 So, we are, we're well into our keynote time this morning. We have one main session left that is really a power-packed panel from our headline sponsors. And you know that we are all about the users here. So this panel is going to be moderated by a user who has spoken at previous summits, and it's going to be really, I think, getting some good information out of these panelists that is relevant to all of us who are interested in how we take the most advantage of OpenStack and how we work with partners and the community overall to do that. So I am really excited to welcome our moderator from Comcast, Mark Mule. Thank you, Jonathan. It's great to be here. How's everybody doing? Good. All right, we're on the tail end of our keynote session today, so let's not keep you all waiting. Although I was tempted to take a quick jaunt off in that BMW, I got to go talk to Stefan about that a little bit later. That's a pretty sweet car. We got to look at it backstage. You definitely want to take a few minutes and check this puppy out. So let me introduce some of our panelists today. Our first person hails from Dublin, Ireland. He's got a huge, well-known passion for open source. He's contributed to a variety of projects over the years, and he's well known to our OpenStack community. He's been the PTL for Oslo. He's been a member of the TypeMegal Committee and an individual director on the board. He's welcome Mark McLaughlin, the OpenStack technical director at Red Hat. Welcome. Okay. Our second panelist has got a tremendous wealth of experience. He started out at SGI doing interactive television, which is near and dear to my heart, kind of before it's time. He was working at startups before startups were cool. He was employee number five at Vitria Technology. He was a founding member of the VMware team that did the vCenter product. Everybody's probably familiar with that. Today, he's the CTO of cloud computing at Huawei. Please welcome Haiying Wang. Next, we have a person who's held a variety of technology R&D roles, big thinker. He led Ericsson's strategy shift to the cloud, including selecting OpenStack as Ericsson's cloud platform. And he's done a number of open source projects at Ericsson as well. Today, he's in the CTO office as the vice president of networking and implementation architecture. Let's welcome Matt Carlson, please. Okay, and our final panelist joined Intel in 1991, and she's held a variety of roles there. Started out in the systems engineering group, did some pretty cool stuff with factory automation apps. Maybe we'll see if we can explore that a little bit on the panel, or you can talk to her afterwards. And now she's back in a systems engineering role, where she's responsible for building, deploying, and running their hybrid cloud for Intel IT. Please welcome Intel IT's product owner for hybrid cloud, Ruchi Bargave. There we go. Welcome. All right, gang, here we go. So I thought we'd start with a pretty reasonable introductory question. We won't get to the hard stuff until later. But why is OpenStack important to you and your companies? Mark, maybe you can start us off. Yeah, sure. Well, I guess from Red Hat's perspective, fundamentally, we're trying to be a full spectrum provider of solutions for the cloud data center. And obviously, OpenStack is a key technology part of that. But I think what really excites us about OpenStack is just this amazing, diverse, vibrant open source project we've built here. And we're a pure play open source company, and this is the exact environment that we know great stuff is going to happen here, and we know how to bring that value to our customers through our products and our partner ecosystem. For me, at a very personal level, I mean, like I've been working on OpenStack for our open source for something like 15 years, and I guess I've set my mind to try to figure out how to prove that OpenStack is the ideal way for cross-industry collaboration on technology. And I think OpenStack is just the really ideal example of that happening. So I'm really... Realization. Yeah. Hanying, how about you? Oh, yeah. I think Huawei is a big technical company. We're benefitting from the standard. I think for new technology to get adoption open and then standard system is crucial. So I think OpenStack, playing that rule, from our perspective, for most, we want to make OpenStack to be successful, to become a standard for more wide adoptions. So that's why we have a dedicated team contributing on OpenStack. Great. Thanks. Matt, would you like to take a shot? I mean, continue on there. I think it's really important to have the open API that you have something that can drive industry alignment, the open source models at such driving innovation and speed. I think a third aspect that we really see important is that OpenStack is actually the only one that provides a true multivendor orchestration engine, to say, because you can actually add whatever third-party software or whatever third-party services or whatever third-party equipment that you want. And I think that is pretty unique. It's a pluggable architecture. I think that has been important for us. It's fantastic coming from a vendor that might want to sell every component or might want to be able to play in every component, allowing us to plug in different pieces. As a user, that seems like it's pretty, that's one of my goals anyway. Ruchi, how about you? Why are you excited about OpenStack? So Intel, as you saw in the video before in the presentation, that is driving the vision of a software-defined infrastructure and redefining the data center. And with that, it's very important to have a single control plane as the orchestration layer and nothing better than OpenStack to drive that. So I think that's where Intel is really vested in this. Awesome. That sounds like a great vision, but you and I are users. Let's talk about some of the pain points that we feel at Comcast and at Intel IT around OpenStack and what we need to do as a community to make it better. What are the pain points? I'll talk about the pain points and, like, I see three key pain points. And I think a lot of folks here from the enterprise world will relate to that. One is to have rolling upgrades. You don't want to upgrade from how many of you are probably still on Havana and maybe some on Grizzly. You want to do upgrades without bringing downtime to the tenants. So that's one. And then enterprise shops have a lot of investments of infrastructure. So how do we integrate the legacy existing infrastructure investments into OpenStack? That's another very important area which is a big pain point for me. And the third part is, as a hybrid cloud product owner, to do a federated identity authentication with a public cloud. That's, again, a big pain point. We are done in POCs, but still something to grow on. Awesome. Stability and scalability, key themes, I think for probably both of our companies. There has been this emphasis on features. So OpenStack has been running forward, forward, forward, adding new features as we go. I wonder if we could solicit some thoughts from some of the supplier side of our panel around what we can do to increase stability and scalability around the OpenStack infrastructure. I can maybe start. Now I think talking to a number of operators, a lot of them are kind of stuck on Havana because they can't do an upgrade. So I think just getting upgrades, getting automatic backup and restore and all these things into place, I think it's a key because when we are starting deploying these telecom workloads, this is services that requires five nines of availability and including planned downtime. We need to have, for instance, rolling upgrades in place because we can't afford to take down the system. How about, let's talk about installation and deployments. I think, Hanying, you had some thoughts that you wanted to share in that when we were prepping. Yes, I think as OpenStack getting mature, we still have lots to do. People are starting to adopt. I think it's a challenge for them that this kind of software-defined infrastructure will move the function from hardware to software, making it more dynamic, but at the same time make it very hard for users to adopt. So we see more of the effort on the standing up of the OpenStack, but that's just the first step. If the software-defined infrastructure without run-operatedly correctly and updated correctly, it will become obsolete or often hardware very quickly. So the installation and running the infrastructure and keep them updated as a software itself is a moving target. So I think the next way is the tools to help people to adopt OpenStack, to run OpenStack, to keep them updated as the software itself moves forward. Mark, what have you been hearing from customers? Sure, I mean, many of the pain points we hear from customers are similar to what Rushi talked about there in terms of upgradability and scalability and performance, operability, installation, deployment, ongoing management. It's all key parts of what's needed here. And I think we are in the early days of OpenStack, and pain points like this is not something we should be afraid of talking about. It's just part we're all here together in the sausage factory all publicly and we're all keen to make it work. You talked a little bit there about features versus working on the core aspects that we talked about there. And you know, I celebrate both, right? The community is growing here. It's a diverse set of interests. Some people want to work on kind of expanding the scope of OpenStack and adding new features and competing with some of the other cloud providers out there, but there's still a very strong core group of people on the project wanting to improve the core. Yeah, let's talk a little bit more about some of the stuff around the core. So it's not uncommon when you build software that you sort of get the core features right as you were saying, and then you start to sort of flesh it out. As an operator, one of the things that stresses me out a little bit is a lot of the tooling around the infrastructure is later to market, so to speak, running it in an automated way or sort of automating this piece of the puzzle versus just kind of doing things manually. The monitoring tools, the accounting tools, things like, you know, salameter and so on. Any thoughts about what the next sort of wave of priorities need to be and how far along we are getting the supporting infrastructure working? Somebody's got to jump in. We can't stay here for 20 minutes. Yeah, I think I just followed what I just said. I think the tools that help people adopt OpenStack is crucial because the traditional IT tool is you know using this environment. So I think I already see the effort, lots of startup and lots of an open source project in working on this area to address the pen pointer for the end users. So you can around the systems, you can stand up and keep it updated. You have a punch list, do you want everybody? We've got like half the development community here. Yeah, well I say many open source projects and the triple and others and Huawei have one called Compass addressing the standing up, operate and I think we need more feedback from the user to drive the adoptions. So you know, we at Intel have enterprise use case and then we also have I guess in the industry the HPC use case where you could have 10,000 servers and you want to deploy code and on all of them within minutes. Today with Python, you know, I think it's going to be a challenge because it's interpreted and so from what I have heard by some of my engineers that Facebook does, I don't remember what the technology is. Basically it's compiled Python kind of stuff. So maybe that's something which our community could go address. So there are high scale use cases that even though this is a highly scalable architecture, we don't quite address them well. 10,000 VMs at one time seems like a big task to me. We don't have that kind of use case at Comcast. But in the HPC world, not only at Intel, but across the globe. We should get the CERN guys back out here, ask them if they've got any tricks up their sleeves. Good, let's talk a little bit about NFV. You know, there are always these, you know, Jonathan put up the buzzword bingo sheet. I'm not sure if NFV was actually on the list there, but it is one of the things that seems to be on this parallel track to all of our work here in OpenStack. And as a user, you know, I kind of wonder, are these things destined to be parallel forever or do we need to bring them together? Go ahead, Matt. Well, I think NFV is actually, as I said, about virtualizing applications and a network within the operators network domain. And I think it's probably one of the biggest transformation for years now, because we believe that within a couple of years, maybe more than 50% of the applications on the network will be virtualized. When I look on the demands, I think a lot of the demands are really the same as the demands that we will see for enterprises or high-end enterprises, that is audit, logging, troubleshooting, all this. Maybe the only thing that sticks out is maybe that we need to orchestrate the wide area network. We need to have line rates to the virtual machines, et cetera. But I think actually, I think NFV, to a large degree, will most likely drive the same type of requirements that we will see for enterprises. So I have a big hope that the kind of world can make combined forces in making this happen. Yeah, I mean, for me personally, I wrote a blog post on kind of my NFV journey there recently, and it started for me in Dublin about a year ago when I visited a local partner company there, and they mentioned NFV, and I've never heard of it. So they explained it in terms of the application requirements for these really highly specialized applications in NFV, and my instinct was, no, OpenStack, that's not, you know, OpenStack's never gonna satisfy that. But I think I've learned over the year that this isn't just taking these existing applications and trying to make them work in a cloud. It's actually the telco market embracing the notion of an elastic cloud, but the applications do have some very specific requirements, say, around deterministic performance and networking requirements. And actually, what we're seeing evolving now is really interesting ways that they can fit in behind the OpenStack abstraction and make total sense for OpenStack to support. And so I think it's really exciting to see a huge market like the telco market about to be transformed by OpenStack and OpenSource. You know, as a buyer of a lot of those services, one of the questions that I kind of wrestle with a little bit is, there are some companies that have well-established big iron, you know, routing infrastructure, switching, technology, and so on. And they start from a different point than sort of the upstart startup who is building new software from scratch. I wonder if you guys have any thoughts about sort of approaching NFV from the point of view of a well-established network infrastructure company versus starting from a startup and being able to build software that's meant to be virtualized, meant to run sort of in a distributed fashion. Hi, Ying. So yeah, I think you mentioned whether they're parallel. I think when they started two years ago, it does seem parallel because OpenStack is more a bottom-up approach and engineer-driven and modular and NFV is more top-down but I think now they're getting closer. OpenStack society more pushing, more try to meet in the middle, provide functions, they have a subcommittee for NFV in OpenStack. At the same time, NFV society is also more realistic to choose not specific platform, but the current platform, but using upstream contribution to figure out what they need. So in the between, I think the challenge here is that when you move the functions from box to software, it gives you dynamics, it takes advantage of electric, like scale-up and scale-down and the commodity hardware, but at the same time, you lose of control for deterministic computing and SRA. So I think of this area that are being pushed. For startup, they may just scratch from a new application fitting the clock platform, but for exception one, they have all they have apps, you try to fit in the current new infrastructure. So this is a challenge that a little bit different, but I think they're getting closer. I think it's very promising that maybe next year they will see them more working closely together. Matt, you have well-established, vertically integrated systems at Ericsson that you guys have been selling to a variety of customers for many years. How do you guys think about NFV and the move to software? Are you starting from sort of your foundations and what you've got, or in some cases are you starting fresh? I mean, we used to, as I said, we used to, we are still delivering boxes with like these five nines of availability, as we say, on the service level. I mean, everyone expects that the phone works, whatever you are, traveling the world, that you should pick it up and it works. And of course, the characteristics will not be less when we're going into cloud. So we actually see this as a big opportunity to really kind of speed up service, creating new services, speed up deployment and rollout of new services because if we don't have to ship in the box, we just have to kind of put the application there. I mean, I think there is a lot of benefits going this journey in terms of speeding and creating new service, speeding deployment, which can't be done with a kind of box-centric approach that has been the history of telecom. What does a community like the OpenStack community need to do to make it easier for a company like Ericsson or Huawei to embrace the model that you just described, where we have this rapid iteration happening in the OpenStack community, but we're trying to build very stable, very reliable lifeline services, in many cases, on top of that infrastructure. What does the community need to do to make that easy and palatable for a company like Ericsson or Huawei? I think in OpenStack, I will just see the sign. It's a more coordinated effort because before each module is sort of independent, so the telecom usually is a whole spectrum standard. The application at OpenStack right now focus more on private enterprise cloud, but the telecom is sort of between enterprise and consumers, big apps, so cross data center, coordinations. Those are the features. If Neutron, Nova, all this group can work together, and then at the same time, MPN3 or those guys, can more collect, so we can see this can be better fit quickly. So I think it's getting closer, it's got there, but if you're asking what's got small, I think we want OpenStack to be stand up to more coordinated focus on some of the realistic requirement from NFV side. Maybe I can add on to create this OpenNFV, which is a kind of collection of operators and vendors in creating an NFV. Actually, a working reference platform, which I think is a good step, so to say, that you actually, the target is to have a working reference platform. But I think, going into what Haying said, what I actually think is that we can actually try to bridge the requirement, the things that needs to be done in asynchronous ways over the life cycle, so roadmaps over maybe a number of OpenStack releases. I think that we can help out in groups like OpenNFV, to help out the community and the foundation to have more, to drive the NFV requirements in a bit more structured way instead of per product. So let's talk a little bit about that tipping point if we can, and maybe Mark, you can lead us off on this. Have we hit a point where the way we do this development and the way we sort of manage the product, so to speak? Have we hit a tipping point where stability and scalability issues need to be addressed? And maybe bumped up in priority relative to the feature, feature, feature priority that we've been marching down? Yeah, I mean, in terms of those cases, I think what the developer community, which I guess where I've come from, what they're starving from is real kind of concrete feedback from operators and users, and so we need to figure out how to build a closer feedback loops between the developer community and the operator and user community, and kind of similar to your previous question there, it's not so much what should the community do for users or operators or what should the user or the community do for NFV vendors, but more how can the operators and vendors feel like they're more part of the community and contribute in different ways to the project other than just code? So typically in open source projects, it's very difficult for anyone to figure out how to contribute something other than code, and I think in the case of NFV and in the case of the Windows Enterprise Working Group, we're finding ways that product managers and people with kind of an understanding of the requirements can come together, collaborate and really provide very useful and concrete feedback and input to the developer community to feed off, and so that's a really exciting development, I think, within the project right now. Pretty good. So, you know, something which Jonathan said in his keynote really resonated with me and what Mark, you talked about is the software defined economy, you know, and that's really going to drive this because in order to do things really fast, fast and fast, you want to take the IT group out of the picture, and that's where NFV would, you know, where you have the proprietary stacks of hardware and you can, you know, do NF, orchestrate it with the software layer, that would be really cool. Cool. I should mention that we are taking some questions from Twitter. I don't know if we're able to get something up on the screen, but if you tweet with the hashtag OpenStackPanel, there are some folks in the back that'll pop up some questions onto the down-screen monitors, down-stage monitors for us. So let's talk a little bit about the Intel IT Cloud, if you don't mind. How do you decide between splitting a workload or a project between your private instances and some public instance? Okay, so we don't have a production hybrid cloud going, but we do have public cloud usage as well as private cloud usage. Initially, it was the power of the credit card, as we heard earlier, so people just went because IT was bureaucratic, we have our own stringent requirements, so they just went for ease of use, but we have put our collective thoughts together and put systems in place so that they only go to public cloud because we've already got a huge amount of investments in our data centers, and we don't want to waste those. So as long as we offer the capability and we have the location available, we prefer that they stay in-house. When we don't have the capability, yes, public cloud is the option, and what we are trying to get with the hybrid cloud solution is, it's become seamless to the customer, to our end-users, that they get what they need irrespective of where it's residing. What kind of conversations do the rest of you have with your customers as you talk to them about private versus public cloud options? Yeah, I mean, I guess at Red Hat, we've been talking for years about the open hybrid cloud and kind of our vision for that there, and I think there's huge opportunities there, and I'm particularly excited around OpenStack, the opportunities around an open hybrid cloud that's actually built on the same kind of underlying technologies, so we can expect plenty of OpenStack-based public clouds and plenty of private clouds, and I think that gives a really strong basis for making this hybrid cloud use case real and really kind of experimenting with all the different ways that you can use that, whether it's spillover into public cloud, whether it's kind of doing your dev test in the public cloud, but running production in a private cloud, whether it's, if you've got kind of large-scale Hadoop clusters you want to run for a short period of time, you've the choice of whether to run that in-house or on the public cloud. Yeah, I think it is obvious that cloud will be the big part of a future infrastructure, but I think at the same time, enterprise always exists, so public or private cloud will be always exist at the same time. The bridge there is nature, so on demand in those futures, hyper-grad can realize for the enterprise, and at the same time for the governance and security, the private cloud can satisfy the company need. So I think OpenStack bring a great opportunity for that, and they try to mirror the success of public cloud in the form that most enterprise can consume. So like Amazon, you can compatible with OpenStack, and now some people even do Microsoft and Google. So this is inevitable, and there's lots of challenges because both sides are moving target. So synchronize them and it will be challenging. I think Intel doing good job, we have some customers, they also trying to bridge them all, but they basically use OpenStack as a general platform to API to connect them all together. So at least for management perspective, operation perspective, it's a one interface for both sides. Although behind the scenes, they are still a bit different as time goes, and you actually will be more consistent. So I think a private cloud will be the future form. Great. So let's jump in maybe into our last topic, we've got about eight or nine minutes left. Let's talk about talent. I'm a big user of OpenStack, I need to hire people in order to help me figure out how to build it, what to build, where to build it, how to operate it, keep up with all the releases and so on. There was a great happy hour last night for the women of OpenStack. Wow. I think you got the opportunity to be there, you wanna tell us a little bit about it? Sure. So there was the two events being held at the summit. One was a happy hour for women of OpenStack, hosted by IBM. I think a great event. I believe they've been doing it since the San Diego summit and hopefully they will continue doing it or somebody else will. And it brings to the point of bringing diversity to not only our developers, to our users, to all aspects of the community. And it's true for any business, the more diversity you have, the better it is. And from a talent perspective, as we all know, 50% men and women graduate out of college. So if we don't target that segment of the student graduates, we are gonna be missing the boat. So I think as a community, we really need to focus on that. However, if you wanna talk about general talent developer. Well, let me just comment on something that you made me think of here. So isn't it interesting that one of the strengths of open source in general is the diversity of all the use cases, the developers, the perspectives that come to an open source environment. And here we are in the OpenStack community really trying to drive diversity into the development community. And there is, I think one, it's no secret that there's sort of one area that we always seem to fall a bit short in. And that's having women inside the technology organizations in general and specifically inside OpenStack. I think it is fantastic and I applaud IBM for hosting the Women of OpenStack event last night. So why don't we make it a little bit more generic and talk a little bit about generic talent for OpenStack. And that's again, I'm sure all of you from whichever group you are in, whichever company you are and you find it extremely hard to find developers, deployers of OpenStack. And what we at least at Intel have started doing is tapping into the talent early. Starting from high school, we plan to host hackathons where we'll get high school kids to go fix bugs, try to teach them. And we heard about the Linux student there. I think we are trying to get... Well, to get under seventh grade somehow. That's the new bar. Yeah, well, that's something new which I learned. So that's maybe going there is next. And then globally wherever we have teams, development teams at Intel, we have partnered with the universities. And in Poland, we've got a grant to go with them and work with the university to develop OpenStack talent. In Guadalajara, we work with the university and besides the US universities. So you're really thinking of it as a pipeline that you have to start very early in the talent selection or talent development process and you're starting all the way back in high school. Absolutely. And if we go back to the Linux days, I think we have to learn a lot from them and we went through the same path. We've got a huge community there and OpenStack needs to just follow that path in my mind. So once we get this great pipeline going, we've got people that are sort of flowing out the end of this river into the OpenStack community. What do we need to do to mentor them as developers in the OpenStack community? Yeah, I mean, one example of what's going on is just over this weekend, the OpenStack Foundation organized upstream training session, I think, and they had over 60 new developers join that and the foundation and the volunteers, they're mentored these new developers and showed them the processes. I mean, I read that we're, I guess, used to trying to grow new talent and find people who have a real passion for open source and, you know, really understand the collaborative environment that's needed here and it's really in exact science and I've yet to understand how to predict whether a really strong developer is gonna be successful in an environment like this. So I think it's really for, you know, anyone getting involved with the project, it's just a question of giving it a shot. Don't be shy, just step up, get to know people, see where you can kind of find your niche and contribute and see what happens. It's really good fun if you can kind of figure out how to play in this environment. Now, and I think it's also give new designers a lot of spare time and just tell them that you can sit down and review OpenStack code for participate in reviews because I think that is often the best way in and just give them the time so they get acquainted to the process, et cetera. Give them the opportunity. Yeah, and I mean, you have to spend a couple of months, I'd say, to train in and that is probably the investment you have to take in a lot of cases. Well, maybe that's one of the challenges. There's so much to learn that it does, there is a long spin up time. And one of the things we measure at Comcast is how long does it take a new developer to sort of get productive. And it might be that it's long for a new developer to get productive on a piece of OpenStack. What other challenges have you guys encountered as you've tried to bring development resources into the OpenStack ecosystem? So I think Huawei's OpenStack journey is very companion. And then we, OpenStack really is an international movement. So you need to get all people to get involved. But we do have unique challenges like many other people company have. We have language challenges, it's English predominant. So if you're not English, you can have some challenges over there. And there's a culture clash because it's more open discussion and for Asia countries, it's less comfortable for the style. But as we goes, and I think community is very welcome. So as we goes, we overcome all those. And I remember the first team member checking the code from China would take a couple of months because there's lots of things that it goes through. But after you build a confidence, you have a core team, you build on the success, you give people have a choice for freedom of choice so they can do what they feel comfortable. It will come out now. So as we are and we've seen in two years, we're contributing dramatically increase. I think that's a tale story. I think I hope more people in the world really make this global movement to make this stand more successful quickly. Excellent. All right, well, with that, I'll just say thank you very much for participating in the panel. Can we give our panelists a round of applause, please? I think we're gonna go back to Jonathan. How was that? Do we do okay? Yes, that was great. Thank you guys very much for joining us. Thanks for having us. Thank you. One of the things that I actually really appreciated about that panel is all of these people are very involved in OpenStack. Now these are people that are working in the community and building products and working with customers every day. And that's great to see all of the different perspectives that we can get here at the OpenStack Summit.