 Live from Vancouver, Canada, it's theCUBE. Covering OpenStack Summit North America 2018, brought to you by Red Hat, the OpenStack Foundation, and its ecosystem partners. Welcome back to theCUBE's coverage of OpenStack Summit 2018 here in Vancouver. I'm Stu Miniman with my co-host, John Troyer. Have a welcome back to the program. One of the rock stars in the networking world, really. Not just because he's given it hard, but I've been lucky to get to know JR over the years. We worked together on some projects back in the day when I was on the vendor side, and one of the people I love talking to in the networking world. You're the founder and CTO of Cumulus Networks. Great to see you. I am, great to see you, Stu. Good to see you again, John. Thanks for having me on the show. Yeah, absolutely. So JR, gosh, we get together, we talk about scalable architectures and what's happening in the network world. We're here at OpenStack. And hoodies. If you want to give your feedback on hoodies, yeah, we can totally go there too. But yeah, let's start. We're here at OpenStack. How's OpenStack fit into the overall picture for you and Cumulus these days? It's really interesting and good. I got to admit, when we first started out, there were a whole bunch of people at the company that were insistent that OpenStack was a science project and nothing was ever going to happen with it. I was kind of lucky to say that I was always on the fence, but there were a lot of people that were, they kind of had that perspective around it. As you know, the industry had this perspective that OpenStack was going to supplant VMware and VMware was going to die and shrivel up like a little raisin. And luckily, both of those things were wrong, right? VMware hasn't shriveled up like a raisin and OpenStack serves a very different purpose in the world than something like VMware does, but OpenStack is incredibly relevant. We got about 20% of our customers are building OpenStack clouds for either public cloud offerings to other people or for private cloud usage themselves. It's great technology, it's moved along quite a ways. I think it's still got a bit to go, but it's got legs for sure. Networking in general has been going through a lot of changes. I was actually presently surprised. I went to Cisco Live Barcelona earlier this year and they're positioning themselves in the future as a software company. Imagine that. Which for those of us that spent our careers in the networking world, it's like, wait, it's boxes and ports and billions of dollars worth of cables and optics and things like that. Yeah, exactly. So JR, you were quite early in the software movement and networking, so bring us up to speed as to where we are with that, what you're seeing in the marketplace and from your customers. Yeah, so I can't lean back into my chair because there's a bunch of arrows in my back because I was that crazy dude standing on the corner saying this was going to happen and as you're pointing out, Cisco's saying it's going to happen. We run into other companies' names withheld for the sake of being cool who are now going to their customers saying hey, we'll run our software on Edgecore hardware or other hardware platforms. So the industry is embracing this not just from a technology perspective, but customers are asking for it because it makes sense for them operationally. So in fairness, I'd say the war is on for who's going to give customers the best kind of software experience in a disaggregated networking world. Yeah, I know John wants to get in, but one more question for you here is every time we come up with new technology, it seems like oh, we're going to simplify the world and make it really easy. Well, virtualization, we spent a decade trying to fix networking and storage in there. You worked on a bunch of projects helping in the virtualization phase. Containers, once again, oh, we'll extract away my application. Oh, I need to do Docker networking, I need to do all these things. So serverless, once again, we're going to do this. Networking is one of those core things, especially when we talk about multi-cloud. It's right in the middle of everything. Why isn't it simple yet? Yeah, that's an interestingly fair question. How to say it right? It can be way simpler than people want it to be. I think the reason it's not simple right now is most people, as they start to deploy, are stuck in older paradigms. For instance, I see your cell phone sitting there. When you turn that thing on, you connect to probably 15 or 20 different services. All of your connections to those services are all endpoint authenticated. All your security is not, you have no idea what's in the middle of the network. I mean, you could be going through the worst malware storm ever, and it just doesn't matter because everything's encrypted and authenticated end to end, so you have a really secure connection. That's not the way people think about developing applications. That's not the way classical enterprise applications have been developed, enterprise tool sets, that sort of thing. Eventually, what you're going to find is people realize that, hey, if I take care of this stuff on the endpoints, the middle doesn't matter, and so networking in the grand scheme of things matters less. We're seeing customers starting to do that now where even in the context of an OpenStack cloud or a containers deployment, they may have a traditional enterprise network architecture. Mark Andreessen had this great term, which I'm going to hopefully don't get in trouble here. It's called Wall Street IT. His observation, which I believe is completely fair, is that Wall Street came up with this architecture that included VMware and Cisco and EMC and like HPE on the server side, and they proved it out, tested it, made it, vetted it, and then other people embraced it and just kind of copy and pasted it and laid it down in spades. Google, Amazon did another thing they called, Mark called WebScale IT, and it's a very different way of thinking about how your infrastructure comes together. Well, if you're a new company and you're starting from scratch, cool, you can do the hoody way, like the skateboard kid model over here on the WebScale IT, but a lot of enterprises need to get that level of efficiency, but it's going to take them a while, right? So they have that enterprise stuff, when they stand up new things, they can think about it in these new WebScale principles, and as they start doing that, we're seeing the networking gets easier and easier over here because they're solving their security problems and isolation problems in the end. Nice. So JR, one of the things that we've been seeing here this week at the summit is real customers, real deployments, right? Productive out in the real world. Maybe if you could take us through a customer story that with OpenStack and with Cumulus Linux and how it all fits in and kind of what the benefit is and what they're doing with it. Awesome. Interestingly enough, I don't know if you guys talked to them, but there's a guy named Roland Cabana that's here today. You did talk with Roland. He's a customer and they just did a press release. The reason I'm bringing him up is we have tons of customers I can't always remember which ones I'm allowed to talk about, but since the press releases this week, I can remember that, like they're brains not that old. You know, Roland and team, they went out, they wanted to be able to build scalable, cost-effective, high-performance infrastructure. He has an integrated ops team, engineering and ops team, so they wanted to be able to roll out capacity, network capacity, compute capacity, storage capacity, from central operations tool, they wanted to be able to monitor it, get alerted, AAA, they wanted everything to be kind of one model for that whole piece of infrastructure. Not a separate network operations team with their own rules and policies and procurement of equipment and things like that. Exactly, yeah, InfoSec or SecOps, whatever you want to call it, it's a collective problem, not an individual silo problem. And so when they go off and look at it, which like a lot of our customers do, they said, hey, if I have an operating system that looks a lot like my endpoint operating system, i.e. Cumulus Linux looks just like the Linux host that you're running OpenStack on, this stuff is all like super, super easy. So you get people that build these kind of Ansible or Puppet or Chef type playbooks or recipes or whatever that are like super easy to follow, they put them to get repositories, they push them out every day, they use a continuous integration type environment in their production world and it's not just for the network, it's for the whole system and it gives them tremendous amounts of power and predictability on what's going on inside their infrastructure. JR, the two areas that really have expanded this year at the OpenStack Summit are edge computing and containers. I want to get your viewpoint on that and from a Cumulus standpoint too. Cool, we're seeing the same thing, the place where I have my fingers crossed is someone figuring out, hopefully it's OpenStack, but we'll see how it unfolds, the world cannot exist on beat only or veggies only, so you don't get only VMs and you don't get only containers, the world's going to consist of both. And so finding that right marriage of how do you orchestrate both of those two things together in the same environment, I think it's really what people are going to start niggling against and bare metal for that matter as well. Hopefully the OpenStack community is able to really rise, I know people have an intent but it's like, will they get there? So I think we're going to see a lot of that happening and to the edge compute question, we see that all the time, a bunch of our customers stand out, points of presence in various countries because of privacy laws or locality, all kinds of different reasons we want to stand up capacity. One of the reasons why desegregated networking makes so much sense is if you, you can go to a reseller in Cambodia and say, hey, what server vendor do you have access to? And they say, hey, I get access to Dell. It's like, great, here's the bill of materials for Dell. I need a rack with these servers, these switches, these cables and everything like that. I need to stood up in this colo that I've got and I'm going to image it in a way that's super predictable and under my control. And when you go to Yugoslavia, they may have HP and if you go to Taiwan, they might have super micro and so it's access to the system IT, hardware supplier that makes sense in the right region and but you can image the whole environment and operate the way that you want. So it's cost effective and you leverage locality. Yeah, when I'm thinking about what's getting a lot of discussion in the networking space these days, micro segmentation seems to be one of those top topics. Kind of that insecurity, what's your take on where we are today and what you're hearing from your customers? Well, it's like we were talking about, right? The micro segmentation is kind of redoing the old world but in a slightly more automated way. Hey, I get VLANs on demand and they can stand up whenever as though the VLAN is the thing that provides isolation, it ends up oftentimes you need maybe a little bit of isolation around the edge of a system or whatever but in the insecurity and encryption I think are where you start finding the more of the simplicity in networking kind of coming back up. Credentialing things on purpose helps out so much in simplifying the system. JR, you touched on this a little bit with you're talking about the customer story and the familiarity of Linux. As you go out and talk with companies and with network engineers, it's kind of a new world. That's been another theme here at the show. Operators, they don't need to become developers but they do need to start to have a better sense of skills than they used to. Any thoughts about that? I'm only laughing because like literally 45 minutes ago I was talking to a customer and he said they're going through, they have an incumbent supplier, one of the traditional suppliers and their networking team had gone through tremendous amount of training with that traditional supplier and they all had a bunch of advanced certificates, blah, blah, blah and they're highly paid compared to a CIS admin and he said, hey, I want you guys to do cumulus and the networking team says, oh, I don't know. You know, we don't know that stuff. It's going to be hard. We're going to have to retrain ourselves and he looked at them and said, wait a minute, you mean my CIS admin can do this stuff and you can't? Maybe I should just fire you guys or maybe you guys should go figure this out and fairness, they may not be, you know, they may stay with their incumbent supplier but his whole point of it is look here at IT professional, your job is to figure out new technologies, figure out how to leverage it for our business, move our business forward and make things work really well. If you are past the point of learning or you need to go to some degree program from some supplier, you know, I need someone that can think on their feet and help me solve real business problems here, not, you know, kind of live in the past. Yeah, JR, I think back, you know, years ago I used to present it like the interop show and when I talked about changing the networking industry I would put up charts that showed decades. Literally, it was like from, you know, 10 gig from the adoption to the, you know, standards to, you know, oh, we got to get the copper cabling done and all these things, it takes years. Feels like things are moving a little bit faster these days yet, you know, when you hear certain stories like this, you know, that people, you know, people technology process, you know, can always bog us down. So what's going well and what do we as an industry need to do better on? Yeah, so I think the piece that's going well, and I don't know if we can take credit for it but we happen from a timeline perspective happen to hit it just right is the, the, I don't know, it's a land war between, you know, various silicon suppliers is amazing right now. If you look at what you get in a chip from a Broadcom or a Melanox, that used to be the bandwidth capacity of a telco not that long ago, right? You know, and it just keeps going. I mean, these people have roadmaps that can continue to apply more and more bandwidth to each, you know, subsequent chipset, better densities, bigger forwarding tables, bigger, you know, buffering tables. It is just absolutely amazing. So it benefits our customers dramatically, you know, long ago is, as you know, that used to be like Cisco would hold onto that stuff for 10 years because they would want to milk the old dollars as much as seemingly possible and drive things in at a really high cost and then finally make it available to the masses. But that democratization of capacity has been tremendously amazing. And I like to think that we helped participate in that. I don't know if it was just us or whatever, but it's definitely moving along. In terms of making things better, you know, one of the interesting things I hear from customers all the time is, which I believe is a bit naive, is they expect a, kind of a standard that's going to make everything work exactly the same. And, you know, in a, you know, altruistic universe that would occur, but it's very rarely is ever going to happen. You know, if you look at any piece of technology, it's driven by a bunch of interests that somebody has and the customers need to really recognize that yes, all these things are driven by interest. What things am I willing to make a compromise on and kind of follow that path? And, you know, we at Chemulus have made a bunch of choices where we think customers will choose Linux over other things. That's a choice that we've made and maybe they will, maybe they won't, but you can't expect everybody to adapt to one common overall standard. Yeah, absolutely. I think the last thing I want to ask you then is, building on what you were just saying there, it tends to get more complicated and people are choosing more things. Remember cloud was originally supposed to be super simple swipe a credit card. Oh yeah, joy. Now it's like buying a compute instance and the public cloud is probably tougher to do than if I was buying a server that goes in my data center. There's just more options out there. Absolutely. And people are choosing from an application and you know where their data lives. It's my data center, it's multiple public clouds, edges doing in. Talk to us about kind of the multi-cloud world and really the role of networking there from edge to cloud and beyond. Right, hopefully you don't think I'm boring or slap me down for this, but feel free to do it if you think so. We see a ton of customers that are multi-cloud. There's this premise that multi-cloud is this large fluid thing where you're moving pieces back and forth in this like, and what we see instead is people are super thoughtful. I'm going to take this workload, maybe it was an application or a data crunching exercise or whatever. I want the ability to put it in maybe a few different places, but when I put it there, it's there for a while and then it will consciously come out and go somewhere else. I'm not splitting it up amongst different resources or trying to cost optimize or anything. Typically people are doing it for capacity optimization as opposed to cost optimization, right? I've got a box of in-house resources. If I'm starting, I want to use those for high value exercises and I will expand into the cloud or maybe other companies will do things like, I get a new customer, I always stand them up in the cloud and once I get my head wrapped around how much capacity they need, I'll build that internally and then bring it back inside. But people need to be able to move back and forth, which kind of speaks to things like, need to be able to move data, need to be able to move virtual machines and pre-configured virtual machines intact or containers, that sort of thing, but not so much that fluid thing. Yeah, well JR, I'm still waiting for one of these really smart people in the industry to break the laws of physics so that we can handle that because absolutely data gravity is there. And even if I had infinite bandwidth, it's going to take time for you to get there. So JR, always a pleasure to catch up with you. Thanks so much for joining us. And for John- Awesome Stu, awesome John. Yeah, I'm Stu Miniman, getting towards the end of three days wall-to-wall coverage here at OpenStack Summit 2018 in Vancouver. Thanks for watching theCUBE. Thanks.