 All right welcome everybody we'll get we'll get going so the title of today's panel is really just talking about open source and how to how to make money how to build it how to build a business and My name is Ryan Floyd. I'm a I'm a venture investor with storm ventures and a fund at a bunch of open source Companies, but but really I think who you want to hear from our people that have actually built them So Adam Johnson's from Midokura Chris McGowan is from Piston now Cisco and Ben Cherian was at ink tank and now part of red hat So three experts I think in terms of really thinking through the business models around around open source So maybe maybe just to kind of get it going to set kind of set the context We had a conversation before the panel about the different models So maybe that's a good place to start just talking about how you how you can build a business around open source and take it from there Okay, I can start Midokura had started a Network virtualization platform. It was fully proprietary originally and at the open stack Paris summit we open sourced the code We built, you know a GUI and some other features on top. So we ended up having an open core model So I think that's definitely one of the models, which is is quite popular. There's a bunch of other companies in the space doing that So that's that's what I know Just for a little bit of background we can go through other. Let's come back to Yeah, you guys open source in a second. Yeah, yeah, Chris much talk about the model at Piston and how Cisco is gonna open source everything So Piston's model was actually an entirely proprietary central or Decentralized distributed controller that used and orchestrated open stack So we were really about like anything that touched open stack We would open source anything that touched our distributed system that we called moxie was ours And since we've been acquired by Cisco's there's talk internally about open sourcing moxie and releasing that it still Has to go through lawyers and you know executives and all the fun Happy fun thing that happens when you work for a company that has more people than a small city So at ink tank what we did was a open core model actually we started out being straight open source And as we started getting some traction in the market and actually kind of right before the acquisition We decided to take our our management tools, which we call calamari and make that proprietary Once we got acquired by red hat we open sourced everything But I think to answer the question that you're asking Let's sew out a context of what we've all done but more in terms of what are the business models You have companies that do pure open source and generally that's a professional services and support play you have open core Which means that maybe a big chunk of it or or a huge part of the value prop is the Open source part, but there are pieces that are added on top of it that are have different licensing structures depending on what kind of Model that is there's some companies that do dual licensing where they might have an open model Like my sequel did do a licensing where you have an open model, but if you want to put it into a commercial product It's a proprietary license and as a whole You know a lot of these tend to be Professional services heavy companies and as they start to build out usually the proprietary piece of it That's where those sorts of subscriptions scale out And then you have businesses like Google and the such that rely heavily on open source but deliver a you know a service or a proprietary solution and I think what Chris was doing at Piston is interesting because they were delivering a full open stack Distribution that but it wasn't just about open stack. That was completely proprietary So the pieces of it that they used were open, but what they delivered to the end user was proprietary Yeah, and I guess to some extent I'd like to go back to the dual license thing It often tends to be from a business model perspective that you start evaluating and start describing commercial uses more and more liberally to the point where if you're just hosting a website The my sequel sales organization actually would come to you and say hey, you need to pay us Which sort of gets away from the open model and the the GPL license that they were using So so how so let's so and we'll take questions here in a second You just kind of bear with us to set set the context So the most simplest form is this services model and I think most people think of red hat, right? And that's not all red hat does but that you know, it's kind of the epitome coming There's other companies like Hortonworks that are that are newer Do you guys think you can build a company with that sort of model today? I mean Hortonworks sort of made it work Let me answer the red hat question red hat would not consider themselves a services company Even though, you know a big part of the value prop of red hat is having good good technical support and good services function They also have professional services function but what they do as a whole is they sell subscriptions and Because it's just the complexity of Linux or main product is Linux and the complexity of Linux and being able to Not only certify what you know hardware platforms are there, but be able to handpick what packages there are and have influence And what is bundled in there? That curation I think is important and valuable for end-users. So When red hat goes to market They're considered more of a subscription Business rather than a services business Cloudera on the other hand I Think when you look at the numbers when you look at What what people are at least anecdotally saying about that is it seems to be more of a professional services Paying less of a subscription play and I think that in emerging markets where technology is not as defined as much It tends to be more professional services rather than subscriptions I think just to some extent about red hat specifically is that red hat Modeling themselves as a subscription license is more of an arc an accounting artifact Not what people are paying for nobody wants a license a subscription license to red hat what they want is at 4 30 in the morning when your email server fell over or Whatever to be able to call up and have someone fix it and it's the services aspect that people care about and what they're actually paying for And I would just say if as a startup if you're looking at a She's got to keep him honest So as a startup if you're looking at a services business model I think that the biggest challenge is finding someone to invest in that right because the challenges in doing that is how do you scale out? Services because it's tied to humans right so I think that that's that's one of the bigger challenges So generally people are looking at are one of the ways to to build a business on open source That's not services dependent even though there it's I think it's pretty hard to escape any amount of services It's it's a necessary part to get the product out there especially on open stack for example, but Still I think that's the hasn't the the reason why people hesitate to you know focus entirely on that as the main model and Horton works in cloud era managed to sort of scale their services business by doing it around training and Licensing and consulting around those sorts of not One-to-one so you can have a class of 30 people you only have to hire one Trainer for that, but they can also do the next class and the next class in the next class And as you move more and more of the training which is professional services income into self-surveys self-managed Web classes like a MOOC You can actually scale professional services income in a way that we previously weren't able to do You know as an investor I guess what I'd add to that too is when you think about like cloud era Horton works or red hat They're enormously large Projects with a very very large user base when you think of something like open stack It's a combination of many arguably smaller projects and a smaller base of Users and I think that you know there's elements of that that makes it difficult as well potentially in terms of how you think about building And around services versus You know some kind of product, so can I can I add one thing? Okay, I Forgot it I forgot to talk about service providers as well like managed service providers particularly in the open stack context All right very valuable business model. It's very successful whenever there's a product of high complex complexity And there's a handful of people in the world that know how to do that and there are people that want to consume that But they don't have the operational expertise to there's a gap over there And I think managed service providers fit well Cisco is doing this of course You know IBM is doing this with blue box. There's a lot of a lot of people that are coming to market with this as well and then you also have you also have the as-a-service so you have dream host and Arguably rack space that are doing the same thing where they're taking the open-source bits and they're providing it as a service That's scalable to the end user and scalable from a Accounting perspective While at the same time they have to hire a bunch of people to manage it themselves So it's talking about that about one of the other you know models here Open core which seems to be kind of where a lot of people have been drawn to I think if you ask Cloudera that's probably what Cloudera would say with Impala and so forth That's certainly what like coming like Swift stack is doing What as you guys have been through some thinking about the challenges with customers There's this fine line you have to walk with open core where you want on the one hand maintain the community and deliver value But on the other hand figure out how to get paid. So how do you think about that? That's that's something we at Meadow career. We definitely struggled like what model we go. Do we go with and I think for us? a pure pure open Model was attractive in some ways, but then we also feared that you know people are just just paying for support And they will be you know, they'll be able to calculate more easily what that value is Versus your our original proprietary model was based on the software Not just the support. So we were afraid that people say why am I paying this same price now that you open sourced it? So that's why we landed on an open core model where there is Something visual that you can actually see, you know, you're getting this value out on top and actually does provide value But it's not where most of our development goes Most of the development still goes in the open source because that's you know the foundation Well, that was kind of our thinking when we when we went that way and after we open sourced We didn't see a lot of pushback on the pricing. We didn't change the pricing We kept everything the same and you know, we got marketing effects from that and it only helped us We'd originally built our what we initially called piston enterprise open stack as if though it was going to be an open core Distribution that all of the technology would eventually be opened up in a rolling fashion contributed back to the community and we tried twice to to Handback our cloud boot bare metal provisioning system and the opens that community wanted to go down a different path and wanted to do originally Bare metal Nova and that ended up becoming ironic And so they weren't interested in the technology we'd built and it would have been too difficult for us to have like Released it and then retroactively tried to retrofit it into the model that ironic had and so we didn't actually pursue that In retrospect, we should have actually provided a mechanism by which it was escrowed The code was escrowed and then released like automatically Because when you're building your business you get busy and things that aren't immediately important fall off the way side that's good point and well before I answer the intangue question on the open core model and in struggles around open core Everything tends to when you have a big part of your product that's open and then you have a small sliver like we had the management layer that was An open piece. I mean a proprietary piece If you're not bringing a whole lot of value at that top layer people really question why it makes sense for that to be proprietary and And I think even over time even if you do bring value to that layer What'll happen is you'll have the community come up with reactions to that over time because they tend to if you've got a Big enough community they'll tend to want to create pieces of it that are open fit within their licensing model fit Within their business without having to have that relationship. So We saw that particularly at ink tank. So Seth as a whole the whole lifestyle life cycle of product everything was a hundred percent open And it wasn't until ink tank came along and particularly towards the last, you know, maybe year of ink tank Did we decide to keep a piece and make it proprietary and the reaction that the market had to it was interesting We saw several people come up with management consoles that had similar functionality We saw a large partner of ours also come up with that Intel came up with something called virtual storage manager That did that and there was also some talk at that time This is back when Intel was doing a lot more software work and had their own distributions There was talk about Intel doing a distribution of stuff at that time So some of that is in reaction to the open core model When you go a little bit against the grain of the system the system pushes back a bit the market pushes back and forces you to Change what's open and where you put your value bits and what pieces you can keep proprietary And really when you capture value or when you create value, you can only capture about 10% of that piston was actually working on a Component that was going to basically be our version of calamari because we weren't willing to Spend the price that ink tank was asking for and we were really good partners with ink tank. They try to sell But our licensing model was actually only $3,500 per node per year Which was a lot less than everybody else in the open stack ecosystem was charging for their more open Open source or the open stack deployments and distributions So there wasn't enough money in our margin to be able to afford to bundle calamari And I think the other the other challenge with open core is where do you draw the line of proprietary versus open and How do you deal with it when the community does? Encroach on that proprietary bit that you have are you fighting against that? Are you accepting it? You know, there are a lot of cases of other open core companies in the past where they fought against it, right? And it caused all kinds of problems. So I think that's that's something that early on I think if you can if you can find a line that's very clear That's not fuzzy so that if you're trying to make a decision on a new feature You should easily be able to figure out is this an open or closed feature I think that's important to decide early on because you get busy and you may make compromises later on That could hurt you Clarity the line is important communication of that roadmap in both independent roadmaps because when you have a proprietary piece And a non proprietary piece the roadmaps are handled differently So communication around that roadmap is important The ability to move the line because as the community comes up with alternatives You might have to open source was previously Proprietary so that means that you still need to keep good coding habits You need to make sure that everything that you're doing can be open sourced in a very easy way afterwards and You have to innovate at the proprietary development piece So you have to invest in proprietary development if you want to stick with an open core model Because you have to continually bring better value to the end user through the proprietary model as a whole So I think all of those are important things to do if you're going to do open core and do it do it Properly particularly within the enterprise although completely unrelated to the actually act of doing the open core If you ever want to see an engineer clean up code Overnight tell them that they're you're open sourcing the module. They're working on and then just stand back. I Might not actually be a nice person. I'm not sure What What what do you why do you guys think? Enterprise is actually value open source like what's what's we know that the trend is undeniable right in terms of all that You know so much infrastructure is moving to open source, but why why do they value it? They have money, right? They're willing to pay for things what what drives it there. Do you think? Well, I used to work a dream host and I don't know if we represent the standard enterprise We really had an open first point of view particularly being a service provider margins were Very low, so we need to eke out a lot of margins. We also had a lot of very unique internal systems that Being open helped tie into our own management platforms our monitoring systems things of that sort that had years and years of Work behind it that we needed the flexibility of it From a service provider point of view I can say that it was very important for it to be open for it to actually even function within an Organization now, I don't think that's the way that all enterprises have they don't have that same type of legacy but what I've been seeing is that the Infrastructure level of the market is now all Tending to be open first the conversations start at what what's the open way to do this? Versus, you know, let me use this closed system now sometimes in closed ecosystems where they're only used to that They don't know how to ask that question, but those workloads seem to be shifting over time So I think that if you're building something in this day and age in the infrastructure space that category We have to answer the question why be open and I was having conversation with somebody else about this I think unless if you do build something infrastructure space and you want to Further commoditize the market may be being open is great if you do something that bills is creating a new market and You have a substantial amount of value Over-existing open source solutions that I think proprietary might not be a bad way to go For us Having having a network plug-in in the open-stack ecosystem where open stack is open source But the networking plug-in was not was causing us You know problems getting traction in the community with customers when there is open alternatives like open v-switch plug-in and DVR and things like this So, I mean that was the driver for us for open sourcing and it had a major impact, right? We were able to differentiate against other proprietary Solutions out there. We got a lot more visibility and it it basically, you know checked that box of ok You are open. We're not having the lock-in problem, it's You know it it just helped us tremendously get a lot more traction So if you if you extend it Further if you think about the application layer, right? We talked about this a little bit before that There's very few open source examples at the application layer, right with sugar Really being the only one I can kind of think of off top of my head. That's had reasonable success So why is that? Why why is it such a different story in infrastructure yet at the application layer? proprietary rules the use case the people the people who are actually using the software that are applications Care more about getting things done than being able to prevent being locked in like a big part of why? Enterprises tend toward open and the infrastructure level is the systems administrators and the engineers who are responsible for maintaining the systems Have come up in a world where things are tending toward more open but they also have a an aversion to vendor lock-in that Your application your end user doesn't doesn't have that. I actually Was building, you know infrastructure software But I use an iPhone because I don't want to have to recompile my kernel every time I get a you know firmware update And you know install sanogen and then oh no Now I have to actually install a kernel update because someone left a 40,000 character password root hack on in the initial thing I don't want to deal with that. I just want to you know text my wife and you know Facebook with my parents And at the same level like your marketing team Your sales team they just want to get their job done and the open-source alternatives for at the application layer get in the way I think that on-premise software is The the amount of on-premise software sold versus software as a service sold You know it suffers a service is starting to dwarf everything else And if you're going to present something as a service and as an application that is hosted You can consume open all you want but What people want is an easy way to be able to use it without having to have their IT staff deploy throughout the enterprise In in such a heavy way like for example, that's one reason why Applications like Expensifier really killing it in the market right now because while there's lots of applications that can do that type of work that our internal systems When you have a really good hosted platform that is as thoughtful as great user experience that as Chris said Can get stuff done quickly your productivity is high because of it people want to use it and To answer your question your your question is a good question and your ticket is 7,543 and we'll get to you in the next 40 to 72 days All right Is there any questions from the audience? Yeah If you can make it we're gonna repeat it from from an investor perspective How do you measure the value of? intellectual property And how much you want to expose that? Do do you want to open-source that what is the value? What's a trade-off and how do you look at that in particular in an investment? Yeah, I give my opinion. I'm sure these guys got an opinion about that, too I actually may have got a controversial view on that I actually don't value IP very much especially in software. I just don't think there's very much IP. I mean You know if we're talking about you know Intel and transistor design There's some IP there right but in in code. It's just it's a question of time I mean, it's just a question of time in engineers And there's nobody that has a lock on all the great engineers The the craziest thing I hear from startups is they think they've got better engineers than you know Cisco or Red Hat it's not true, right? They may be able to move at a different pace because maybe they don't have the same organizational constraints There's a lot of great engineers out there. And so I actually don't value IP very much I I look more towards like the business issues. So thinking about the open-source question It would be if we were to open-source something What what value does it drive right? Is it is there a community reason that we're doing it? Are we not going to be able to really get paid for it anyway? So even so someone's all sorry even if like, you know, we didn't open source We really can't charge for it anyway Well, we may as well open-source it then right because then maybe we build a community We get some additional goodwill and value out of it So it's more kind of business questions than maybe technical ones So I was on the previous panel on being famous in open source apparently I am But Randy bias actually made a very poignant comment about how in the early stages of building cloud scaling that all of the customers that he went to Didn't it wasn't an open-source or closed-source thing. They just went and re-implemented his number one competitor was not piston was not Morantis was not red hat was not, you know That other company nebula It was companies like Walmart and you know, J.P. Morgan Chase and all of these people going and re-implementing the products that we had built and Once you've told someone the idea you have they will go and do it mass and juju exist because We both cloud scaling a piston sort of paved the way for oh, that's a great idea People should be able to bear metal provision stuff and it should be, you know, hands-free seamless and at a certain point you have that You the decision is not how valuable is this IP? It's what can I build that's valuable with this IP? An example is keybase. I'm not sure if anybody's familiar with them. They're an encryption like they're making PGP easy and they just launched a new beta service that's Peer it's basically centralized storage, but it's entirely encrypted on both ends So if you and I were to decide we're gonna share some files You I make a file a folder on the keybase file system with your username and my username and when you come online will mute our clients will mutually create a Key and encrypt all the files and if I can if I upload them first They'll be encrypted with my key and then they'll be rekeyed later when you decide to accept my invite They're not charging for that It's entirely open source and one day they will be charging for that because they're finding ways that the core IP That's entirely open source all it's Apache to it's all in GitHub They're finding ways to build value on top of that that's completely separate from the actual code I think these days it well in general it just comes down to execution I agree with Ryan, you know, we we open sourced our stuff and you know Some people were worried that you know our competitors will steal our ideas around distributed networking or something that didn't happen, right? We actually work with some of the other open source Groups out there and guide them on here's what we did here's what we learned They could go look at the code, but they don't because everyone's busy, you know So in reality, we don't find that people are going and stealing your IP Even if you give it all away You actually have to put effort into spreading it if you want it spread also all your code is in Japanese and It's actually Catalonian, sorry Any other questions? Yeah Actually somebody mentioned escrow at one stage so and most of the time when you've been talking about business models You've been talking about the shiny new thing at the beginning So I'd like the panel to talk a bit about the end of the life cycle the sort of When it's not a shiny new product anymore, but you still need maintenance on it. Are there business models around there that make sense? No Not unless you want to get into a services business if you want to have a scalable business like it's expensive Every piece of like every piece of code you build is a debt for the future and if you have to maintain that forever It's basically it's impossible. It's too expensive So you have to life cycle it you have to tell people all right after a certain point in time This isn't supported anymore. We'll put the code in escrow. You can maintain it yourself, but don't call us. We'll call you I mean, I think I mean, you know the there are certainly large companies out there like NCR That make quite a bit of money Maintaining legacy code bases But I don't really see that as an particularly interesting business and I think it's hard to get into as a small company because you don't have the credibility the reason NCR wins all the businesses because it's NCR and it's their Terminal from 1960 that someone's got to do something on right? Same with you know IBM mainframes. I don't know how big that business is inside of IBM They sure don't sell many mainframes anymore They sell a couple but it's that services group that probably does you know pretty well I bet it's incredibly profitable with an IBM is my guess But it would be hard to start that outside of IBM Because how do you get the customer list? How do you know but it is I think Chris's point ultimately going to be a services business Microsoft has a has a good business. That's an embedded business on Windows 95 still but I Wouldn't call it a growing market And I don't think anybody would be terribly excited about it. I actually worked at a company that was using you verse which was a multi-value key Proto key value store database non-relational database from the 70s on that up until like three years ago was owned by IBM and IBM spun it off and sold it to someone else because it was an expensive business that wasn't making a lot of money and This company won't move off of it because they tried to move off to something more modern in 2008 and it nearly killed them Just a part parting parting shop from my standpoint to I'm not so concerned about growth I mean, it's not gonna be business for me as a venture investor if it's not growing Is that's just that's what I'm looking to go do but there's plenty of private equity firms that that's their whole business, right? I mean solar winds if anyone's familiar with with that company Tomor Bravo has bought that and is rolled in there's a service called Pingdom. Maybe some people might be familiar with that They were a vibrato. They basically are bringing together all these disparate pieces and they're gonna milk the Service right, you know the maintenance stream out of these companies and they may talk about investing in R&D But they're not gonna invest in R&D. I mean, it's an it's an asset That's kind of that they're gonna basically squeeze the value out of and that's a profitable model It's just it's just not growth Any other questions? So so question for the for the for the other for the panel so we talked a lot about startups I think we're all geared towards startups You know Ben you may be starting now to go to the dark side. I don't I don't know but Maybe I don't know. I'm just happy. I'm seeing next to someone famous There's a lot of big companies that have come into the open-stack ecosystem and Certainly from you know 2010 days, which you know Chris was there at the beat that the very well you guys were there at the beginning How do the big companies best leverage open-source from a business model standpoint because they're they're in this unique position They don't necessarily need to make any money from open source. It can drive I don't know hardware sales or something else like how if you were inside You know a big a big company today, whether it be IBM or HP or Cisco or Dell or EMS What advice or how would you sort of think about the open-source business question? I Think that what we're seeing right now is we're seeing a lot of big companies Getting it to open-stack I've heard this actually Chris you brought up this term earlier about the idea of some people look at open-stack as a lost leader and They're attaching it to because they have other things that they can sell there's hardware and switches and Professional services or whatever but there's a lot of complexity in this and the amount of engineering work It takes to continually maintain this and have have a team that Can deliver good quality product and be competitive is very expensive So if people look at as a the lost leader that means that the price on these services or price on these things can go To free it's not a sustainable business model So what I predict over time is that if that is the goal of some of these larger companies Just like they tried to do that with Linux in the past before I don't believe it's going to be a sustainable model And I think that only pure play Platforms are going to be able to succeed in that space There'll be an ecosystem of other people of course services providers and people who might want to consume it, but I think from a From a top-line perspective here We'll probably see a shake out and there'll be a handful of companies left standing and these are the ones that Have core value in building these open-source systems and distributing it So I mean I think for some of the large companies who are trying to do things like build their own public cloud You could debate whether that's you know a good thing for them to do or not But for these companies trying to go down that road There's a lot of logic for them to embrace that and build up the expertise in house Because they can control their own destiny But on the other hand is this gonna be successful for them anyways? Yeah, I don't know But but I think that that's why they're doing it But I probably wouldn't do that if I were them personally so I'm gonna take somewhat the opposite approach I'm actually an open-stack skeptic and I have been since I found it in 2011 Sorry, Tony 10 at the Austin Summit I was one of the four four the 12 technical talks that were there and I've been amazed at watching how it's grown but If I were one of the big companies I would be building in a public cloud and taking one of my existing revenue streams and putting it on there and saying if You want to adopt this revenue stream in cloud You have to use ours. There's a large database company in Redwood Shores, is that where they're located? Starts with an O ends with rock recall They're building a cloud They're using open stack as their development environment to do so And they are going to put or go on the or sorry their database on to this platform And if you want to use this database in cloud You'll be using them and they have a large enough revenue stream This is a business a legacy business as making them lots and lots of money and it allows them to move Holy into a new Cloud-based ecosystem while still not abandoning or offering their existing revenue stream. So if I were HP IBM Cisco I would be doing that as well What is one piece of software technology that we can move entirely into the cloud? And how can we leverage our own ownership of that revenue stream to make this cloud money? Great thing about it Yeah, I saw a question over yet quick It's not a it's not a question so much as adding on to the question that was just post Because I used to work for HP and I work for Red Hat now So I found your question to be very interesting just wanted to share what I have seen HP do what I saw HP do So HP has the software You know their own proprietary software so the strategy that I saw work there to some extent was let's complement Let's go leverage open-source components to kind of complement the HP proprietary software stack That was one but what I found more interesting was internally within HP to kind of drive that open source mindset So kind of have an internet of open source You know stores and you know getting the employees to collaborate more using the open source culture So that was something else that I saw HP do I'm not saying it made them a lot of money But that was an interesting application of the cultural aspects that was it and if I was HP the the The revenue stream that I would move entirely into the cloud would be the reverse polis notation calculators that they had in the 70s I think that one's got wings. It's a killer application All right with that I see it's it's 420 listen. Thank you all for coming. Thank you. Thank your panelists Thank you