 Live from Los Angeles, it's theCUBE, covering Open Source Summit North America 2017, brought to you by the Linux Foundation and RedHash. Welcome back, everyone, live here in LA for theCUBE's exclusive coverage of the Linux Foundation's Open Source Summit North America. I'm John Furrier with Stu Miniman. Our next guest is Erin Welch, who's the co-founder and head of product at Packet. Welcome to theCUBE. Thank you. Innovation's booming here. We're the product guys who love to have that product founder perspective of the collision between Open Source, accelerating at a massive scale, not just in the classic sense of all the normal projects that are getting more and more derivative projects, but new projects. You get the hyper ledger, you got IoT, you got a massive amount of collision going on between software and your world is about hosting all that and making sure it's on-premise support with low latency and multi-cloud architectures. So there's an architectural battle happening while Open Source is massively accelerated. What's your take in reaction to all that? Yeah, I mean, it's pretty interesting and I think especially with the sort of advent of containers on the scale that we're now currently seeing and obviously that's a technology that's been around for quite a while, but I think Docker finally fixed the user experience side of that and made it comfortable for developers to deploy on. And so now all of a sudden you have a sort of portability on the application level that the cloud sort of always promised but didn't ever really deliver. You never really ran a AWS instance image on GCE for example, right? You never had that real portability, especially across clouds or across facilities. But now with the advent of containers, both your development pipeline and your CI CD pipeline, once you've obviously made the investment to get that all running properly is so much more accelerated and so much more isolated from and doesn't rely so much on sort of the traditional infrastructure gatekeepers. So I think that development cycle is accelerating in that regard, but also it's enabled people to get sort of come full circle and now you have the ability to deploy your workload on specialized hardware and target that specifically, right? So we're going from a sort of like very abstracted cloud environment where it's a certain amount of RAM and CPU, you don't even necessarily know your clock speed to I want to push my SSL offload to my network card and people are able to do that. So that's an interesting sort of thing of the last I would say. So Aaron, I want you to take us back to the founding of packet. What was, what was it going? I mean, we look at technology is changing so fast. We're talking about containers. Heck, you're in New York City. We're going to probably be there. Serverless conference is going to be there. Amazon's pushing the next generation. There's always the new, new, new thing. And there's companies that come out with a new, but the big guys are also jumping all over it. So, where do you guys fit? What was the impetus for the start? Yeah, absolutely. Well, you know, it was an interesting time. You know, and most of the people when we're starting the company were like, are you completely out of your minds? Why would you start, you know, that game has been won. The cloud game is over. You know, AWS has taken the prize. What are you doing even getting into that space? But, you know, around that time, 2014, was when sort of Docker was coming out. Kubernetes was really shortly after we sort of started the company. And we felt like there was a real value around the economics and performance of dedicated bare metal hardware that just wasn't accessible by people who had cloud native workload requirements. But with the advent of containers, now all of a sudden they could do that. And, you know, we really felt like there was a gap in the space where all of those big players are still virtualized. They are still co-tenant. They're still, you know, using a hypervisor. And so if you're running your workload in containers, you're paying a virtualization tax. You're paying for co-tenancy. You're paying for noisy neighbors. You're paying for generally poor shared network. And so there wasn't anywhere in that space for people who are either on-premise or in a dedicated environment that wanted to move to a more nimble cloud infrastructure and vice versa. People that were getting pinched on the higher end of the economics and paying for poor performance in a virtualized public cloud, wanting to go the other direction, but not wanting to give up all of the flexibility and tooling that they'd invested so much in. So being able to actually automate bare metal, give that experience of the cloud with the performance and economics of dedicated servers was something that was missing. Aaron, take us inside the customer viewpoint here because I've been looking at surveys and they say, you know, how much are they deploying? Things like we used to talk past. Things like cloud foundry, you know, containers, Kubernetes, you know, Mesoscones going on here this week, you know, lots of companies that are using it, but, you know, from a production standpoint, you know, the data is still a little bit muddy. What are you hearing from customers? Where are they with that kind of, you know, cloud native, if you will, or kind of application modernization? And is that a necessity to leverage your offerings? Because I guess I look at most companies and say, oh, I've got thousands of applications. I've got some that I'm doing new stuff with, but I got a whole lot of stuff that I need to do something with. And there's lifted shift or platform it or anything like that. So how does that whole, you know, maturity fit into your picture? Yeah, for us, we're very agnostic. So we really focus on sort of automating that fundamental infrastructure. So we have plenty of people that are bringing their own hypervisor and standing up their own sort of traditional virtual clouds using KVM or Xan or VMware. We have people deploying OpenStack on our platform. But I feel, I definitely feel like in the industry in general, the excitement and the inertia is definitely behind containers. I would say it's still very early. I think most large enterprises that have a cloud transformation strategy or initiative internally that have their own team are still in the proof of concept phase right now. You know, they're still playing around with it. I think it's probably going to be a couple of years before they're really deploying serious workload. But I do think that sort of there's a large, but still I would say mid-level enterprise who are paying consulting agencies to boot them into microservices framework using Kubernetes and Docker and those other frameworks. So still spending real money, still doing it, but are paying some of the kind of leading edge pro-serve firms to do it for them at this point. And then of course there's just I think massive amounts of just development and experimentation on the sort of individual developer front. Eric, talk about the impact of Amazon web services because we're seeing a lot of workloads certainly going there. Stu and I have commented that, kind of, well I've commented, I won't speak to Stu. All rows lead to the cloud, eventually. Because as startups, start companies, we are all in the cloud, no one starts a data center and says, I'm going to spend all those cat-backs. They might have an operating model at some point, realize something on-prem needs to be there. So you have that kind of mindset. A lot of hybrid going on, which is basically true private cloud as Stu points out. Your thoughts on this trend, how much of that debate of Amazon workloads being real vis-a-vis their competitors and on-prem, what's really happening in terms of the trends? Is Amazon running a lot of workloads and what kind of workloads and where's the on-prem gateway to the cloud happening? Or do you even agree with that? Oh well, I mean I think without question, I mean Amazon has been validated by the market. They're running plenty of production workload, absolutely. I think that their strategy of sort of having a diaspora of supporting, sort of hosted services has served them very well. I think EBS and obviously their object store S3 has been like fantastically successful in keeping that revenue sticky. So I don't really, I don't hear anybody debating that anymore. The debate in my opinion is much more like how fast in is the Kolo world going away? And the number show it's still growing, right? Data center providers are still building more data centers. People are still selling more Kolo. So the question is, is that new workload that's going into those sort of on-prem or physical data center locations that aren't necessarily on a cloud provider? Is that hybrid workload? I think most people at this point either have something running in Amazon or one of the other cloud providers. And either that's fine and they're going to expand there or on-prem or have a DR strategy or something. But I think most people aren't going to put all their eggs in one basket with one. I agree Amazon is definitely kicking ass. Yeah, absolutely. Check the box now. Get your thoughts on architecture. So things that we're seeing a lot of practitioners doing is saying, okay, look, I got to protect my jewels. I got to get my own house in order, which is usually their on-prem situation. And changing their operating model. Same time, you mentioned the virtualization tax. There's still the aspect of I still want to run software define something that's either servers, hyper-converged to stoop. Let's talk about many times. A lot of that, they look at a new architecture. What are some of the new, cool reference architectures that you see emerging that give the enterprise a great solution to maybe take baby steps to the cloud. Maybe certainly bursting is in the cloud. We see that a lot. Maybe some production workloads make sense there. We've seen that too. But there's still the legacy enterprises. What's the preferred architecture? Yeah, it's tough. I mean, for those legacy, I mean the problem with most of those monolithic applications is that they're so bespoke. That it's pretty difficult to kind of, you know, say, this is your solution. This is going to solve all your problems, right? And a lot of times, in the container world, they cause a lot of problems that you may not even be aware of, right? Like especially in the security space. All of a sudden that perimeter is disappearing, right? Like your services move around, they're dynamic. So, I mean, I don't know if there is a reference architecture, at least that I see that's sort of coming out as a standard. What are your clients doing? Well, like I said, you know, our clients are kind of all over the place. We have people that are running containers directly on the bare metal. I think that Kubernetes is probably the, you know, leader in at least the orchestration space. I think, but there is still a lot of workload that is just virtualized. I think also there's, you know, there are some interesting kind of like, I don't know, dark courses out there in the virtualized container space like Hyper. So these super, super lightweight actual virtual machines that behave similarly to containers and sort of have, you know, are compatible with the Docker API, but that actually provide the isolation that true VMs do. So I don't know exactly where that is going to fall out. And the other kind of piece that's floating around out there is all the serverless, right? I think you eventually you're going to see people that are literally just deploying functions to the cloud and those functions are likely going to be deployed into some edge compute network of some sort. And they're going to be used absolutely for things like IoT, smart cars. I mean, anything that has, you know, requirements where you have to have a response time of 100 milliseconds or less and you're deployed in major metro areas. So, you know, it's just extraordinarily disruptive. And, you know, when I speak a lot of some of these, you know, events and that's the question is like, so what tool should we use? Like what should we do? There's no silver bullet. There is no silver bullet. And quite frankly, it's still the Wild West out there. Like there is no lamp stack equivalent for the container space right now. And I think that will eventually happen. But I think we're a little bit away from that. Yes, sir. And one of the things I think we've seen the maturation of Kubernetes has happened pretty fast. One of the announcements this week was the Kubernetes certified solution providers trying to make sure, where do I get it? Is it, you know, got an open stack where it used to be? Is it pure? Is it forked? It's, well yeah, do you have a solution that doesn't work in for Kubernetes? Does it actually give me that portability that you talked about? What are you seeing in the Kubernetes community? What would you like to see even more? We've, you know, watched this kind of tsunami come. So, you know, what's good? Where are the customers? And I'd love to follow up on some of the serverless stuff. Yeah, absolutely. I mean, I think that it's great that the certification program is coming out and like kudos for the CNCF to kind of get that rolling. And I think it needs a lot of work. I think anytime you put a certification program together like that, especially when you're starting and the technology is so volatile, you run the risk of it being a waste of time and or irrelevant. So, you know, I think that the Kubernetes project in general needs to look at some sort of longterm non-breaking releases. I mean, I think, I think right now it's still pretty dangerous for you to deploy anything like serious on it even though I know that there are definitely people that are doing that. But those are always early adopters and they know they're going to probably have to pay the price. But for Google, Google does it, you know. Google has some resources that a lot of people don't. Well, I mean, you know, like Google does a lot of things a lot of people don't do. They got a clause in the goodness they can just bring out. Sure, and you can either be, and you can be fine with that and still be successful with that. You just should be aware that you're going to have to pay that price in terms of resources and potential downtime and breaking changes and whatnot. But I think for widespread adoption, especially in the enterprise, there has to be a certain roadmap and a certain amount of security there to feel like you're going to be, you know, you're not going to lose your job for deploying something, right? Yeah. Well, my big thing that I look at, Stu and I look at is that the true private cloud report that Wikibon put out recently is getting a lot of traction. Certainly at VMworld a couple of weeks ago was they went viral and that was the revelation and I've been having a lot of debates with the Amazon folks here. I'm not trying to say Amazon's not winning. They're blowing it away and there's a zillion workloads moving to the cloud. But, certain enterprise for the bespoke applications and many more other reasons of legacy baggage that they just aren't moving as fast. So they're doing a lot of retro onsite, changing their operating model, which takes some time. Then they kind of move to the cloud as quick as possible where they can. So that seems to be the trend. So with that in mind, you're starting to see people jockeying for position. We saw Pivotal launch the Kubernetes Container Service or KPI, the Pivotal Container Initiative which is tied to Google Engine. Is that real? I mean, because now you have a lot of different things happening. It's so confusing. You got Docker, now you got Pivotal, but Pivotal's got some cred, VMworld's involved and you're tied to Google Cloud Engine. Making it management easier? Sounds good on paper. What's the reality? Everyone has their own Docker solution, right? And some of those are more bolted on than others. I think you saw Mesosphere with their DCOS. I think you see Docker appearing and Canonical and even Susie over here has something that I hadn't seen before. Everybody wants to kind of get on that Docker bandwagon and some of that is more native than others. So yeah, I mean. At the end of the day, it's got to work. It's got to be manageable. It's got to be manageable and I think that's right. I think in the end, operational reliability is what's going to win and lose, right? I mean, the bottom line is a lot of those foundational technologies are quite old and it's all about having visibility into your stack, being able to troubleshoot it quickly, being able to deploy it seamlessly, being able to truly segment your sort of, if you're really going in that full DevOps mode, giving your developers the freedom and nimbleness that they want and making sure that the DevOps and traditional security teams have the good, warm, fuzzy feeling that everything is in out of control. And so whatever that is in terms of integrated, CI-CD pipeline, virus scanning, security, integration. So the better mouse track will win. Yeah, absolutely. Final question I have to ask you. I look in your bio and it says something about fire eating. So how do you? I didn't bring any of my fire eating equipment. So sorry. How do you make some of this bleeding edge technology safe so that customers aren't feeling like they have to eat fire? Yeah, get burned. Yeah, I mean, I think it's a really good question. I mean, I think with anything, it's important, especially if you're deconstructing a monolithic application is really trying to find those pieces that are safe and easy to carve out and use those as a little skunk works internally and really put it through its paces. I think, I don't think I'm breaking any NDA or anything, but I went to a very great event at Moody's a number of weeks ago and they did basically an internal all day summit. They have a cloud transformation project that's running there that's going very well. And they did something fantastic where they invited the entire company and they did basically a day long camp just internally and they invited a bunch of people out of the space to just come and inform everyone and making sure that everybody sort of understood what the ramifications were, give them an opportunity to learn. And it was really clear that from the very top there was buy-in around making sure that this works and making sure that people have the tools and that those teams trust each other and understand what's happening as much as possible before moving into something that they may not fully understand. So it really was kind of a kumbaya messaging, gathering to get people on the same page. That's pollinating ideas, testing, getting feedback. Yep, they ran multiple tracks and so you could go talk, they brought in experts from various areas of the field and gave their staff opportunity, staff questions. And which is... Should have jammed it down that throat, it would have been more open. It was open policy and it was really great. They streamed everything so people that were working remotely could watch it and they had a really fantastic turnout. I think they had many hundreds of employees to sort of take the day. Thanks for coming on theCUBE, appreciate it, congratulations. You've been three years as co-founder and you know it's a battle to have your own startup. You're a warrior, great to see you on theCUBE. Thanks for coming on, sharing your perspective. Absolutely. I always love to have the entrepreneurial perspective here on theCUBE at Open Source Summit North America Linux Foundation. I'm John Furrier, Stu Miniman, more coverage after this short break.