 Now is it on oh look amazing what happens when you turn things on How's everybody doing? sleepy yet Back on the time zones figured out It's been a good week. I think great content so far in the show. Thanks for joining me So I'm Jesse Proudman CTO of blue box. You can follow me on Twitter at blue box Jesse I'm here today To talk about the death of the distro in the future of open stack My belief in where the future of open stack is headed from a service provider perspective How we make organizations the most successful with open stack. So I'd like to start back at the beginning Way back in 2010 when open stack was born project was initiated from rack space and NASA and And it's designed to create an agile infrastructure platform to make deployments of VMs more easy So Nova and Swift for the foundation important foundation we need We need compute obviously we're gonna have a cloud although today you might argue that to be containers but back then we needed to compute object storage, I think was an interesting choice, but Going compared to what was going on with Amazon at the time, but Swift was a great contribution and a great set of code and arguably some of the most Reliable and some of the best code in the open stack project today along with Nova, but we were missing quite a lot We're missing block storage. We looked at what was going in EBS with Amazon We were missing installation capabilities upgrade capabilities orchestration capabilities monitoring logging all of these pieces that it takes to actually Run and operate an open stack environment weren't in the original set of code And that's fine because it's an open-source project and we got to start somewhere and that gap Left a market opportunity Left a market opportunity for organizations and so we all jumped in Bunch of vendors said hey, here's here's an opportunity to fix a problem How do you make money in this business you find something? It's painful and you solve it And you solve it in the most reliable and simple way and you provide a great experience doing that and so The distribution was born People looked at Open stack all the different pieces and bits and Services and you need to actually deploy and operate one of these clouds and realized to some extent it was a huge set of spaghetti You look at just Python dependency management to get all the open stack pieces working How do I track each individuals? Individual specific version of each individual specific Python library for each individual specific project and that becomes a giant pain And you've got to do that not just against Python and all those libraries But the database engine the message of us all of the additional services that you need and If you think about it, there was no HA an open stack when it was born so People had to figure out how do I take this? Single service and make it highly available because what's a cloud if the control plane isn't highly available. It's useless so The distribution was born Open stack is hard. It moves very very quickly and we have a lot of debate about this We're on a six month release cycle today Which if you think about from an enterprise perspective that that's insanely fast if you think about it from a public cloud perspective It's insanely slow But it moves quickly and it is growing at an ever-increasing rate. So not only Do you have a fast-moving project? But you have an ever-expanding project with ever expanding complexity? That's a challenging problem It's a distributed system Distributed systems are hard It's very difficult to find application engineers period It's harder to find application engineers who understand distributed systems And it's even harder to understand application engineers than understand Python distributed systems and open stack. It's a talent crunch We're all aware of that anybody who's working in the space whether they're trying to hire Talent or they are an engineer and are being approached by 50,000 recruiters at every summit It's a difficult space And that makes open stack itself even more difficult Python Python's actually not too difficult, but it's a unique skill and It's one that much like that application engineer problem You've got to go find the people that understand the language that this technology is written in And then lastly time So all of these things are solvable with time But nobody has any time right time is the most valuable Resource to any organization more more so than money because time is money for these organizations And the whole point of adopting a cloud platform is to increase an organization's velocity and increase an organization's agility And if I'm now taking my best engineers in my organization and having them focused to learn all of these pieces How am I actually benefiting my organization? What business difference am I making? So that was a big challenge. So How how we go back to 2013 how are we gonna solve that? Well Very early on there were very few organizations that actually had patience for all of those components Many organizations are saying look this is this is very different from how I'm used to delivering my IT I have VMware and I use put a ticket in and then two weeks later I read your ticket and I create the VM and I hand you back credentials and we're good to go or maybe we're out of capacity So it takes me four months to buy another server from Dell and put it in the data center And maybe didn't work and but it works like that's how we've done it for years And that's how we're comfortable doing it. So why would I try this whole new thing? That's that's very difficult particularly when I don't have time to learn all the skills Isn't there a faster way? Can't we have the pudding without eating meat first like it people just we're trying to experiment and figure out what we were gonna do So around the Diablo and Essex release Timing's approximate a bunch of smart people had a great idea if we could simplify All those pieces of OpenStack the velocity challenge Python all the dependency management And we could provide more choice than just taking the source and Using that or going to rack space and getting a public cloud account on OpenStack There might be a business opportunity there. That's a problem. That's a challenge We think people might be willing to pay for that And so with that in mind the OpenStack distribution was born and we saw a lot of great companies come in very early and They packaged everything in a way that that was meaningful Then the sea came pouring in so we had a bunch of independence Many of those were companies that were founded by folks originally involved in the creation of the OpenStack source code to begin with No, it's he came pouring in we have red hats who see HP Ubuntu cannot or canonical Everybody started making an OpenStack distribution. That was the popular hip thing to do IBM had an OpenStack distribution If you actually look if you go to the OpenStack foundation website, they have a marketplace And that marketplace lists all the vendors that sell OpenStack services So go to this website click on marketplace click on distributions I think there's like 27 different software distributions listed there in different vendors So 27 people taking the exact same source code Packaging it and trying to sell it as something that's differentiated and adds value that's a pretty pretty thick market and at the end of the day You've got to convince me that that's that's the right approach We can look at the history of some of those companies the three I I show that started aren't in existence anymore Cloud Skilling was acquired piston was acquired nebula was unfortunately went out of business now that team is part of Oracle not working on OpenStack and So that model Feels flawed and you look at some of the challenges in the other major enterprise OpenStack distribution vendors Nobody is sort of shouting from the rooftop that OpenStack is driving Material portions of the revenue So people are using these distributions. It's not as if they are going unused But that that wasn't sort of this panacea to generate revenue from OpenStack If you look at 451 research every year does something called the OpenStack pulse And they take they survey all the OpenStack vendors and they ask how much money are you generating? What are your projections look like and from that they put together this forecast of what the market will look like an OpenStack I think in 2017 or 2018. I'm gonna get this totally wrong, but it's something like three billion dollars projected In the market and the interesting thing about that study is actually that the bulk of the revenue There is derived from service providers Not from software distributions not from sort of networking plug-in companies not from All these these individual peak monitoring companies or tooling companies or configuration management, but from service providers So something there tells me that software approach isn't isn't what the market's asking for But this shows did solve some of the problems installation got faster commercially licensed bits and some of the distributions filled holes that existed in OpenStack for various components Think billing or metering or telemetry or monitoring and they provided software support, which is helpful It's helpful to have somebody that knows what they're doing who can answer your questions So if I call Red Hat and my systems down they can say hey, here's three or four things you can go try That's a that's a benefit. It's worth some amount of money But I don't think that they solved enough of the problems So many of the distributions would I call them open ish, right? We we're using OpenStack the idea here is that each of the each of the bits are all Cross compatible the work we're doing the foundations doing with ref stack. I think is actually adding a ton of value here But a lot of the early OpenStack distributions had enough proprietary Capabilities in them that they weren't really open they weren't open source you couldn't go get The pieces without paying paying for them on many of them Operations was a huge challenge. So installing the cloud like that's That's a solve problem too large except now many people can go install OpenStack It's the Sunday morning at 3 a.m. Problem when that cloud falls over that really is the painful part and We think about the Linux distribution When Sunday morning at 3 a.m. You had a server fall over with your Linux distribution and you needed help Okay, you called the software vendor you were able to get some assistance Maybe you changed some setting and they kind of generally fixed your thing Or maybe they didn't you had to wait until Tuesday or whatever it was But that was one system that was impacted and if you've architected things, maybe you can fail over to another system The business impact there wasn't as large If you think about a cloud that cloud generally begins to underpin everything you're doing in your business Which means the blast radius when it fails It takes out a lot more than that one system and unless you're running two clouds and you're gonna fail everything over to the other cloud Which is a ridiculous amount of work That's a big problem. How do you limit that blast radius? Those people that the organizations that used the open source initiatives or that used the distribution itself that Operations challenge became their problem and it's a huge pain in the neck The monitoring the logging the operations the upgrades those are all big challenges in OpenStack and Particularly when you think about all of the flexibility that was baked into those distributions Merantis has done amazing work in the space. You go look at what they've done with fuel And you can literally use almost every networking plug-in you can install every OpenStack service You can configure things in 14,000 different ways. That's awesome But how do you reliably upgrade that environment to the next version when you've got so many different configuration flags becomes a big big challenge and So it's that that's not how we do it here upgrades Like all right You turn this thing on and we didn't have support for that driver version into the next version or we see we've seen This a lot with cinder drivers where one cinder driver and one version of OpenStack works great And then you go to the next version of OpenStack and the same driver that's commercially supported by the same storage vendor Doesn't actually work in the next version So when you've got all of this flexibility all these options that upgrade process becomes a huge challenge we saw it in the keynotes on What day was that Tuesday? I'm losing track of my days the first day of keynotes that There was the the cloud provider that had four different OpenStack clouds on four different versions of OpenStack going all the way back to Grizzly or earlier Havana So that that's a it's a big pain and when we think about Using OpenStack for the long term for our businesses Those are the things we need to be thinking about how do we operate this environment? How do we make it reliable? How do we upgrade it? How do we make it survive? More than just one single OpenStack release. So at the end of the day the distribution model Didn't convince me. I'm not convinced the distributions are the right way And I think you begin to see that there are critical business barriers that block these open source deployments Because these are big challenges that have open-ended questions and it's hard to actually answer those in the marketplace Or you began to see a bunch of these original These original distribution based installations fail Remember a year or two ago. We would see all these articles about people saying oh, I tried OpenStack and it didn't work It failed my installation failed down back on VMware. I went to a public cloud That's a problem that's derived from that ongoing operations problem and so you begin to see enough of these stories in the marketplace and as an organization you begin to lose The incentive to take that risk to go try that open source project If you've got VMware and your installation and you're running everything there and it's working What's the driver to go try something new? Particularly if all you read our articles that things are failing like that. It's a big challenge that we as a community had to work with And it's one that I think still exists to a lot of to a large extent We also didn't see if we back up a couple years a lot of competitors a lot of companies competitors doing using these types of technologies Those competitors were looking at other options in the marketplace. They were looking at public cloud providers And so the market had time to wait The market isn't necessarily ready and they time to wait so they did and in the meantime the public cloud vendors in the marketplace Evolved very very quickly Because what do you get out of a public cloud you get an experience you get an SLA and you get an API You don't worry about what hard work to choose. You don't worry about what operating system to install You don't worry about what distribution of public cloud Do I buy and install my thing and you certainly don't worry about at 3 a.m. When the data center fails how are you're going to fix it you worry about how you get your application back online? So it moves your spectrum of concern up to your application not to the cloud itself And So these are big challenges that that open stack had that private clouds had And in the meantime public cloud began to get more and more adoption But open stack also began to evolve very very rapidly and evolve not just in the technology but in ways to consume that technology and While that change was happening the market began to get more and more acceptance of open stack So if we think two years ago or a year and a half ago When you went to go talk to a CIO or VP of infrastructure at these companies Nobody said it was no more longer. What is open stack? Everybody had heard of open stack But the question became how do I actually get going with open stack? How do I make sure that if I want to try this technology? I'm not going to have the same experience the same failures the same frustrations that I've read about in the news More and more customers began to say hey Can I consume open stack without actually having to touch open stack and from a public cloud perspective? You had a lot of options you had rack space you had HP public cloud at that time You had a bunch of vendors in Europe smaller vendors that have started I was just in China last week and there are two or three Pretty material public clouds pretty materially sized public clouds running on open stack So there were ways to do it from a public cloud perspective But that private cloud perspective being able to put that environment in your data center Next to your systems of record next to where all your data that you may want to actually interact with in the application You're building resides there wasn't necessarily a solution there And all about the same time you started to see companies like Meta cloud Blue box come to market where we're doing this managed model and more recently you begin to see companies like platform 9 Or a zero stack or easy zero stack come to market and The thinking about how we actually deploy and consume open stack in our data centers we begin to get actual options there So as open stacks evolving as those those different types of consumption methodologies began evolving a funny thing happened Organizations realized that ABS wasn't a distraction that public cloud thing wasn't going away It's a it's a Direction that next generation applications will will be written in cloud native formats The differentiation through infrastructure was a losing game that I've got my data center And I'm going to make my business successful because I'm going to operate that data center better than anybody else that wasn't a viable strategy anymore and that Sustainable success for an organization Rests in the software that they're building not on the platform that they're building it on And so as organizations began to realize that they began to look for alternatives. How are we going to how are we going to make this thing work and? at the same time open stack continued to evolve and We got into the ice house release. We got into the Juno release And we started to recognize that that open stack was ready that open stack works It was no longer a matter of the technology doesn't work Like we were struggling with in early versions like I it literally didn't work But but that operation problem was still a challenge and it wasn't it's not just me saying that like we start to see Gartner And for folks that don't follow the analyst community Gartner has been one of the largest skeptics of open stack and his success For a long time. So it's open stats. It's a success. Sure. It is so Gartner is beginning to say, okay There's something there's something here to this technology you've got Forester is open stack or open stack is ready. Are you so that the analyst community is beginning to say Open stack is ready. I say open stack is ready Gartner says it's ready forester says it's ready IDC says it's ready. Users say it's ready. The market says open stack is ready. The technology is there The technology works that solves customer problems That the question continues to be How do we make this work in our enterprise? And so there's an enterprise disconnect at the same time Some folks and organizations were continuing to call it a science project. You continue to see Articles about open stack failing. There was that nice article that came out last week about British telecom and how they're ripping open stack out of their their infrastructure. The whole article was kind of a Miss number since British telecom was just experimenting with open stack. So there was nothing to rip out They were saying here's the things we think we need to see in the technology before we use it But you see these articles and I think to some extent the media likes to use open stack as a whipping horse Because it's easy or has been easy But we've been struggling with that that installation operations problems So today if even if you've used the distribution, you've had to go hire train and build that installation Or you can let somebody else do it for you And so what we see is that most enterprises that we're talking with wanted to consume open stack without having to touch open stack Most enterprises want a cloud a cloud experience and remember what is a cloud? It's an experience an SLA and an API whether it's public or private. That's ultimately what what we're looking for And so some organizations have solved this and they're rolling their own clouds We see them at the keynotes. They're impressive numbers. You look at the things that CERN is doing You look at the things that Walmart's doing targets doing PayPal eBay these organizations are building engineering teams big engineering teams to focus on their installations and For those companies Building that infrastructure is a differentiator for their business But you look at many other organizations that are saying look ultimately all I want to do is deliver business value I want to be able to build applications. I want solutions powered by my software that run on infrastructure And I don't want to worry about that infrastructure a few are using distros We've seen that model work here there So it's not that distros won't ever work, right? But I think you have to go back and think about that first point Do I have the expertise and the capability to actually fix and operate that environment when it fails? And we've seen a growing number of organizations now looking for an alternative that managed model so Why are they looking for that managed model? Well, it's the complexity It's the tooling. It's the operations challenge But it's okay They know those are challenges now and that's making this this as a service model even more successful So you look at even rack space in the key notes today. They're private cloud is listed as a service the the Zero stack folks they raised the series a in the space six months ago, and they just raised a series B So investors are putting money into this managed model. That's a model that makes sense to organizations so in closing I think Distributions are not the future of open stack I Think service providers are the future of open stack and whether that's service providers from a public cloud perspective or service providers from a private cloud Perspective that's what will make the most material impact into the the wide-scale adoption of open stack I think a few distros will exist maybe two maybe three certainly not ten or twenty five like we have today And it's it's somewhat like the Linux distribution in its early days that we had this huge fracturing Of the distribution and each one had its own sort of special thing, but ultimately the market Collapses onto a few options and those options actually have differentiated values see that today between what's in a bunty or what's in red hat Or what's in Susie? I think business models will continue to evolve So we saw the creation of do it yourself. We saw the creation of the distribution We saw the creation of the appliance we saw the creation of the as a service model There will be future models That they come out to help customers adopt and use open stack And I think that's that's exciting and at the end of the day cloud will keep being cloudy the things things will change we're in a market that is moving insanely quickly and Not just velocity, but the breadth of technology that we're using is increasing at an ever ever faster rate And so it's hard to say what the future will hold you've looked at what's happened in the last year you're alone With Docker and containers and how that's transformed the industry There will be future dockers. There'll be future change in the marketplace and so Ultimately, I think it's the most exciting time. It's an exciting time to be working with cloud It's exciting time to be working with open stack. It's an exciting time to be in this space It's why we're all here It's why we come to these summits every six months because this is a it's a great technology to be working with So I like to quote Randy bias pretty good friend of mine and markets to find standards Customers will come and they will ask for something and they will expect it to work And it's our job in the open stack community to respond to those requests and come up with a solution that works It's also our job to figure out what they're not asking for and to figure out how we We adopt and innovate in that blank space in that space that that isn't a solved problem today Whether it be from a commercial vendor or from an open source solution so with that I Will open it up to any but questions And if there are no questions, I'll give you a few minutes of your day back and thanks for joining me Yeah, great question. So if I'll summarize it for the video. So There are service providers that have existed and failed Because of the complex the same complexity that we all struggle with Organizations need to decide if I get rid of all the talent that I have that actually knows how to operate these environments How do I actually continue to innovate in the application space? Is that a fair summary? Okay, so I Believe that those individuals don't go away from an organization You don't get rid of your team that knows how to operate things you have them operate at a higher level of value in your organization So your application developers are gonna write let's talk about it. Let's talk about a bank Banks are I mean my bank three weeks ago rolled out a new mobile banking application They had engineers in house working on that application And they had to go deploy it and operate it in their environment that environment surely connects back to the transaction system or record The team that they have has to operate that environment So they may have two teams one operating the infrastructure to power that environment and then the actual application itself It's my belief that if you consolidate all that talent and technical expertise in the application delivery layer You can be more successful. So instead of splitting your talent across doing something that that ultimately isn't a core competency Versus moving that core company's competency into something that actually brings you business value You've got opportunity that the other thing I didn't note in the in the talk that I think is important to that same On that same line is that if you think about operations of a cloud environment from a single Customer perspective when something fails, they'll fix it and that fix may get baked back into their environment But that fixed often is challenged to get back up stream into either the open-stack project or into the vendor That's selling the technology The service provider model and this is what's made public clouds in my opinion largely as successful as they've been Every time something fails in a service provider model that fix goes back into the same product That's distributed to every customer So every cloud failure that we have in our managed model that fix goes out to every other customer So every installation we have makes every installation more reliable more successful more scalable And you don't get that when it's only your organization focused on on that component. So That that's my thoughts on that topic great question though Yeah, yeah, so the question was yesterday we had a talk around Open stacks future and in a collision with VMware So I can't remember what the panelists said but my my perspective was that I think they are on a collision course and Not necessarily open stack becoming sort of the VMware replacement But if you look at where customers are going where the market's going that's in cloud native applications Cloud applications cloud native applications need a cloud platform to operate on And VMware recognizes that Open stack recognizes that and so ultimately we're going to end up building similar solutions to solve the same problem Now VMware's business isn't going away, right? They're not going to work VMware is not getting ripped out of every installation that exists But I think that's a declining business Because I think the next generation of applications that we're writing are being written against these next generation technologies I also don't think it makes sense for open stack to focus on Replacing that VMware installation because if it's not broken don't fix it, right? If you go into organization you say you say I've got free VMware You've just got to move everything from the VMware you have to this free VMware. That's exactly the same It has all the same functions But you've got to move everything What does that take? Takes time. Do we have any time? No, nobody has any time, right? So nobody's gonna do that people move things because they're broken or they're making something new And if we focus on what's already there in the marketplace as the open stack community versus what's coming I think we're doomed for failure But if we focus on that next generation application platform We build all the things in that make containers of first-rate citizen that Make bare metal management of first-rate citizen that make VM management and we get that service catalog Spun up and we're seeing that with the big 10 initiative now. We've got a platform. That's ready for those next generation applications That's my take So the question was do I see a way to make upgrades from one release to another easier? Absolutely Absolutely, I think over time all these pieces get easier, right? So what our installation was once a challenge that's become an easier Problem or upgrades are once a challenge every release of open stack that comes out has solved some of some of that hurdle But at the same time we're adding additional complexity and each of those new projects has its own round of upgrade challenges So where today may become easier to upgrade Nova all of those those new initiatives because I haven't reached that same point of maturity in the spectrum And so it's sort of where it's the dog-chasing a tail to some extent We're gonna continue to add services that each of those services will lack maturity that the services that have been there have And so I don't know if we ever sort of reach a point where it's it's a push button simple Yeah, do you want to do it? Yeah, we have to speak in the microphone Introduce yourself There you go. Hi, I'm Sean Roberts. So I work in the part of the product team Which is a I guess it's a working group now. So One of the things I'll just generally say tagging as a general thing There's actually two versions right now and that's being worked out, but in general tagging is is trying to approach the Initially, there's stable release tags but for other things of interoperability type style tags and tags that show levels of maturity in the project and Well, just leave it there in the project itself I think over time is that starts filtering through all the different projects and people get more used to not only accepting that in the projects, but also in Downstream products that are based on that. I think you'll get more of that that the I'm not sure what I'm talking to but So so I think over the next couple of release cycles, you'll start to get the language will start to change over Projects that have accomplished that upgrade ability success will have the matured that some tag of Tag that means maturity or enough will actually say maturity, but So you'll see that over the next couple releases. It's happening now So that the project navigator actually has some sort of maturity rating. So it it does. Yeah, it's show So I think we'll yeah, we'll see that and again I think that the other key piece there is you don't have to use all the pieces right just because they're in open Stack doesn't mean you have to deploy all of them. So certainly if you just use the core primitives compute network storage Those those will get better. They will get more stable. They will be easy to operate over time absolutely Any other questions That is the is that the container? Yes, I think it's a great idea. I think it's absolutely a great idea We experimented with it about a year ago before the project was created. We ran into some challenges in the way Some of the open stack projects like neutron and cinder need to talk to the kernels So that there are some things to figure out there But I think that process certainly plays a or that concept certainly plays a material role in making upgrades more easier and even even making installation and distribution of open stack easier. It's a great great initiative Anybody else? Awesome. Well, thank you all for spending some time with me this morning. Look forward to seeing many of you around this conference this week and Thank you very much