 Live from Boston, Massachusetts, it's theCUBE, covering OpenStack Summit 2017, brought to you by the OpenStack Foundation, Red Hat, and additional ecosystem support. Welcome back, I'm Stu Miniman, joined by my co-host this week, John Troyer. How come you're welcome to the program? A first-time guest, professor at Boston University and lead of the Massachusetts Open Cloud, Oron Krieger, thanks so much for joining us. Oh, my pleasure, thank you. All right, so we're here in Boston, the center of culture, the revolution, a lot of universities. Tell us a little bit about, just quick on yourself, your role at BU, and then we'll get into the MOC stuff in a little bit too. Sure, I mean, I sort of came back from industry after 15 years in the industry to this incredible opportunity we had to create this entity. I mean, there's no other place like this. If you take the universities in the city, it's equivalent to all the universities on the Pacific West Coast, right? The concentration of high tech is unbelievable here. I wonder, it reminds me, my wife was actually involved in partners healthcare. First got launched here in Boston, was it early technology and collaboration here in Boston? Sounds similar, what you were doing with some of the universities in cloud. Maybe you talked, you came from the vendor side, just real quick, your background, you worked at a company that John and I know quite well, maybe just give a quick background? Sure, I left academia, I don't know, how many years ago, ended up going to IBM Research and was there for about 10 years, and then I joined this little startup called VMware and started up and then worked as sort of one of the lead architects for a VCloud director and the whole VCloud initiative. All right, great, let's speak. Today you're also the lead of Massachusetts Open Cloud. We actually had a couple of guests on from Red Hat that talked a little bit about it, but tell us about the project, the scope of it, how many people involved, how many users you reach with this. Sure, so what, the future's in the cloud. I mean, you look at sort of the fact that users can use what they need when they need it, producers can get massive economies of scale, the future of computing is in the cloud. And when I was on the industry side, what really concerned me, what was going on is that these clouds were really closed. You couldn't see what was going on inside them. Innovation was sort of gated by the single provider that operated and controlled each of these clouds. So the question that was struggling with back then is how can we create a cloud that's open, that multiple technology companies can participate and certainly when I came back to academia, a cloud where I could do innovation in, where not just me, but many, many different researchers. You look at how much research has fundamentally impacted our field, it's dramatic. Even in just sort of the very area we're talking about from what Mendel and Team did with VMware and then Zen coming out of Cambridge and I mean, Seth coming out of, just like technology after technologies come out of academia, but now clouds are these closed boxes you can't get into. So we had this incredible opportunity. There'd been this data center, the Massachusetts Green Hype Forms Computing data center, MGH PCC, 15 megawatts. That's more than half the size of one of Google's 16 data centers that had been built right next to Hydro Dam, one third the power costs of what it is in Boston by five big institutions, MIT, Harvard, BU, Northeastern, UMass. And we thought, wow, couldn't we create a cloud there? Couldn't we create cloud with some 157,000 potential students as well as the broader ecosystem. So we started discussing the idea, all the universities kind of signed up behind it. The model of the cloud is not to create another single provider cloud, it's not going to be my cloud. The idea is to have many vendors participate, stand up different services and create an open cloud where there's not just multiple tenants but there's also multiple landlords of the cloud. Great, yeah, could you talk to us a little bit about how does some of those pieces get chosen? How does OpenStack fit into it? And if you can talk about some of the, some of the underlying pieces, it'd be good to understand how you sort that out too. Sure, so in doing that, it's actually being sort of this cool, you know, you have to kind of build different levels simultaneously. When we started the project, you know, our first thing was, oh, you know, we'll be able to just stand up the cloud. Wasn't that easy. OpenStack is actually a complicated learning curve to get up. Now it's matured tremendously, you know, we've been in production for about 10 months with no significant failures. I'm almost thinking that we need to kind of bring it down for a couple of hours just so that people start realizing this is not intended to be a place where you run it like you would a production data center facility. You know, that we don't guarantee in SLA because people are starting to assume we do. But we started off and we sort of sold OpenStack, got it up and running, took us a while to get it to the production layer. Started hosting courses and users and stuff like that. And some taste that with sort of two other tracks. One is developing some of the base technologies to enable a cloud to be multi-vendor. So Mix and Match Federations serve our core of that, which is this new capability that we've, after like five iterations on the right way to do this to allow multiple different clouds with their own keystone, because different administrators say from MIT or Harvard or from companies that might want to participate in startup and service. So to have a capability of feathering between those things and allowing you to, for example, use storage for one and compute from another. We started off with OpenStack because OpenStack already had the right architecture. It was designed as a series of different services, each one which could be scaled independently, each one that had its own well-defined API. And it seemed natural, geez, we should be able to compose them together. Have, you know, one stand up, you know, Nova Compute, another one stand up Swift Storage, another one stand up Cinder Storage. Turned out not to be that easy. There was assumptions that all of these services were stood up by the same administrative entity. After three iterations of trying to figure out with the community how to make it, we finally have a capability of doing that now, that we're putting into production in the MOC itself. You talked about the different projects inside OpenStack, that's been one of the discussions here this week at the summit, different projects, the core, which ones are important, and also the whole ecosystem of other cloud native and open source projects that have grown up around OpenStack over the last six or seven years. Any commentary on which kind of projects you're finding are the most useful and do you see as kind of the core of OpenStack going on, and also which projects from other ecosystems do you think are natural fits into working on an OpenStack-based cloud? So in our environment, we've stood up sort of all the core services you think of, obviously Nova and Cinder and Swift, and we're using SAF in most of our environments. Sahara, Heads. We've actually expanded beyond in a couple of different dimensions. I guess that one thing is we've been using extensively SAF that's been very valuable for us, and we've also been modifying it actually substantially. It's actually kind of exciting because we have graduate students that are making changes, they're now going upstream in the SAF community as a result of their experiences in doing things within our environment. But there's other projects that sort of tied in at sort of two different levels. One is we're working very closely with Red Hat today around OpenShift, and we're making the first deployment of that available in the very near future. And the other thing is that it's very important for our environment, we have, I think, three different talks related to this, to have data sets in the cloud, to have data sets that's shared between communities of people, data sets that are discoverable, data sets where you can actually, that are citable. So we've been working very closely with Harvard and the open source dataverse community, and we've together created the cloud dataverse, which is now actually in the MOC. So researchers from all these institutions can actually publish their data sets, as well as researchers from around the world. So there's over 15,000 data sets today in the Harvard dataverse, for example. Curious if you can give us any commentary on how open source fits into education these days. Talk about the pipeline of the next generation of workers. Do your students, you talked about upstream contributions, how do they get involved, how early are they getting involved, you know? Well, actually, that's sort of a bit of a passion of mine. And it's also, so multiple different levels, I guess. One of them, I think, is this is a great way for a student to sort of get exposed to a broad community of people to interact with. I think it's, you know, rather than going into sort of one company and getting locked down doing one thing, I think it's just enormously valuable. What we've been, there's sort of two different dimensions, I guess, educationally and from a research perspective. And both of them were very tied to open source. So from an education perspective, we have a course, for example. I mean, one of my frustrations, having come back from industry, was students had done a lot of great, learned how to program, a lot of these individuals, and, you know, they really didn't learn how to do agile, they didn't learn how to work with teams of people. So we have a large course that's sort of by multiple institutions today that's sort of tied to the MOC, where we actually have industry, we teach them agile methods, we teach them a lot of the sort of fundamentals of cloud, but we also have industry mentors come in and mentor teams of five students to create a product. There's actually three different lightning talks by different students that have taken this course that are here in the OpenStack forum today. So it's kind of exciting to see. So we've had several hundred students that have learned that, and at least in my experience, learning how to deal with open source communities, mentorship is a great way of doing that. You know, first year we started teaching this course, we had sort of struggled finding mentors. Now we're about twice as many mentors applying to mentor teams as we can accommodate in it. So that's been kind of exciting. That's great, that's super important in learning, right? And not just learning how to program, but how to operate as an engineer and a team. Well, so in the MOC itself, a lot of it stood up by students. We have like 20 to 30 students, we have a very small core development and operations team, and most of it is actually students doing all the real work. It's been amazing how much they can accomplish that environment. You mentioned OpenShift. So another conversation that's been somewhat confusing in the broader industry is talking about containers, versus VMs and virtualization and OpenStack. Here this week, I thought it's been a fairly clear message that you can be containerizing the stack itself, and then there's also a role for containers on top. Obviously been involved in virtualization for a long time. How are you seeing the evolution of both containerization as a technology, but also container-based platforms versus kind of the infrastructure and provisioning of the cloud part? I mean, there's three levels that all have its role. There's actually people that want to control all the way down to the operating system and want to do customized things. They want to use SRILV and want to use accelerators that haven't... And so there's people that actually want hardware as a service, and we provide a capability for doing that that's got its limitations today. There's people that want to use virtual machines and there's people that actually want to use containers and the ability to orchestrate, setting up a complex, multi-tiered environment on that and doing fine-grained sharing in a containerized environment is huge. I think that actually all three are going to have a continued role going forward. And certainly containerized approaches an awesome way to deploy a cloud environment and scale the cloud environment, even the IIS environment. So we're certainly doing that. Love the idea of all the collaboration you have both internally with all the universities. You're getting reached out by outside of Massachusetts. How do you interact with the broader community and share ideas back and forth? So there's of course multiple streams of that. One of them is our industry partners are very broad. Second, we've participated in sort of the open stack summits and all those kinds of things. The other thing is that the model that we're doing I think has a lot of excitement and interest from very many different segments. I mean, I don't think people want to see the public cloud be dominated or can always be dominated by a very small number of vendors. So the idea of actually creating an open model of cloud. Lots of other academic institutions have talked with us both about setting up sister organizations feathering between clouds and replicating the model. We're still at an early stage. This model still has to be proven out. We're excited that we have users that are using us now to get their work done rather than just courses and things like that. But it's still at a very early stage. So I think as we scale up, we'll start looking at replicating that model more broadly. Is there any public information about what you're doing? And I'm curious, will this tie into like MOOC delivery, things like that? Oh, absolutely, yeah. So I mean, it's all on our webpage info.massopencloud.org. And we, so everything's done in kind of the open, I guess. So all the projects, they're all, everything's on the websites and you can sort of discover all about it. And we welcome participation from a broad community and are excited about that. All right, Aron Krieger, really appreciate you sharing with our community, everything there. Congratulations, local. We'd love to stop by some time and to check out even more. John and I will be back with lots more coverage here from Open Sack Summit 2017, Boston, Massachusetts. You're watching theCUBE.