 Mike Davis, he is the team lead at dedupe marketing Right, I do many things. Yes. I saw that I was trying to work through all the things that you do here Why don't you explain it? Like I've done a lot of things actually, so I ran marketing for Ocarina Networks. Okay. You came You actually came from Ocarina. Did we meet? We probably did one. Who are you by the way? I'm John McCarty. Oh, John McCarty. And I would know you from. We probably met in past lives. I was at before Ocarina. I was at it was in the content delivery business and I was at Sun Microsystems Okay, that's storage something called honeycomb, which you know. I knew honeycomb, right? Yeah, so I think we crossed paths. A lot of friends over there, it's snorkel. It's snorkel. Some of them are still there. I know, I know. But they're all calling me now. Yeah, some are. Some some are. That's not clear. We're storage is going over in that shop. So I was part of the acquisition. Okay. And the acquisition process, which is, you know, I can talk about that. Yeah. I know all about Ocarina and its legacy product and now I'm involved with all the different projects, the platforms that are now integrating that technology into the Dell portfolio. So this is one of those situations where you've got a market that becomes a feature, right? Yes, that's right. So it was a standalone platform that worked with other people's stores that shall now be nameless. Right. And so we had, you know, and the real IP in there was the algorithms and the workflows, the policy-based workflows that shrunk the data. Right. And we did a better job of taking data than anyone else. And we had started working back at the beginning of, let's see what year is it now is, 2010. So in 2010, the beginning of 2010, we had started working with a few OEMs to actually embed our intellectual property in their platforms. Okay. That was one of those guys. Okay. And did that stop then after the acquisition? Yeah. Or you still have an OEM opportunity? Yeah. So Dell kind of pulled that trigger and took us off the table, as they say. For now. And there are, I mean, clearly we wouldn't go and engage with HP or some direct competitor, but there are other OEMs that are, say, vertically market-oriented. It's like if I was in the medical space or you were looking for oil or, you know, maybe a VMware or someone like that, where we have a real strong relationship with Dell, we would look for opportunities to do tighter application integration with those guys. So, you know, there's still the possibility of open discussions there. You know, but the thing we're no longer interested in doing is shipping a co-processor that goes and helps sell more network appliance. Okay. And that makes all the sense. And do you have other jobs in addition to that? We mentioned the long list. Well, what's interesting is... Yeah, you said the long list, so that was short. We, at Ocarina, were very much focused on the NAS marketplace. Right. It was all file-based workflows. We compressed and deduced files based on policies and attributes of those files. And so, most of us at Ocarina were experts on the NAS business and the file system business. People came from my salon and people came from Sun and NetApp and elsewhere. And so, I immediately had an affinity with PG at Dell. PG being the enterprise products group and the storage team there under Darren Thomas and helping them because Dell has been... Dell was very block-based. Very block-based, very block-oriented. All the IP that Dell owned was essentially block-based. And they purchased the Exynet NAS, which is not... We call it the DSFS file system. It's really a full NAS stack from disk all the way up through the protocol layer. And so, they'd acquired Exynet not too long before Ocarina. Very... Maybe four months, three or four months prior. And so, they were down the path of getting those products to market. And so, I jumped right in on those to help them with use case-based marketing and building more collateral. I actually wrote some of the white paper stuff. Okay, so you're a solutions guy. I'm kind of like a consultant into the product marketing effort. And as of, say, a month or two ago, I officially jumped over to be on that team. So, right now at Dell, I report up through Darren Thomas. I manage the team that does all the NAS and file system work. And that spans these platforms, right? So, we have Equalogic and Powerball denounced today. And then other things, you know, implied... Yeah, Dell has been discussed. It's bear a bit on here. It's like, yeah, I mean, it sort of makes sense to, at some point, integrate scalable file system and deduplication into the... into component. All of the above. And so, owning the file system piece as a key IP component that goes on different platforms, but also the Ocarina piece. And, you know, we're starting to talk about how that's going to roll out over time. I'm giving a presentation tomorrow morning. That's going to go into a little more detail. Okay. But each of these platforms is, I think Darren was discussing this morning, have different operating systems. The embedded operating systems are different. So, there's a fair bit of work to be done. It's non-trivial. It's non-trivial. Right. We have the Ocarina development team, which was based in San Jose. We've done it to a Dell office in Sunnyvale. So, the Silicon Valley Dev Center, that's where DC, is mainly that team of Ocarina guys now times two, who own that team to support all these different programs. Right. And so, they're working really hard. I mean, there is, I counted, maybe I shouldn't name a number, but there are something like 10 programs in different states that are now in the pipeline moving forward that will all use some form of the technology. So, we're really leveraging it across the whole portfolio. It's going to be, you know, it takes time to do it right, especially in a big company like Dell. But, you know, what we have on our side is, we don't have a lot of legacy baggage to deal with. I mean, everyone's marching in the same direction and everyone's enthusiastic about, you know, about adopting the best-of-breed components to do the right thing. Ocarina had a lot of technology. It didn't have a huge installed customer based on the time of the acquisition, right? No, not really. It was really more of a technology play. Right. Not in a creative revenue opportunity. Right. Although some of those customers still have the Ocarina gear. We still support them and they're now, you know, we get to go sell Dell's storage to them, which is great. We have a good relationship there. And as we move more from, say, SMB-oriented NAS products, file system products into more scale-out, large enterprise-oriented file system products, you know, those were kind of the nature of the Ocarina legacy customers. And so we, you know, we keep those relationships alive and they've been giving us valuable feedback. You know, so they, you know, you know, there's, you know, a lot of interesting dynamics right now with the acquisitions and, you know, we're still, there's still a degree of, you know, we spend a certain amount of our time kind of merging processes and meeting new people and, you know, you know, we can kind of look at, oh, we didn't do this very well. Well, we have to re-engineer it anyway because we have a new organization here to deal with or a new go-to-market model because you have to think about new things. You know, it's, it feels like a start-up environment. I mean, no one coming from Ocarina and our 40-person company, start-up company, feels like we've run into this monolithic wall of goo that's going to just slow everyone down. That's good. But what's been the one challenge that we've talked a little bit around challenges, but can you delve into... The one challenge? What, from an Ocarina perspective, would you say? Yeah, into the transit. No, yeah, from Ocarina with the transition and to Dell. Yeah, that's a good question. The biggest, I mean, so there are challenges from an integration perspective. There are challenges from a business perspective. I mean, I can touch on different things. From an integration perspective, you know, I thought Cisco was kind of the model in terms of, you know, really killing it. You know, they've done it a hundred times and it's just a new company. But, you know, I think the experience personally and speaking for the team was tremendous. I mean, just the respect for the local culture. I mean, we get free snacks in our kitchen in Sunnyvale, but no one in Round Rock would ever get such a thing. No, no, no, no. So they did respect local culture. Absolutely. And so just don't tell anyone. You certainly haven't done that. We're live on the queue. All right. Sorry, Round Rock. Everyone. Sorry to Sunnyvale, you said? Sorry to Sunnyvale. Loose the olives. So, you know, things like that have, they're kind of symbolic in a way. And the executive leads that have kind of, you know, they put a top-level guy in charge of the acquisition. I mean, these guys go to bat for some of those issues. And it's worked. We haven't lost anyone. We've doubled the size of the team. We did double the size of the team. We talked about the investment this morning. Yeah, and that's gone. So there's five, I mean, I can speak just for, just for stores. There's five dev centers, right? We've got Minnesota. We've got Nashua. We've got Israel. We've got Silicon Valley. And then we've got the most in Round Rock area. And that's a non-trivial management. Are you spending more time on planes? I spend a lot of time on planes because the team I lead is based in Round Rock and I live in San Jose area. And are you part of the quarterly meetups in terms of integration that Darren was talking about earlier? You were talking about the wonderful meetings in Minnesota. Yeah. Did they want to do Minnesota? No, it says not Minnesota in winter. Oh, that's right, right. Well, you know, you've got to know which meetings you don't want to go to. Right. I'm busy with that. Yeah. Yeah, you had to pick your battles. So there's surprisingly little challenges on the integration side. From a, you know, one of the challenges was really uprooting what we did as a company at Ocarina, which was sell this appliance de-dupe and compression box into third-party storage environments. And we kind of uprooted that and we had a sales team that kind of had to figure out how to reposition them into Dow. So there was a little bit of, you know, based on the strategy. I mean, it's just a fact of life or some things that you have to get done. And all those Muslim people are still happily employed here and have found now storage-oriented positions. So once we got through that repositioning, then it's really just down to execution. And now the big challenges are how do you get across five development teams who are going to use our, the Ocarina technology components in their platforms. You know, how do we do that the best way, the shortest time to market without ripping and replacing code that they've already written. And optimizing for those platforms. Because it's not, you can't just say copy, paste, paste, paste across a block array, a NAS system, an archival system, or even a server product. We have to think about the use case and the workflow and the performance expectations and what kind of data is going to be stored there, what's the throughput it has to have. So there's a lot of data in the cloud. Because some of the cloud suppliers, cloud services companies, one of the things that they're doing when they build out their infrastructure is they're actually just using internal storage in servers. So they buy a Dell server. They buy an X86 with a bunch of disks, right? What's your play there? Are you part of that? Are you working with the server team as well? Well there's an interesting end state vision where a customer, whether it's an enterprise or a cloud provider has just a whole bunch of commodity compute and storage resources in their X86 boxes with just in them. And you can kind of dynamically provision them hey I'm going to assign a VM, a server personality to you 1,000 boxes and I'm going to assign a storage personality to the other 4,000 and it's all going to work. And that's a viable view to have. I mean we have line of sight into how we can go implement that. Now the question is how do we step along that path to get there? I don't want to fixate on the cloud providers because that's a tiny market in aggregate. Tiny market? Yes, absolutely. Look at the number of people out there and the terabytes that are actually being utilized as compared to the global market for enterprise. I can see that at this point. But it's not going to be tiny. So there's clearly You have a direction there. So there's lots of ways to answer that question. So I personally look at that market as hey these guys are early adopters in terms of driving multi-tenant secure wide area accessible storage. Now if that's the goal in mind then that more clearly frames it for me. I need wide area capabilities. I need security. I need multi-tenancy. Regardless of whether it's privately deployed or publicly deployed. We were talking this morning to one of the cloud providers, Island, who said if I've got five data centers or seven data centers, whatever they've got today I think seven now. Two in London and they got to move data. They got to replicate data around. So one of the things they want to do is reduce the amount of data that they've got to move around and still be able to restore all the instances. So one of the jobs I have is is defining what we mean by our fluid strategy. And when you saw Darren and Michael talk a little bit about our fluid strategy it's you're not seeing all the underlying details that we kind of have a roadmap of capabilities. Actually here's the physical capability we're going to deliver over the coming three or four year period. That's one of my jobs is to try to pull that together and then drive that through our architecture teams to make sure everyone is bought into it and aligned and building the right components and sharing the right components to make that happen. It includes obvious things like tiering within platforms and across platforms. Kind of the classic ILM stuff we've been talking about for a long time but I think there's a way to do it in a simpler way, a way to do it that has less egregious licensing terms and forklift upgrade issues and it just hasn't been done right but you ask every customer in the room they'll say, yeah I'd love to implement tiering. There's some obstacles to doing that that have happened. Another one is use of the cloud. I mean that's absolutely attendant of what we're talking about fluid data. Now you'll start to hear us talking about the Dell cloud rather than say, hey we'll support an S3 interface and you go do what you want to do. I mean we think that there are real opportunities for innovation on the cloud services side. And so we're going to be looking closely at those kinds of things. I mean there are services today for example that an enterprise could license at great cost, which means that for an average SMB customer, mid-sized customer maybe not such a good fit. We're paying six figures for a license for a piece of software that does index and search and so forth. But if you can amortize the cost and complexity of those kinds of things across petabytes of cloud back in and deliver them in a pay-as-you-go model which is precisely the value proposition of a utility storage service like that. And that's really interesting. We always intend to support openness and the support for third-party software products and the workflow and third-party storage products will attach to our servers. I mean we're not going to do away with any of that. But there's distinctly going to be a better together story. So if you're using an ecological platform... Sorry that was an HP term there. Oh yeah that's right. So you need to strike that for me. Dell on Dell value proposition. There you go. Dell on Dell. Dell on Dell. And so you'll start to see us talk about things like, hey every storage platform should be cloud-enabled. And to simplify that, to make it totally painless and reliable, to eliminate complexity, to have coordinated transfer of optimized data. So this system communicates over here and says, hey I already have those chunks, don't send them to me, those kinds of protocols. We can do that much easier more effectively when it's Dell talking to Dell. These are some of the things we'll leverage the Dell cloud for. And it will become extremely easy for a customer that doesn't have... We find even customers that do have multiple facilities that could set up Dell replication targets in the other facilities, it's still a hassle. You still need to negotiate bandwidth from your facilities people. You still have to deal with standing up the other system and administering it. There will be a point where it will be easier and more cost effective to just click on the menu. Oh, use Dell replication target service, for example something that's precisely aligned with that platform in mind, that's optimized transfer, that model that gets built through the same mechanism as everything you've seen from Dell before. That's the kind of a cloud service that will become pervasive and widespread. There are early adopters today that are starting to use the cloud in the enterprise but until you sort of reach that ease of use on integration level, that's when you see the knee of the curve. You made a claim earlier when you first arrived that you have the best de-duplication, most effective de-duplication compression technology on the planet. You've reduced and compressed better than anyone else. Can you back that up? Tell me why. You're a hard hitting reporter then. We will stand by that claim. Are you offering a guarantee? Well, you're not shipping it right now. Sure, there's warranty on our products. The claim is in any Apple's Apple scenario we will shrink better. There are scenarios for example where the workflow for that customer may not be suited to a given platform that does optimization. That's an apples to oranges scenario. We won't talk about those. But in an apples to apples scenario we can compare two things in the same workflow. Yeah, we will shrink better. Now, there's a couple of reasons for that. One is people. We're the only company that I know that actually has algorithm PhDs. These are the kinds of guys I'm sorry. I think Diligent made that claim. He hired more people. I think Duran Kempel made that same claim. How long ago? Diligent was acquired by IBM. Right? They're doing the same thing. Absolutely sure. IBM has plenty of PhDs to be involved in algorithms. So anyway, we will certainly make a claim that some of the guys we've hired to do algorithm work for us are top guys in that field are renowned as leaders half in academic, half in commercial environments in terms of inventing new algorithms like compression algorithms for specialized data sets. There hasn't been a lot of development in compression algorithms per se in the last 15, 20 years. We still see deflate, LZLZ77 derivatives dominating a lot of the compression landscape because they tend to be generic and good enough for some applications. They have good performance so you don't have to think about optimizations there. It goes into some use cases data's pre-compressed or something like that and you get very little in the way it results. So can you operate on pre-compressed data? Absolutely. One of the things that the other reason we shrink better besides we have good people we have good algorithms that we invented and we're not patenting those. Those are trade secrets. We also have a lot of work on the logic that picks algorithms. So at runtime you have a ton of algorithms and you choose based upon the data type. So we can actually shrink data that we've never seen before. The system will actually test different algorithms at runtime and pick the one that's winning and it will remember those choices. It's actually I hesitate to use the term neural network but a context-weighted learning algorithm which is technically the definition of neural network is something that we've used in the legacy of Karina products. I just want to make sure I got that. Context-weighted learning algorithm and that means whatever is winning over time we weight that out with that thing more and so when we see that data again we go directly to it. Give us an example of a data type that was thrown at you that you hadn't seen before. So to be effective at compressing DICOM images you need to understand the headers and the file structure. So we write special algorithms that actually don't do shrinking of data but actually do delaring. They'll take apart the file, get to the image, apply a different compression algorithm to the image from the metadata and so forth. So there's this kind of a delaring concept. We've done specialized algorithms for the oil and gas industry for files like SEG-Y which is commonly found for files which is a file that comes off a TELIS-NE scanner in the film business These aren't your everyday home applications This is not in the ZIP software, right? Do you get compared to PKWare or ZIP? Does that come up? There are specific tools for specific industries like for JPEG compression that do reasonably well on their own and if all you want to do is shrink a bunch of JPEGs and you're willing to write the scripts around how that tool runs and it doesn't have to be particularly reliable and you can embed that code because it's freeware and you can put it in your app Great, that's perfect but if you want something that has enough cores to run at a good throughput that's HA now you need to start thinking about how that's implemented and so forth so we've won a lot of deals that way and we don't want to do it but how to scale it and how to have it be reliable so these are some of the things now that all being said we have to take what we learned at Ocarina and plunk it into platforms that are limited in CPU resources that are limited in RAM that have different kinds of interfaces We're not dealing in HPC environments here where they're trying to take the scalable file system right now and you've got to think about in the legacy Ocarina product which is very much oriented around the petabyte scale customers big archival so I'm going to store three petabytes worth of film data that's not the model for say a backup target a backup target which is focused on reducing backup windows there's a mix of performance requirement how fast can you throughput my backup set as well as the ability to shrink and so we would balance those things instead of being a policy based post-process would be an inline high speed Ddupe oriented and you are inline we're all of the above we're all of the above it depends on the platform are you the only I'm inline and I'm post-process we've got a product that extensive research I will claim we're the only let me see that video there's a great video the kids were showing it to me the other day it was a kid trying to start something and he gave a great speech he just couldn't get it started he just had too many thoughts going on he had too many thoughts going on one of the things that we are the only ones that use both Ddupe block level Ddupe