John. Thank you both.

Appreciate it as always. Lots of fun.