 Okay. Everyone, welcome. Sorry to get into a few minutes late. I have a, we have a great panel here for you today to talk about driving ROI in OpenStack and how they did it, especially from a startup standpoint. And we're also fortunate to have Martin here who can also give you, I think, some really perspective in how he sees it from HP. So let me quickly introduce the panel and we'll get, we'll get right into it. So starting off is Joe Arnold, who's CEO and founder at Swift Stack. Prior to that, spent some time at cloud scaling and at Engine Yard. Next to him is Jim Morris Rowe, who's CEO at Piston Cloud. Prior to that, he was Senior VP GM at Zimbra, before getting acquired by VMware. And then last but not least is Martin Mikos, who was CEO founder at Eucalyptus, was recently sold to HP and is now Senior VP and GM of Cloud and all things Cloud at HP. Prior to that, he had been founder at MySQL that grew to be one of the biggest open source companies that you're all probably familiar with. So with that, just, you know, I think it'd probably be great to just jump right in. What I thought I'd do first is have each of them give, instead of talk about their companies, give maybe a couple of examples of some customers and how they drove ROI and how they showed ROI in those customers. So Jim, maybe start with you. You've got, you know, some customers like Candy Crush and, you know, King, I don't know what you can talk about. Zulily, what can you tell the audience? How you want to, how'd you show ROI? Yeah, sure. So I think it's becoming, you know, somewhat well known, certainly at conferences like this, that private cloud infrastructure done well can drive cost savings even versus Amazon. So, and there has to be a certain scale until you get there. It's not at, you know, 5,000 a month, but at 25,000, 50,000 a month, customers can start to see value if they build efficient in-house systems. And I think we certainly take some debate on that, but we've seen it at Candy Crush and at Zulily and at ChartBoost and these are, you know, six-figure Amazon bills that have come down by 60% in some instances. And how did you tell, how did you convince them? They must have been skeptical at first. Yeah, so how did- What I think Piston ultimately does well is we enable the private cloud deployment with minimal upfront costs. So we have a piece of software that installs super quickly on commodity gear in a very scale out web-scale centric model and it allows the customer to see that it's not no ops, but it's fairly limited ops because if you have a huge professional services engagement, a big hardware buy, four admins that you have to hire, you're not going to realize those savings. And I think the way Piston goes in quickly and easily in a POC fashion gives customers that comfort level. I think the other point is that the, you know, one of the best benefits of cloud in the public sense is its linear scalability. It scales indefinitely, but the cost scales at that same pace. And that's not, that's very difficult to do in-house. And I think the one, you know, really unique advantage Piston has over some of the other OpenStack solutions is everything runs, network compute, storage, and the management plane all runs on every node. And, you know, there's certainly challenges where there is huge object stores are a challenge for us and so we partner in those scenarios. But in general, for most of the cloud's requirements, the costs will scale linear as well. So as you need, you start with 10 nodes, as you need more capacity, you add 11, 12, 13, 14, and you have this very kind of consistent cost curve. And I think those two kind of attributes, very hands-free system, not no ops, but close to it. And this linear cost curve really helps us kind of drive that mindset. So, Joe, does it work the same way in storage as it does on the compute side? I mean, there's some slight differences there, right? And I echo this statement. It's really easy to figure out what the CAPEX costs are gonna be when you're doing your own private cloud. What's a lot harder to figure out is the operational expense. How many administrators are you going to be adding to your team or augmenting to your team to run that? And I think the goal on both of our fronts or everyone here on the panel is to really reduce that operational costs of running these larger scale clouds. But particularly with storage from our point of view, there's actually a bigger change going on, to a commodity-based, open-source storage environment. That's new for most people. And so, especially in that context, making it easy to consume storage in that way is a big challenge. But once you realize that challenge and realize that it doesn't necessarily need to be a project that you need to take on, that's where it gets a lot of benefit. So, Martin, you've got a very interesting perspective here, having sold to some degree against OpenStack and aware as a competitive framework with Eucalyptus and having to convince customers to trust you on that. So, maybe you can give us a little bit of insight from that point of view, but then you've also now got this insight being inside of HP. So, can you compare, contrast? Okay, how do you kind of think about that? First, from a Eucalyptus and OpenStack, similar now with an HP? Yeah, I'll be happy to. First, if I may correct a little bit. I'm not the founder of Eucalyptus. I was not the founder of MySQL. I'm a hired gun. I'm smart enough not to follow my own ideas. But anyhow, I think you could ask, why do we have to have a panel on ROI? Because it should be obvious. So, if we have a panel and the room is full, it means that we have some concern or question. And I think it has to do with the fact that the first benefit of cloud is agility. It's only later that we get into higher utilization and automation, which bring the direct financial benefits. But early on, it's agility. And they say, what the heck is agility? But it is for companies the ability to move faster, to change plans faster, to do things in smaller increments and be much more responsive. And it means that the organization becomes more productive. Their R&D is faster, or they serve customers better. And then when we go back and say, so what was the ROI of the private cloud or a cloud in general? It's difficult to know because it's not exact dollars. So that's one problem. The other problem is that private cloud projects are messy. I know, I've seen two. And it means that as you are deploying your private cloud, you spend a lot of time just getting it going. And then at the end of the day, you must say, so what's the ROI if it takes this much work to get going? And then you have to rationalize it for yourself and say, oh, I'm learning, that's a thing. I'm developing my skills. And soon it will be so automated that we get the real benefits. And I think all of those are true. But that's why we are sitting here today not knowing what the ROI is because the real benefit, first benefit and the reason to move the cloud is agility. The straightforward financial benefit comes a little bit later. It's even agility when you go to AWS if we use that as a contrasting point here. You don't go to AWS to save money and you don't. You go to AWS to get more agility. You move much faster. And I know a customer where they make large telecom equipment and they use to test them four times a month. Then they built a private cloud and they started doing tests 4,000 times a month. And today they do about 4,000 tests per day. And most of them take less than 10 minutes to run. So they split the R&D function into such small pieces they can test everything all the time. How do you do ROI on that? The result is faster R&D, higher quality sooner and not necessarily a saving in the actual IT cost. And that's why I think we must ask ourselves, how do you do ROI calculations when you have to measure something that's very difficult to measure? So either you could let this right now, the HP is maybe newer or harder to talk about since it's so new. How do you sell agility? I mean, how do you go in and pitch agility to a customer, especially against me if they're looking at AWS which is the ultimate agility in a way? Yes, that's a very good question. How do you sell agility? Well, first you say I sell only to those who know that it's agility they're seeking. If they're not asking for agility, don't sell to them. Because there's more customers than we need today. So we can select those who really understand it. And then why would you use something other than AWS? The thing is that there are customers who need control. And that is the thing you sell to them. You say get the agility that you get on the public cloud but you're under your own control. And at that point, they have to choose something that they can control. Well, I see this agility bit. There's a little bit of twist, at least from our point of view at SwissStack with storage. And we're seeing customers like in life sciences doing cancer research, a lot of video and video distribution. And what they're being asked to do as operators inside their organizations is to 10x the amount of data that they're needing to store but they're not getting any more headcount. They're not getting any more budget and they're just having to figure it out. And so to them, it's this deluge of projects that they're trying to take on. And this is a way for them, an alternative path for them to go down in order to get done and enable those projects that they weren't able to do before. Yeah, so I think it's a great difference in kind of Martin's spot at HP and the resources that he can bring to bear for a customer that has to understand their business problems and translate that into agility and translate that into private cloud or public cloud. As a small company, I don't have the benefit of that sales cycle. And so we're agility spot on and I think even more of an indicator for us is that the applications that our customers have created or are creating are cloud applications natively. So we're not helping people port their legacy workloads into a piston infrastructure. We're helping companies that have already built their stateless tiering and their redundant data structures in the application and we're helping them achieve this linear cost savings as they scale out their application. And I agree, agility is the key and it's hard to quantify at that level. It just for my team and certainly when we talk to customers, if that agility light bulb hasn't come on, we run away and even more clear than that, if their application set isn't very clearly already kind of cloud architected, we also run away. And so I think that slightly puts us in kind of different engagements, but I think agility is key. I think we kind of, some of the same perspectives coming from doing products around it and who's heard of like the bus number, which is we used to use this in development parlance. Like how many people in your team that they got hit by a bus would the project be totally like screwed? And what products do around an open source ecosystem is that it gives you a vehicle to take on these projects and it de-risks taking on these private deployments for yourself. That means you have more organizational support, you have the project is on a different set of rails so you can get support throughout the project. And I think that helps also on return on investment because you need less team members in order to get the same amount of functionality and benefit out of the cloud that you're building. I'm gonna save time at the end for questions so everybody please think of some questions, we'll make sure we'll make some time for it. Just switching gears a little bit. I imagine a lot of people in the audience are thinking about how they build a case internally. You spend collectively all three of you, you spend a lot of time convincing customers that OpenStack makes sense in respect of solutions. What advice would you have for engineers, for people inside of companies that are champions, they wanna do it, they're trying to convince the organization to do it, there's some resistance, there's someone that really likes VMware or whatever. What advice do you have for them? The first advice is to go for open source solutions because then you have less lock-in. Not that you avoid lock-in completely, just have less of it and you have more opportunities to choose among vendors and do things. So for most customers that's the main choice and then when they go into the OpenStack world they can grab any of us. I'm happy to recommend Piston and all the other companies in this space. That is the beauty of the ecosystem that you can try out with any of them and switching from one to the other. Sure, it's a little bit of a headache, but not a huge headache and you will have headaches anyhow, some headaches anyhow. So just picking up on that, a lot of people, they resist open source because they view it as just being difficult to deal with and a bag of bolts and internally it's just so much easier. I can just buy a solution and we talked about de-risking things. So what advice, I mean, you've been in open source for a long time. What, if you've got an organization that's just resistant to some of the things, how do you help them get over that? Those who resist open source shouldn't use it. Let them overpay for proprietary crappy software. That's what you must let them do and they'll lose in the market because if you, I mean it seriously. So it sounds like your advice to those engineers is leave that company. No, no, I mean it not as a joke because you must pick your battles carefully and you must know where you can win and if you try to go out evangelizing open source then you need to sell your own product. The beauty of the open source movement is that it evangelizes itself. You don't have to sell that notion but if there's still people who haven't understood it or don't agree with it, leave them. Because selling to them, that will destroy ROI both for them and for you. So just leave them because there's so many and I've seen this from my skill days when open source really wasn't accepted but now it is mandated by large corporations. And the fact that you in the open stack ecosystem you have all the major IT vendors, IBM, Cisco, HP, Red Hat, and so on and very few are not there. So if you still face resistance to open source then just leave those guys. You will never be able to sell to them. So leave your job or leave your job or change your job for a change your job type of thing. Yeah, I mean kind of more pragmatically we use a lot of spreadsheets and help people through it and we get asked that question by people who are in the seat of trying to convince their management and here we go. Here's what this looks like. Here's how to amortize the cost. Here's how many operations. So I mean there's an exercise there that you can go through really to kind of pragmatically build it out. And let me tell you, it makes a ton of sense across the board whether you're replaced so in the storage base so with Swiss stack. It's super easy compared to storage appliances that you would buy and everybody has lower than Amazon costs when they're deploying with us privately as well. Yeah, so on the ROI assistance, very tactically we have a page on pistoncloud.com. It's a TCO calculator that allows you to put in your cloud requirements, storage, memory, compute capacity and then compare that to your Amazon bill and it's a fully loaded operational cost model three year amortization on the hardware and you'll be able to see your direct monthly savings with a set of assumptions. Have some questions on how we built the model. It's fully loaded, pick up the phone and we can talk you through that. So that's one help there. Back to Martin's point on the broader open source technology, I think we in this room certainly in the community have a particular challenge with OpenStack in that it's a framework that really by itself does very little but with all of our assistance it does so many different types of things from a use case standpoint. There's obviously tremendous amount of NFV discussions here. There's this private cloud conversation that we're having for both enterprise workloads, web centric workloads. There's the VMware oriented environments and so we're particularly challenged in that we have to have some kind of focus inside of this framework and then we also I think are going to see this adoption pace continue for a little bit kind of because of the complexity lag behind some of these other things. I bet when Martin was four years in at MySQL there were millions of downloads at that point and that because it was solving a very focused problem. We put up Zimbra and the first month we had 18,000 people trying a mail server and we don't have that pace yet because this framework is so vast and there's so many different use cases and I think we'll see it. I'm already seeing it this year but it's I think a particular challenge this piece of open source technology that we should be conscious of. Well, I think it's a little bit different too because of who is trying to adopt it like with a developer tool or something like a database it's downloadable, it's consumable, you can run it on your laptop but when you're building out a cloud environment that's a set of operators and what I think is so great about OpenStack as a movement, as momentum is that it allows this generation of operators to learn about it and get trained up and I think it's a very elevating thing for to bring on because it's a very portable skill set from job to job. It's not like learning a proprietary technology and getting certified in that technology. It's, I think that's a really neat thing about the OpenStack as an organization because so many companies can get involved. So Martin, I'm interested in your perspective. Do you think that the case, the ROI, the justification for OpenStack is it increasing? Are we kind of peaking? Does it just get, does it get better over time like a good wine? What's kind of your experience in looking back on your history with open source? It will definitely improve. It is improving and it will keep improving and one reason for that is automation. That we are building a very complex product and we are not done yet with making it easy to use. It's complex for those who build it and complex for those who use it. When it reaches maturity, it will be complex for those who build it and easy for those who use it and that's when you will see the scalability where you can scale the cloud without scaling your organization and ROI will only improve and I think also we will learn to reach higher densities of whatever it is, hardware density, software density. Using containers, we will have more efficiency if for certain workloads, not all of them and we will learn to sort of pack more stuff into the same space and that's why I used the analogy of maybe containers are like mopeds and hypervisors are like cars in a big congested city but we should look at the biggest cities of the world like Sao Paulo or Delhi or something and see how the heck do they get so many people to move around so quickly every day and it is by optimizing the system and using the same street area for multiple different purposes and sort of really maximizing the heck of it which we have never learned to do in the Western world. Los Angeles is heavily congested partly because every car is of the same massive size and there's nothing else. There are no buses practically and no mopeds and none of that. So it's an analogy but I do think it will apply to the open stack world and we'll learn to optimize the heck out of it and I think it will happen through vendors and we will specialize in our certain ways and then it'll come to us because we make it very easy to consume and we figure out a way that's suitable for your workloads. And of course HP, we sell to very large enterprises who will do massive, massive things so we optimize for them and I guess Piston optimizes for a slightly different target market. Yeah and I think we're like, so Swiss stack is in a little bit different position because we're just focused on open stack Swift and the object storage workload and after working on the project for several years at this point, we've now we've moved to more of that download model particularly for testing and development and now we're at a point where you can go to, so Swissstack.com you can go create an account, download the software and not really engage with a project manager to get up and running and so I think with this project with Swift, with what we've done around Swissstack, you're starting to see what the maturity is gonna start to look like coming from the ecosystem. So I'm looking forward to the stats that I hope you'll publish one day about how many petabytes are managed by Swift in the world and showing a curve growing. That's an awesome idea. Do you have that? We have the data. So just picking up, Marta, you mentioned about how HP thinks about selling open stack versus maybe like Piston. Just pros and cons, how do you guys think about it? Where does Piston fit in terms of a ROI to a customer versus where does HP fit and where does it overlap? Good question. I can't speak for Piston but I do know that at HP we serve very large enterprises. So they're giant IT operations and they do everything you can ever do with computing. They have their own data centers, they outsource some and they use managed clouds and public clouds and we serve them across the whole range. So one of our unique selling points at HP is completeness. If you need a private cloud, we got it. If you need a managed cloud, we got it. If you need a public cloud, we got it. If you need hardware, we got it. If you need services, you got it. Whatever you need, we have it. And we have it worldwide. We're already selling in every major region in the world. So, and of course, large companies need that and that's where it's what we specialize in. Whereas I would assume that Piston takes a more limited geographic scope and limited scope on certain types of customers. I imagine, Jim, you don't have people on every comment. Not yet. Not until I buy HP. That was Josh's joke, I kind of stole it. So, yeah, I think we obviously have to pick our battles. And I think it's not just geographic battles. It's also this light bulb that has gone on for the teams that we work with that, their application. The agility light bulb has gone on. The architecture of their application has been cloud from day one and they're looking at either rapid growth of in-house infrastructure in a more efficient way or looking at bringing public cloud workloads back in-house without having to hire a huge staff of SIS admins. And so it's that use case that we're kind of laser focused on. I think Martin's point though, because I think at the end of the day, the underpinning technology is the same and we completely agree that whether it's VMs or containers or bare metal, big data instances, the infrastructure shouldn't really care and the customer shouldn't have to do specialized work to get that particular workload onto their infrastructure. And so I think a lot of the technology challenges, even though the use cases are different, are very similar. How about Joe and storage with you competing? How do you think of that? What do you use at SwissStack to differentiate and what's a better fit for let's say maybe an HP or maybe even a more traditional solution like from EMC? So like with HP, you're a customer. So that's easy, right? Well, let me back up a bit. We focus on a very specific use case, which is object storage for unstructured data. And that means we're not taking on database workloads. We're not taking on virtual machine workloads. We're taking backups, archives, documents, media, and that's the work that we focus on and that's all we do. And so when a use case comes up, that is to take that on, we get calls from other folks in the OpenStack ecosystem to help them to take that on because we're so focused on that use case. And I think where we differentiate against other competitors like with an EMC, they're taking on those types of workloads which are more traditional workloads which are still absolutely essential for applications. It's just as the scale in the data center gets bigger, as applications become larger and larger and you have this trend towards software as a service, which means more users are using an application and that's just growing in the data center versus each enterprise having their own version of exchange server in the closet. That's a situation where you can use general purpose storage. But when that exchange server in the closet moves to Zimra or metaphorically, then the scale becomes much larger and the storage requirements become more specialized and that's where we come in. So let me pause. Any questions from the audience? We've got about 10 minutes left. Yes. Yes. Yes. For legacy. Yeah. Have a short tactical answer then pass the mic. Yeah, so very pragmatically what we've done is we've also realized that. So Swift is an object API and so what we've implemented is a file system gateway that speaks SIFS and NFS so that we can offload the type of storage which is optimized for unstructured data into it and so that's a way that we help out with onboarding and then the transition's important so what we've executed is you can put files in and they get objects out so that way in 2019 or whatever the timeframe we spoke of when you're ready to build those next applications the objects are waiting for you so they can be served out. So pragmatically we've been thinking about that and have built that. Yeah, so first off, our current go-to-market strategy is hopefully not our indefinite go-to-market strategy. It's a focus of a small company to solve I think the most pain that exists in the market for the product and listen, we didn't come to this without talking to hundreds of enterprise customers like yourself and the conclusion that we came to as a small company is that our architecture of web scale everything hyperconverged, the technology is ready for the enterprise but the enterprise isn't ready for the technology if that makes sense, right? There's too many antibodies around CCIE certifications and EMC relationships and pre-bis of hardware and VMware multi-year contracts and that the enterprise is just kind of not ready for web scale open-scale. Have you run it? That's today and hopefully that changes. But if you run it, it sounds like it was more of a how do I move my legacy applications? Yeah, but in the meantime, how do you do that? Then I would introduce you to my friend Martin Picos because companies like his or even what VMware is doing, my old friends there I think are at this stage probably perfectly suited to spend lots of time with you. Thanks for the lead. Thanks for the lead, Jim. That's very much what we do and when HP decided a few years ago to become a leading vendor of cloud, they knew that it meant building fantastic cloud products but also becoming experts on the transition from what we call the classical IT to the new style of IT. So that's why it is for HP a giant undertaking because we are not just building products on top of OpenStack and Cloud Foundry and so on, we're also building a practice of transitioning, helping our customers transition from the old to the new and managing, there are many ways of dealing with legacy. Sometimes you just isolate it and manage it as such. Sometimes you isolate it and manage it with new tools and sometimes you migrate it to the new environment and run it on top of the new environment. The good news with a cloud platform is that you can of course within an image and an instance you can spin up any environment you like. So the underlying platform can be very modern and you can spin up an instance that mimics an older world environment with an older operating system and so on. So for us it's a very high priority so I'd be happy to have a discussion at American Airlines and others. I'd be interested to know at American Airlines, do you still run mainframes? Probably a few? Okay. Yeah, gotcha. Yeah. But it's an excellent point, I mean we still live in an age of mainframes. I mean it's interesting in a history of compute, people rarely ever rewrite applications. It like never happens. So I think living with a diverse IT environment is probably something that will be here. 99% of switch ports by 2019, is that a friend's at Gartner or who? IDC? Listen, I don't know if we'll get there but I would say if that happens we've all collectively kind of failed to some extent because I think the inevitable is that the big guys are on pace to do it more efficiently than we can but that shouldn't be the case and I think these Amazon savings that we're seeing today in real world environments are a great example and I think if IDC's right then we've all kind of screwed up. Any other questions? Any other quick questions in the audience? Well let me get a question then for the panel. So from a startup perspective or from an HP perspective what's the biggest myth about OpenStack today that you deal with in talking with customers? In terms of what they believe to be true and is not the case or what is the case and they just don't believe it. Specifically with storage it's how quickly they can get up and running and how product companies have solved that and I think there's this eye opening that happens when they use a product that's a little bit more refined that can glue together open source software and commodity hardware and I think that's a shock for people at this stage. Yeah, I think outside this conference the biggest myth is that OpenStack is not ready for production deployments and it's propagated by some of our friends at VMware and they're wrong, it is and it is and I think all of these use cases are number one and then the one that I think we deal with the most is that the public cloud is cheaper and it's not, if you do it wrong, it is. If you do it right, you can save a lot of money. What do you think Martin? Yeah, I don't really have anything to add. I agree with those and I'm so new at my job I haven't seen any of them. All right, well with that, listen, thank you everybody. Give our panel a big round of applause. Thank you.