"To Infinity Paradigm and Beyond", a community driven podcast series by the industry veterans that you trust. Stay for the revealing interviews, expertise and education. Enjoy the light hearted conversations about life in and out and of the industry.
May 23, 2018
A conversation with Derek Collison
Derek is an industry veteran, entrepreneur and pioneer in large-scale distributed systems and cloud computing. He founded Synadia Communications and Apcera, and has held executive positions at Google, VMware, and TIBCO Software. He is also an active angel investor and a technology futurist around Artificial Intelligence, Machine Learning, IOT and Cloud Computing.
- Episode Transcript
Mark Thiele: Hi folks and welcome to the second edition of IDCA industry luminary podcast, otherwise known as “To Infinity Paradigm and Beyond”, and this time with Derek Collison. Derek, welcome to the show.
Derek Collison: Thanks mark. Great to be here.
Mark: I appreciate you taking the time. I know that you're a busy guy between travel and your love is outside of work and the work that you do, which you seem to always be finding something interesting to do. But before we get into the geek speak, tell us a little bit about what Derek's been doing for fun lately.
Derek: Oh, well, I don't know about fun per se, but I still love travel, obviously. I was just in Rome for a conference on human longevity, which is a great topic that I love to explore and tinker with a little bit. Living in San Francisco and L.A. Starting to really enjoy the L.A. life and decided to get a boat. I have enough cars so I got a boat and I'm really enjoying it.
Mark: Nice! Alright, well one of these times I'm going to have to talk to you more about the boat itself. But Human longevity thing is interesting. I know you and I have had conference or discussions in the past about A.I. and the influence on humans in the future and you know all the different impacts on potential human life and a longevity conference sounds interesting. Was there anything in particular that you pulled out of it to that you've found new?
Derek: Well, I'll be honest with you. I really think that the overarching theme which might sound funny is don't die stupid in the next 10 years. I think the advancements are coming so fast in terms of genetic modification with CRISPR technologies around mRNA, which actually is the worker bees inside your body that actually are replicating the hard DNA usually isn't doing that. It's telling the mRNA how to do that. And we got access to some folks in companies that are just on the brink, and stage two clinical trials to solve all degenerative diseases to solve cancer or to do some things that are really mind blowing and I think most people feel will take 10 to 20 years to come and I think there'll be here in two to five.
Mark: Oh, that's, that's really good news. I mean, we've all had those, friends and relations that have gone through difficulties with terrible diseases, cancer being one of them and watching them struggle with that and the idea that we might actually now in real terms be close at hand to solving that at scale, across the wide spectrum of cancer types as an example is awesome. You know its testament certainly to advancements in tech research and A.I. among other things I would imagine.
Derek: Yeah. You and I have known each other for some time and you know that I'm an eternal optimist, but even I was short-sighted in the timeframes where some of these technologies will be ready to be used in humans and like you said, affect people's lives in profound ways. So I would imagine now all of your listeners know someone that has gone through something like this or lost a loved one. And so it was great to be there and to see it first-hand. And so I do a lot of that, try to make sure I think outside the box or outside my day to day subject matter.
Mark: Yes, I think it's important. I mean it's funny, I was just talking to a friend the other day when I was in San Francisco, we met for lunch and we were talking about the value of hitting topics outside what would be considered the norm for a bunch of geeks to hit. It never hurts to have influence on your thinking from areas that you wouldn't normally consider as you worry about what to do with your technology next door. The technologies around you. Good stuff.
Mark: One of the reasons we got you on is because you've got a really impressive career in the tech space and you're one of the people that could effectively say, have left a mark on the industry that a wide portion of industry participants would know about. One of the most obvious ones are of course, is starting a couple of companies “Apcera” and now Synadia. But before that, even some of the things you did at TIBCO and then Google and then the founding of Cloud Foundry. Anything during that period, what are some of the most important things from your perspective that happened to you during that period and how do they relate to what you're doing now or what you're seeing now?
Derek: That's a great question. I think as you get on later in life, you start to reflect a lot on the things that influenced you and sometimes you see things that you didn't necessarily see in the moment. So one of the very earliest things that ever happened to me was my grandmother, I think it was like in 1980, got me a Commodore 64 and no one in my household knew what a computer was. No one had ever gone to college. And my grandmother had the foresight to actually get this for me. And kind of peeked my curiosity at playing with these things.
Derek: Fast forward to getting out of college. I was in the applied physics lab at Johns Hopkins University and probably one of the most profound things happened to me, which is I got selected by the second best physicist at the lab, not the first best and the reason that it was such a profound moment in my life and career was that the top physicists got all the supercomputer time and my physicist had 12 spark pizza boxes that they then looked at me and said, you figure out how to make those things do the same thing as a supercomputer. And at the time, I mean, it sounds very kind of general, of course you could do that. But back in the day, 1990, it was always scale up - vertical scaling, not horizontal scaling. And so I was kind of at a whim thrown into, you know, distributed computing when that wasn't a thing. And I wanted to do, you know, artificial intelligence, visual representation, graphics types of things with my career. And I continually just kept running into distributed computing problems. Even when I moved to Silicon Valley in 91, the first job I had, I immediately was confronted with the same problem and so I'm a little slow sometimes, so it took me about three years to realize that maybe this is what I should be concentrating on. But it was a huge opportunity for me that I probably didn't realize as much as I do now looking back, but getting early on into how you scale out horizontally and utilize lots of computers to complete tasks is obviously kind of at the forefront of infrastructure technologies and cloud infrastructure for sure. But in 1990 people were like I'm so sorry you have to do that. We will just keep going up with our supercomputers.
Mark: That's huge. I mean realistically to some degree you could argue that was a foundation for the reason why I joined your company Apsera in 2016 was ostensibly really helping people. I mean to extrapolate from where you started in 1990, to extrapolate from there a more efficient use of distributed compute capacity. Right. I mean, I’m really simplifying, a very amazing tool that the team created, but still in the end, that's what we were doing, right?
Derek: Yeah exactly. When I moved to Silicon Valley, I was trying to do advanced, large data analytics and visualization for a medical trial, right? FDA medical trial for a health startup. And I was very excited about that. And again, like we were talking about, I got run into just distributed computing problem. And kind of said, okay, fine, someone's trying to tell me something. But since 1992-93, my whole career has been spent on essentially horizontal computing; either how they communicate with time at TIBCO, what types of systems can be put together that are extremely resilient, but based on a lot of lossy type systems, again, at TIBCO and then at Google. Vmware was cloud foundry, which is, “hey, if we want to deploy these things into production, it's kind of hard to do right now with lots of moving pieces. How can we make this easier?” And of course all of those trends have continued and had gone well beyond anything I probably could have come up with or figured out. And that's a testament to the industry. But yeah, since 1992 I've been working on essentially the same technology area and that's quite a bit of time.
Mark: Yeah, no kidding. That's awesome. That's funny because you know, 1992 is when I first started in client-server. Before that I was working on a mainframe in an old data center. I kind of restarted my career in 1992, just coincidentally. Extending from there, let’s talk a little bit about your newest, love and startup. Synadia Communications came out of the NATS.io project which is now part of CNCF. Right. So, uh, what do you guys do in there?
Derek: Yeah, exactly. So for the listeners that might not be aware of NATS is a low level messaging system that allows digital software and systems and services to communicate. It's very different from the things I've designed in the 90s and 2000s. It goes back to a very simplistic model of at most once delivery and protecting itself, kind of like a utility, right? So if you think of a utility like electricity, obviously the electric company's job is to keep your lights on, but one of its other jobs is to make sure that any one bad actor doesn't take down the utility for everyone else. And I was seeing a trend in an enterprise messaging systems where they were trying to do so much for any individual client that a lot of times the system could become unavailable for the majority of the users.
Derek: So NATS was created actually as an underpinning for Cloud Foundry, which was a platform as a service effort that I had designed and created at VMware, which is now extremely successful. So NATS is actually new to some people, but it's been around for about 7 years. We built Cloud Foundry with it. We built Apcera’s platform technology with it. Again as a underlying substrate to do command and control discovery, location, addressing and things like that. And so Synadia was essentially an opportunity coming out of Apcera that I wanted to push the envelope a little bit on where I think NATS as a technology might be able to serve a larger population, a greater good. And where that comes down to is that if you look at open source in general, there are some challenges around the way it's commercialized, especially if you're not a very large company that has other sources of revenue.
Derek: So startups trying to be in the information technology and the cloud infrastructure space are mostly forced into open source and then you're running into a consumer bias that has to be free. That's a difficult place for the market to be in but that's where we are. So I thought quite a bit about NATS and a technology offering that that might be different, might be ambitious, but might capture the attention of both myself and some of the early team members and hopefully a larger market. And then that's essentially to connect everything. So every digital systems, service device, a single URL that will work in any cloud provider, any ego cloud or edge is secure by default, right? So we don't have to wait like we did with the Internet for the green lock to appear after 10 to 15 years. Promotes a sharing of data, but also gives you very strict and proper, you know, cryptographic abilities to do account isolation.
Derek: But again, just ubiquitous single URL dial tone just works, right. And we think that there's an interesting angle to making this decentralized. So a global utility that's decentralized by design. So I think blockchain is getting a lot of news and press, but I think there's two different versions, right? There's the cryptocurrency fiat ICO type of conversations where I think there's some challenges there and I think you're going to see a lot of regulatory involvement from state agencies and such going forward, but there's something I think very appealing to the notion of decentralized architecture, meaning there is no one owner of a global utility. It's actually a decentralized managed system that has both a notion of command and control that's on a public blockchain, but also provides for a transparent reputation system to allow other people to put resources into this global utility.
Derek: The way it works on the back end is that it's a utility, meaning you have to pay for it, like electricity, water, all that kind of stuff. But Synadia doesn't take all of that revenue. That revenue is actually put out on the blockchain and immediately redistributed in three big buckets, right? Eighty percent goes to the operators and of course that's divided up by how much traffic and how many people you're serving. Ten percent will go to publishers who are sharing data that the network finds extremely useful. Meaning there's a very large fan out and 10 percent will go back into the OSS community, again, Mark and I have known each other for awhile and he knows I feel that OSS is both a great thing, but it's also presenting some very interesting challenges that I don't think the market and the community will really understand until a decade later.
Derek: But I think there are challenges that, as much as I can as one person in one company try to show a path that might be a little bit different way of thinking about OSS instead of charity or as a support contract tax from big companies. That 10 percent, is kind of like a tax on all of the collected revenue from the consumers of the utility that always goes back right into the OSS community. And so if the global utility launches a no one uses it, obviously there's no money to be distributed. But as that gets larger and larger, I'm sure we will have made mistakes and have to autocorrect a little bit on our path, but I'm really interested to see if it could show a new way of actually showing that if OSS drives value, it should be driven back into the OSS in terms of incentivizing the innovation that takes place.
Mark: Yeah, couldn't agree more. I really feel like you've made a significant headway in the thought process because I know, I had many of these conversations with you and I'm not nearly as smart about it as you are, but I saw the struggle you are making in trying to find the right balance between making something and praying that somebody will give you money for it even though it's technically free. And making something that is really effectively integrated into the underpinnings of the web or as you put it, the new electricity or the new phone service. I think that's a really interesting approach and worth a significant amount of discussion all by itself. I mean, I was gonna ask you a question about the opportunity around Edge and certainly much more distributed compute, much more use of functions as a service, potentially even one of the conversations you and I had about multitenancy with IoT and data segregation or data governance at the Edge and it seems like depending on the situation, there may be a lot of opportunity for a product like yours to help manage that.
Derek: Yeah, I'm not sure if listeners are aware of, but I've been very big on the potential for Edge Computing as a massive opportunity. Now some of the lessons that I've tried to glean from Abcera, which was trying to do kind of an all in one multi-cloud platform technology that was actually mostly proprietary - there were some open source components obviously, but it's proprietary. One is that throughout my career, which is coming up on 30 years, we've ebbed in flow between different states like client server to centralize, client server to distributed. One of the other ones that we've vacillated back and forth between, at least in my opinion, is IT professionals looking for more of an all in one, one company, one throat to choke, so to speak, versus “no, just give me a whole bunch of different tools in my toolbox and I'll put something together myself.”
Derek: And I made a bet, which was wrong, that the market was swinging back to more of an all in one. And I think where I missed the boat, is that because there's so many new technologies being introduced at such a rapid fashion. I mean, think about it. We went from VMs, which were kind of really new and exciting to all of a sudden this notion of platform as a service and infrastructure as a service, cloud. Then we went to, you know, things like Kubernetes and, Serverless and Docker. All of these things have come in a pretty rapid successions and so I think IT professionals are mostly looking at trying to cultivate and prune out certain technologies to put in their toolbox and the three big areas that I think they look for are, when we're talking about cloud vs edge is how do I do compute at the edge, how do I do storage at the edge and how do I do communication at the edge? And so Synadia is trying to solve that one specific problem be that one, hopefully great toolbox technology that says, “Hey, if you want to communicate between things very easily, very, very performant and efficiently, but know secure by default, no matter where you are in the world” we want what we're trying to do it at Synadia to be that, that technology and that offering.
Mark: Yeah, it makes a lot of sense. Lessons learned to some degree and you know, I'm hoping for both of our sakes because I know that my forward looking, thoughts on the future of IT arent nearly as clear as yours are. But hoping that our recent experiences in struggling through proprietary vs open source questions will give us pause on how to approach the next phase. And it certainly sounds like it's given you pause on how to approach the Synadia. So good stuff.
Derek: And I think, one of the things at least for myself and I think a lot of people there, including you is that we always operated with a really high sense of integrity. We built an amazing culture within Apcera, which is something I'm most proud of. I figured out as I've gotten a little older that even if you make the right guess, meaning you know where things are going, timing is everything and timing is the hardest thing to predict. And it's the number one risk factor for any startup. Even given the situation we're in, I think we develop some amazing technology, got some amazing customers and I really do believe that we did the right thing, in terms of allowing a Ericsson to grab the technology and really purposely slotted in for the Telco vertical that they obviously are most interested in.
Derek: I tried to hold my head high. You know, it's tough to say, “hey, this isn't going the way we want.” But I think cloud infrastructure and platform infrastructure, those markets, they’re in a tough spot right now. It's really hard to make a business model that's something beyond support or professional services training or education. And I'm not saying that those companies can't be useful, they can. but I don't think you're going to see VCs keep investing lots of, of their capital, even though they have a lot of it in companies that at best can return maybe a 3x multiple, on a fairly limited based on working a human capital revenue stream. And so it'll be interesting to see, we've got so many things accelerating at such an amazing pace, from just base technology to a couple of trends that I started talking about a couple of years ago, which at the time sounded really funny, but I said, you know, everyone keeps telling me everything's going to software like software is eating the world and everything's going to cloud and I said, actually, I think we've already reversed and I think everything is going to go back out towards the Edge and everything is going to be either generalized hardware but more specifically specialized hardware at the edges and even inside of the cloud. And I think in those two years for the most part you saw CPUs and maybe some GPUs in cloud. And now of course we got CPU GPU,FPGA, TPU, which the new one, by the way is eight times faster than the one that just released last year, liquid Cool. So the pace of innovation is going so fast, but if you're watching some of these trends, things are moving from software back into hardware or firmware and I don't think you're going to see a lot of things moving to the cloud. I think you're going to see a lot of the technologies that you interact with on a daily basis as both an individual and a consumer and as a business are going to keep getting closer to you. Right? So metro offices, central offices, base stations, and then inside of buildings in terms of services and features that are delivered wirelessly but are localized. So I think those trends are going the way A.I. is still moving at such a rapid pace. There's going to be some amazing opportunities that if people can see the right combination, kind of the modern version of an Uber moment where, we've got smartphones, we've got maps, we get GPS, “hey, we can disrupt an industry” that, for the most part, taxis just didn't have a great satisfaction rating. So to speak, nobody jumped out of a taxi and said, “man, I love that. I can't wait to get back into one of those” and I think the way technologies are going and the ability access these technologies at scale for relatively speaking almost no cost as what it was even 10 years ago are going to present some opportunities that are going to be mind blowing. And, one last thing I'll throw in there is I suspect that most of these mindblowing opportunities, these combination of technologies to be put together are going to be coming from outside Silicon Valley to be honest with you. I think Silicon Valley - and I'm still part of that crowd - at least 20% these days going up and down the coast, has a bias. We're biased not to think out of the box. And I know a lot of people will kind of react to that and challenge me on that. That's totally fair. But Silicon Valley has a very biased way of thinking about how technology should be started, funded, matured, grow, all of other stuff. And I think you're going to see some of these developing countries or even different pockets within inside an already developed countries, get access to these technologies and see patterns and combinations that people with that bias, - even though you always want to say I'm not biased, everyone is - aren't going to see. So I think there's going to be some amazing innovations in the next five to 10 years. I think mind blowingly so compared to the last 100, to be honest with you.
Mark: I fully support that. Even some of the things that I've written about recently, on my blog and, and, uh, from discussions and even the presentations that I gave at a recent conference in New York. I would argue that that's true and that I'm more and more people are beginning to accept that. You mentioned Uber and as it relates to taxis and indirectly or maybe directly, but just not directly enough for my little brain to catch entirely. It sounded like a, you're effectively saying that the VC community as we've all known and loved it or fought with it, depending on your perspective is to some degree, right? For that same kind of disruption.
Derek: I have a lot of great friends in the VC community and I have the utmost respect for them and the VC community in terms of it probably being the single most important thing to why Silicon Valley exists and of course people are trying to recreate this throughout the world for sure. But I think most people looking at the VC industry who look at it at all, have also recognized that they are about to be disrupted in. And most people would say they're already being disrupted by ICO’s and of course regulatory things might slow that down. But access to capital is probably at an all time high and people are just looking for different ways to access that capital without having to give up so much in terms of either equity percentage of their company or whatever it be. And so I think VCs are smart enough to realize this and they are going to adapt. But I think there'll be disrupted for sure.
Mark: No, I think so too. I mean, I don't see the opportunity for them going away as much as I do believe that there's some opportunity for improvement in how the interaction occurs and how the influences is absorbed. But taking from those comments, a lot of comments and threads throughout this conversation, like the rapid change of tech, the increased impact on life, the increased impact on how we run our businesses, the potentially huge dramatic move away from, if not away from centralization, certainly in addition to centralization out to the edge and the potential complexity that gives to the average enterprise IT organization. Thinking about that and thinking about what I'm working on with the International Data Center Authority Technical Committee Team, The Infinity Paradigm itself, or what we might call an Application Ecosystem framework. We shared a couple of notes on that framework before the call, what are your thoughts about something like that? Something that gives the average owner of it infrastructure a better model for a viewing and operating and measuring either the effectiveness or the risks about the systems.
Derek: The IT world, the IT landscape, the enterprise IT departments, their complexity will never be simpler than it is today. Meaning tomorrow it's gonna be more complex than it is today and it's going to keep going. And the ability for people to be able to understand the effectiveness, the RTI, the cost is something that there's whole markets that have been created around, nuances around how do we actually kind of figure this out and it's going to keep becoming a more important part of the ecosystem in my opinion. Even in terms of what do you pay and human capital just to monitor systems. Everyone says: “well, you know, power and cooling for on premise resources is probably a large operating expense that we could maybe trade and get better at it, if we go to the cloud”. I would argue it's “what's the human capital and operating expense for running these things”. And I'm not saying that the only answer is moving to the cloud. I'm saying that's one area where I think a lot of people struggle at properly articulating it or accounting for it in terms of cost proposition.
Mark: No, I would agree. There are a lot of different ways that I look at it. One of the reasons why I got interested in the idea in the first place when I was approached by Mehdi Paryavi of IDCA, was just this basic notion of really giving the IT organization something closer to an ownership strategy for their technology use. Right? It's not about whether they can build bigger and badder or whether they have a better data center or whether they have faster networks. But rather a better ownership strategy for their IT as a whole and to some degree, again, a part of the theme that we've been talking about or some of the themes that we've been talking about through this entire conversation around the notion of, it's not about how much you have in one place or whether you did it or somebody else did it, but rather how do you effectively own your capacity to deliver services that are generated from IT. And whether that IT is servers in your data center servers in a Colo, servers in a data center that belongs to a cloud company, whether it's managed, or whether it's delivered via containers or whether it's a distributed app or a legacy ERP APP. There are appropriate ownership protocols in order to be able to deliver that service and those combinations of services to the enterprise that many of us frankly fail at, and the tools we've used in the past like CMDBs and charts on the wall last about as long as it takes to save them and distribute them, at which point they're out of date again. It's a lofty, lofty goal.
Derek: Why I do too, and I think it parallels a lot of ways at least at a high level what the CNCF is trying to do. Maybe not exactly, but CNCF is trying to give a cultivated list of technologies that could be effective in that IT toolbox that I was talking about, I think from at least talking with you and I could have this wrong, IDCA is going to look at, “okay, now that you've curated a bunch of tools and you put them together in some sort of combination, we want to present an Application Ecosystem Framework that can kind of give you feedback on how effective that is at solving a problem you want.
Mark: That's right. Yep. That's exactly right. I mean, at very high level, that's exactly right. And we want to be able to do that from top to bottom, it's not just about an application and it's not a dictate on how to design an application, but rather what to look for and how to build that Application Ecosystem that was the right way for the company that, it's meant to serve so far. So good stuff. We're coming up on 37 or 38 minutes and I wanted to get a little bit more from Derek Collison before we left. And what do you see if you had to pick something that you're most interested in, maybe other than increasing lifespan or..-
Derek: At least in terms of professional or the technology landscape It's A.I. I mean I studied A.I. in school, of course we went into, I think the second A.I. winter right after I got out of school. Luckily as we talked earlier, I got the second best physicist which put me on a different track, but the speed at which we're going, is mind boggling. So even to the listeners, if you want to get a sense of it, Google put a documentary movie together about Alpha Go and what you see is a couple things going on there. You see one, the first competition that they did where they won, but they lost like a game I think, which is part of that documentary. They had never even disclosed that they trained these models in the cloud, that they were running them on TPU’s that they had designed a specialized hardware, but it took up a whole room or so, it was very large and then they came back less than a year later and they trounced the number one player in the world and it was the size of a mini fridge apparently, I'm pretty close. I could be a little bit off, but it was two orders of magnitude smaller hardware footprint. It did everything faster and it just obliterated this guy, right. Then they came out with Alpha Go which had, no pre training whatsoever about how go is played. It goes back and whips the previous three generations of Alpha Go and it can run on two TPUs and those were the slow ones, not the fast ones they just announced today. And so when you look at it, we're looking at lots of data, lots of compute, even if it's specialized like FPGA A6 or keep using for training in the cloud to train these models. But these models are gonna come all the way out to your home to something that you're wearing within probably 2 to 3 years and they'll be able to execute these models instantaneously, zero latency, right for you in terms of speech translation, you know, dictation, all kinds of things. And so for me, I'm watching that and I'm saying: okay, the basic building blocks of the way we believe the brain kind of works. Some of that cutting edge theory isn't even close to what we're doing with deep learning, and so I think what you're going to see is you're going to see a massive proliferation of hardware, specialized hardware devices and chips to execute these models continue, but I think in the next five years you might see someone pop up and go I actually think I figured out the way individual neurons are communicating in a way that my model is going to two fold, two orders of magnitude better than the current deep neural networks that we have now that everyone is watching just get better and better at specialized tasks.
Derek: I think there's a bunch of companies now going all the way to the bottom going, okay, let's figure out this very, very low level simulate neuron, which is a very bad abstraction for the way the brain supposedly actually works. Someone's going to figure this out, a couple it with specialized hardware and we're going to have our eyes open to a world that I think people don't feel they'll see in their lifetime. I really think we're going to see it, you know, within a decade, to be honest with you. I really do.
Mark: Well one last question. Well maybe two last questions. One is, what have you read recently? What's the last book or the best book you read recently that you would suggest to somebody?
Derek: So I buy books a lot. I have not read a book in probably four years, but I have them. So if I ever do retire, I have a lot of books to read. The last book that I actually am reading now is on how to train a puppy because I'm going to get a puppy soon. So that's kind of funny, but I do read lots of research papers, lots of blogs on technology. I read probably two, four hours every single day on stuff like that, but I actually just had never read an actual book and quite a long time, but I buy them all the time. I have them stacked in bookshelves and all waiting for me to have some copious free time.
Mark: Yeah. I'm, I'm sorta similar. I do read still, but I'd probably buy four times as many books as I actually get a chance to read and that doesn't even account for the ones that people like my older brother actually send to me unrequested. So yeah, I've definitely fallen behind. But for your collection, it's actually beginning to change my thinking considerably on the future of work, etc. Look at a book called The Fourth Age.
Derek: You talk about work and what I think people may not realize is that if you zoom way out, the concept that our identity is mostly coupled to what we do for work is actually a very recent thing and it'll actually go away pretty quickly. So there's going to be a very short period of time in human existence where the first question you ask them when you meet is what you do for work and I think that's going to go away. I think A.I. will have a lot to do with that. I think A.I. will totally transform how people learn, not in terms of necessarily formal education, but I do believe that same timeframe as what we do for work also applies to how important universities and colleges are for our population. I really think when we zoom out it'll be, oh, for about 200 years, that was a really big deal that no one cared.
Derek: The reason I feel that that's applicable and again, the people listening might disagree, is about 10% of the population is very motivated. If you give them the tools they will learn and what's great about where we are at the landscape today is there's massive amounts of very high quality tools to learn whatever you want to learn starting from obviously a google search, but MIT has all of their curriculum online, at least for comp science, some of the math stuff. So I think A.I. is going to have a profound effect on, on education. And then of course the automation stuff. I invested in a company called Vicarious, which is trying to use A.I. in a novel way around robotics and automation will allow people to start self identifying with something besides work. I think there's challenges there for sure, even all the way down to the notion of what about, you know, income, do we have a basic income system, what does that look like? ButI actually think it will probably be a good thing, but it'll be a hard transition.
Mark: I would agree. I would agree a lot to see over the next 15 or 20 years for sure. Well, before we wrap up, Derek, where can folks that might be interested in keeping an eye on what you're doing, where can they find you? What's your twitter id or where would they find anything you might have? Your company? Your website?
Derek: So, @derekcollison is my twitter handle, and firstname.lastname@example.org is the best place to reach me. And Nats.io is a website around NATS as a technology, which is kind of the focus that we have right now. As we trying to bring online our global service will probably surface some more details about the company and our mission in this larger ecosystem. But if anyone has questions, uh, I'm usually pretty good at inbox zero, at least, uh, by the end of the day. So feel free to shoot me an email.
Mark: Oh, awesome. Well, I can't thank you enough for joining me today. This has been fabulous. It's been great. I've really enjoyed chatting.
Derek: Mark, I always enjoy our talks and I’m glad we could do it for everyone else to listen in to but we'll have to do a one on one here soon in San Francisco, thank you.
Share this episode:
May 1, 2018
A conversation with Jay Adelson
Jay Adelson is a serial entrepreneur, having built companies such as Equinix, Digg, Revision3, SimpleGeo and Opsmatic. Jay currently serves as co-founder and General Partner of Center Electric, LLC, a venture firm he started with Andy Smith in 2014.
- Episode Transcript
Mark: Hello everyone, I'd like to welcome you to the first podcast of a new series that I'll be doing with “IDCA”. IDCA is International Data Center Authority and we're working on the application ecosystem, otherwise known as the infinity Paradigm ® and as part of that work effort, we've created the technical committee, which I'm the chairman of. Part of attempting to spread the word about what's going on IT and trends technology today, we're hoping to invite industry innovators and visionaries for on a monthly basis to talk about, nuggets of wisdom and what they're seeing in the industry today. And on today's podcast, I'd like to welcome Jay Adelson. Jay has a long history in the technology space and was instrumental in defining and expanding one of the biggest sectors of our industry today. Colocation with Equinix. Jay, would you like to give the audience the long and short of where you've been and what you're doing now?
Jay: My pleasure to be here. Thanks for having me. I've spent pretty much my entire career in the internet infrastructure space, with probably a few little diversions in between for about 25 years or so. Some of my first jobs in tech in the early 90s in silicon valley where we're working for Internet companies that were just getting started before it was really commercially viable. For example, I ran that work operations at a company called Netcom, which was arguably the first free annoying disc you would pick up for free internet connectivity, before even AOL was connected to the Internet early days and I got to know a lot about that world pretty early. The benefit at the time was it was sort of coming on quickly, but there wasn't really anyone else there to receive it as it came. And so we got to learn in real time. So I went to go work for digital equipment and really that is where my journey started, which would lead to Equinix and founding that company began. the short version is that is that I was brought into the network systems laboratory in Palo Alto to work on a project called the Palo Alto Internet exchange. there were a couple-...
Mark: Yeah, I know exactly where that is in downtown.
Jay: Yeah, we were I guess the predecessor or the prototype to Equinix. There was a couple of folks there, Paul Vixie and Steven Stewart who had written a white paper and done a lot of research on whether or not you could build neutral colocation facilities that would serve as the alternative to the historical hubs of Internet traffic which were controlled by telephone companies. And, we tried it out, they brought me in to build it and operate it and make some tweaks to the model. We created the concept of the CNI or the Network Interconnect, which now every data center has, or PNI or whatever you call it. And within a about a year around 1996 to 1997 or so, we took this very inexpensive data center, took us only a few million dollars to build, filled it with, over 20 carriers and networks, realized that there was a huge demand because the internet inflection point had hit and the growth was starting to go at an exponential level and we needed to spend that model out and build something larger.
Jay: And so that's really where Equinix came in. And myself and my co founder Al Avery, we basically quit our jobs at Digital and, you know the rest is history from there. Equinix was great. I did that for five, six years. I then left a Equinix to start a number of other companies. I've actually started close to 14 tech companies, depending on whether or not you count some of the ones that never got funded or never got off the ground. But some of the ones that you might've heard of "Digg" the social news website. I was CEO of that company for quite a while for six years. I also founded a company called Revision3, which was really the first Internet television network. It's now the Discovery Channel’ digital arm.
Jay: We were doing podcasting before it was called podcasting and we had some great shows. if you ever have kids who like to watch the discovery online or discovery news, that's an example of a show that we created. So all of that kind of stuff was fun and that was my first experience building and creating content or having a large consumer oriented business. But my love has always been internet infrastructure and so I went back to companies like "SimpleGEO" and "Opsmatic" and other things that were really around the back end of scaling the Internet. I still have a love of that today.
Mark: Yeah. Well, I mean that's one we share it. It's one of the things that I was disappointed about in a positive sense, if that makes any sense at all. when I first learned that we'd be able to get you on for our first podcast is that here I am, I'm working in the industry that I love so much. I've spent time building my own environments and helping build an organization like Data Center Pulse, etc. And the two of us had never met in person, so I was glad at least we've solved the problem of meeting if not in person. And so that's outstanding. So, you know, storied career, no doubt. And before we get into, you know, more of a little bit of geeking out about what's going on in the industry right now, give the audience something about Jay. When Jay's not trying to figure out how to solve the next big internet problem or figuring out what comes after edge, or how AI might solve edge or something like that.
Mark: What does Jay and the family do on the weekend?
Jay: Well, it's a great question. sometimes I can be a little bit nerdy and I'm not gonna lie, I spent a great deal of time restoring pinball machines, which is probably not a very common pastime, but I just love doing things with my kids. I actually got into that because I have a 20, a 17 and a 15 year old kid and my son and I ( 17), we work on that project together and restoring pinball machines. So I have, I've advised a ton of companIes and so the, the key to sort of keeping things sane is really being able to shut off. And so I spend my weekends in Marin. I live about 15 minutes north of san francisco and my family and I, you know, do, do crazy stuff like, pinball machines and of course, you know, I have a ton of interest still. I actually went to school for film and broadcasting originally. And so I have people who are in the industry and are around videography or in filmmaking, and I still sort of have one toe in the water, so to speak in that world. And I love it.
Mark: That's cool. I mean, it's funny, I was not one of those growing up that knew all the director's names and knew who did the scores and stuff like that. And my daughter is, and it's great when she talks about a movie coming up or something and she can say, oh it's got to be good because Alan Sorkin wrote the script for it, that kind of thing. And I just never paid attention to that stuff before, and it is amazing. I mean, you watch something like West Wing versus a regular TV show these days and the quality of the dialogue, etc. That's fascinating stuff. So I was going to ask you what's your favorIte book is, or what's your most recent book is? But now maybe I'm going to ask what's the last movie you watched or what's your favorite movie?
Jay: I studied that and got a degree in it. I got a little obsessive about art film for awhile and I just been absolute "Stanley Kubrick" fan. I love his movies, they are just absolutely amazing. Although I watched the 2001 with my son not too long ago and he fell asleep watching the movie. You know, but I have my all time favorites. There's just so many. I have a real love of old Westerns, John Ford movies, modern films, I love these new "Star Wars" movies that have been coming out. They've been fantastic. I like everything. I even like Ready Player One.
Mark: Oh wow. I haven't seen that yet.
Jay: and that's a nerdy movie. but I liked the book even more. You know, these are in my opinion, great times for content. It's also, there's just so much of it that you can't possibly absorb it all.
Mark: That's the hard part. I mean whether it's for work or for play, the amount of available information is just staggering. Absolutely. And here we are creating some more indeed. So as I mentioned before, a very storied career, the experience you've built in building startups and helping others certainly founding a company that is now considered sort of a backbone of what the industry's doing relative to data centers and using data centers effectively. I mean because they more so than ever before, even though it grew out of the PAIX, the 'Palo Alto Internet eXchange' more so than ever before. People now are realizing that the data center is much more than just a secure building with lots of power and cooling. And to see how that has continued through companies like Equinix and how that's grown today and how different data centers you've differentiated is pretty amazing. So I imagine you're proud of being able to be a part of that.
Jay: When I see it from the outside, my mind is blown. I mean I remember in the first days when we were imagining what the maximum footprint Equinix could possibly need in order to serve 100 percent of the world's internet traffic. And we were talking maybe 15 to 20 data centers. You know, they have hundreds of them. And back then 15,000 square feet seemed like it'd be plenty.
Mark: now, it's a small room
Jay: now. It's literally the reception area, or the bathroom in a data center.
Mark: The climbing wall at somebody's data center here. That's funny.
Jay: I should've known better because I remember in the "Paolo Alto Exchange" back in 1998, we had a situation where we were running out of space and UUNET who at the time was one of the largest internet service providers wanted to expand their cage. And we took one of the bathrooms and we got rid of all the sinks and the toilets and expanded UUNET into the bathroom. So it's funny, now you're looking back at it, it's probably a bellwether for where we're going. When you start cannibalizing your bathrooms for data center space.
Mark: Well, yeah, I mean as a data center geek, every time I walked into a data center, and not only did they not have containment, but they had hot blowing on cold, right. Those are just some of the more obvious ones. But yeah I've been called back at 2 o'clock in the morning from a vacation camping up in Point Reyes for example, because one of the data center rooms we were using at the company I was at at the time happened to share a circuit with the wall out in the hallway and somebody was puffing the floors and popped the circuit and dropped five of our racks and it's just, those were the days, right?
Jay: And it changed quick, it changed very quickly.
Mark: It absolutely did. So when you think about a data centers and where they're going. I've got an either or kind of opportunity for a question here. What are your thoughts on the data center as a service space, right? In the sense of data centers being much more than just your data center in another place. Right? But more of a true service offering and I think we sort of hinted at that at the beginning of our conversation versus what do you think about the private data center market and what that might look like in five to 10 years. Do you think more people will continue to get out of the data center space? Do you think there'll be some equilibrium that we will reach in another five to 10 years where we'll have a 40, 40, 20 split or something like that? How do you feel about it?
Jay: I was talking to the Equinix people not too long ago and they kept throwing the term Hyperscalers at me and talking about the movement towards these sort of different types of data center products for different types of customers. Right? In 2005, if I built a data center, it really had to serve all the different constituents, whether you're a carrier or a network service provider or fortune 500 or somebody like amazon or google. The sense I'm getting is that we've really changed this market. I know it's not really Tiering because they're all very high quality products across all the different players out there. When you'll build a very specific type of data center for a player who has no intent of any other sub tendency where they're going to have one big space designed to maximize the CPU or "nets per square" foot and get great power efficiencies versus data centers that will never have a single type of one of those giant customers in there. There'll be lots of smaller cages. I guess from a philosophical standpoint, I believe that there should be fewer consumers of data center space. If it's not your core competency, if the product you sell, if the business you're in is not racking and stacking machines, it strikes me that you probably shouldn't need to visit a data center. Your developers should be able to do everything from the comfort of their office instead of putting out a parka and standing underneath the air handler and some data center in a cornfield someplace. I thought that was going to happen when people started moving their applications to the cloud. I thought that was going to happen much faster. And I totally understand that there are challenges around latency and proximity and security that they interfere. But I do believe that there are more data center service providers who are stepping into those markets anywhere on the spectrum from edge to core and it becomes less and less valuable for an enterprise to control their own controller and real estate.
Mark: I would largely agree with that. I don't know if the caveat to, as more of a way to get to what does that exactly mean? And from my perspective, it's pretty straightforward. I see it as, if the data center isn't really your business, and when I say it's your business, that means that you don't, once every five years identify four or five people on your team who have otherwise been doing their day to day work and say, go figure out how to build us another data center. You instead have a team that is consistently looking at where do we get power from? How do we build these most efficiently? Who do we partner with? How do these fit into the long term growth of our company? How do they fit into our desire to acquire other businesses or potentially divest businesses, etc. And if you're not looking at it that way, if you're not looking at it the way that professionals have to look at building and owning and operating data centers, it seems like that should be telling enough that it's not the business that you should be in.
Jay: That's absolutely true. Perhaps this was a little naive or maybe a little too early, I remember in the early discussions at Equinix saying to the folks who are building out these products. What are the common tasks that every one of our customers do and if we can find things that everybody is doing and then offer those as services, whether they be automated or manual, it seems to me that it's a logical direction to go as a data center company to provide those kinds of common services across. I think that in 1998 that might've been power and air conditioning, reliability or security. I think in today though with automated provisioning, SDN, other kinds of really beautiful ways to automate the rolling out of services, it seems like there's way more opportunity to go. I’d hesitate to use vertical, but basically to reduce the headache and the truck rolls for customers. There's lots of places to go there.
Mark: Yeah, totally agree. Again, to expand a little bit on what you're saying, it's things like megaport or direct connects to cloud providers or being able to share within a community of other companies that you're a part of, say from the financial industry or something like that. These are things that you can't easily do in your own private data center. Even if you own hundreds of thousands of square feet. It's just not something that's economically viable as a private data center owner. I think you just have to figure out the right ways to leverage those things. If you've got 200 megawatts worth of company proprietary stuff and you've decided to make it your profession to deliver that, then maybe it's the right answer. But for the average company, owning a data center, and I wrote about this five or six years ago, owning a data center is like trying to write a 15 year business plan and who can do that, right?
Jay: Exactly. And by the way, it's sometimes more fun to focus on your core competency and having to deal with those things. It's a quality of life issue.
Mark: Exactly. So I guess this is sort of related, but I want to ask you a question. Against what we talked about a little bit as we were preparing for doing this podcast and it's specifically about IDCA. We talked a little bit about what the framework is trying to accomplish. For the audience and as a reminder for you, the idea is to look at the technology that customer zone as an application ecosystem rather than looking at it as infrastructure or data centers and being specific about how to build or manage for them, rather it's how do you own the entire stack from site selection to application output for your customers to where data is and to how you do processes. Do you see that kind of framework as being something that's more important in today's world of rapid changes and the potential need to be able really quickly evaluate whether or not you're building what your company actually needs versus what you thought they needed five years ago Etc. Etc. Do you see that as more important today than it was 20 years ago?
Jay: Well, I do. Actually maybe a better way of putting it is, it was more important 20 years ago, but we couldn't get it. I remember back at the Network Systems lab, again dating myself back to Digital when, my bosses wanted me to go through the ISO certifications for offering data center services and it just was a square peg round hole problem trying to get those kinds of things done. And part of the reason why that stuff existed was because you wanted to be able to rapidly certify that when you're talking to your shareholders or your constituents or your customers that you are delivering a product that meets a certain standard and that frankly there's knowledge transfer, so that if at a later point there needs to be other people involved in this product or this infrastructure, there are certain expectations that you would have based on the fact that it was based on some kind of industry standard or I don't know how else to put it.
Jay: So yeah, I do think it's important. It's more important as people outsource these capabilities to others in the supply chain. It has to be consumable by someone who's not a data center engineer. So you have to create versions, sort of explaining each component for different audiences, which is hard work. We tried and failed inside of Equinix years ago to try and do things like that, you know, because we kept getting handed these requirements and so we would put our arms around some other industry standard so that we could try and use that to communicate. You know the classic example would be like in security, SAS 70 requirements. You know, I don't remember like the financial industry after 911 and where people were moving their infrastructure into our northeastern data center facilities, they wanted us to be compliant to something, they didn't care what, they just wanted to understand it. So when they talk to their bosses they could say, hey, this SAS 70 requirement or what have you. But I think it's similar. I think that it's just a little bit more since we're even more abstracted from the service delivery, I think that it needs to be pretty broad. and yeah, it's a lot of work. It's very valuable if you can do it.
Mark: Well it's funny, from my own experience, when Mehdi the president and founder of IDCA contacted me and asked me about working with the Technical Committee. One of the questions he asked me was what do you see as one of the problems in the industry? At the time I was thinking, well, it's the International Data Center Authority that called me. so I'm thinking data center. And so I thought about the data center and I said I can think of 100 really quickly, like we don't need raised floors anymore and we should have containment and we shouldn't push air up from underneath anymore. We should draw air from above and we should use outside air, etc, etc, etc. And all those things. But I said, if I had to pick really one thing, I'd say it's a better ability for the people that own and operate a data center to actually own the entire system. To be able to represent it effectively, to be able to, when they need to build more or buy more or use more cloud to be able to use a common set of practices for figuring out the right way to do that and where to go and how to capture the appropriate information about their company versus you know what it is they're trying to build. And he said, well, effectively, that's what we're trying to do. Only we're trying to do it for the whole stack. And I said, sold, right? I will still admit to anyone that's listening that I think it's a lofty goal, if we can get even 75% of the way done, I think it'll be well worth it because as you pointed out, I don't think outsourcing makes it easier, to own what you own. I think it's just the opposite when you don't have it under your care, when you can't talk to the Mark Thiele or the Jay Adelson in the hallway everyday about how things are going, you're actually putting yourself at more risk and anything you were doing wrong before it gets exacerbated. So I think it seems more important.
Jay: If you're a CIO or CTO and you're looking into moving your applications online or even to private or hybrid cloud or, or whatever, and you're going through that process and maybe maybe this is your first time you've gone through that process, but maybe you're a veteran. Part of the problem is that when you're abstracting, it's hard to know what you're buying. When I was a customer of data center services when I was operating Digg, it was funny, so here I was now finally a consumer of these services and somebody told me that it was being served out of Amazon eastern or some other cloud location. I didn't know what that meant. I didn't know what that looked like, I didn't know what the the standard unit look like from the entire delivery for the application all the way to the physical part. It was completely abstracted from me and so it was very hard for me to feel confident about it. The flip side is I didn't really have any choices anyway. There was only a couple of data center solutions and architectures right now, particularly post containerization. There's Just an absurd number of ways that you can combine all of these technologies together in the stack and you have to have some kind of set practices that you can point to. It is, it is lofty but valuable.
Mark: Thank you. Yeah, I couldn't agree more. So one more question that hints at IDCA as well, but really touches on a broader subject that seems like almost everybody that I talked to in the industry, especially the data center industry is worried about is that is the ability to get new talent and especially to get new talent that includes the appropriate and from my perspective actually beneficial diversity. And what do you think about programs like IDCA from a training and industry involvement standpoint and their value in the industry today to try to get more talent available and the right kind of talent, frankly, available to everyone that's trying to build out their IT environments or build out their data center environments.
Jay: Well, I love the idea of being able know, like when you say data center talent, there's so many different roles along the stack, and I would really like to understand when I look at a resume, what they know, I guess the analogy would be, you know, the Cisco certifications of the nineties, right? And how we wanted to understand what a network engineer could do. Cisco provided this, this sort of nomenclature for different types of certifications and there never really was that for other elements of the delivery. You know, whether it was physical operations or whether it was, you know, the application layer security and I mean there are but there was nothing so standard that I would see it on a resume. So that would be helpful. Uh, certainly some kind of training and then certification associated with it. I just don't know where I would start.
Jay: And the physical side or would you start on the application layer or...
Mark: I think as you described, I think all of it is required. Well unfortunately I have not gone through all of IDCA’s training. I think from a simple kind of external perspective relative to data center training, having an understanding of what the system of the data center is along with a deeper skill in an area of opportunity, whether how air conditioning works or how cooling works or how to manage a successful and safe and secure operations on the data center floor. There are a lot of kind of either obvious when somebody has already wrItten them down, but they're not obvious for those people who haven't thought about them a thousand times. Things like having a five minute training lesson for anyone who's going to come and work on your data center floor for the first time even if they're a contractor, rIght? Having them understand what they can and can't touch, what they can plug and unplug. Who they have to talk to before they can ask for or make a change. Things like that. Those are just little things. But from an operations standpoint, we can build the biggest, baddest, Tier 4 data center on the planet if we want to, but if we don't have good operational understanding of how to operate it on a daily basis, then it's a waste of a good iron and cement.
Jay: I agree. They're also just needs- we need to bring back the notion of apprenticeship into this industry. I feel like we're talking about talent and limitations and lack of access, you can create all the training in the world you want. But what would be incredibly practical is to be able to provide solutions that ultimately result in a certification, that are more In the apprenticeship or mentorship model. as opposed to say go sItting in a classroom offsite someplace and taking notes and taking a test, from some combination of the two. I remember in the 1990s and the 2000s and even more recently how much more effective my talent pool was that started in other jobs and then apprenticed into their new role up and down the stack. Hopefully I'll just put a plug out there, bring that back.
Mark: I think it's a great approach and in fact I know that from the IDCA guys, a lot of times they actually do training on site with folks in their own environment. I don't think it can get much better than that, if I could pull a nugget out of what you just said assuming I wasn't reading too much into it. I totally agree that the opportunity to expand someone's ability to contribute is not just to get them to become the best air conditioning specialists in the world, but rather it's for them to understand how not only their air conditioning works, but how it fits into the larger environment. And so that the training you talked about, whether it's knowing a little bit of code or having worked on helping some teams build server environments on the data center floor or some combination of the above are things I think can really help drive a larger team to be more successful even if only being more successful in how they communicate with each other. Right?
Jay: Totally. I mean, the best talent were all the people who can answer the question why. When you blindly architected some element of your facility based on some ancient standard, that's what you tended to do when you were - we used to joke that when we first had IBM as a customer at Equinix, they sent the people that invented electricity to evaluate their first product and they were just so used to data center, the same sort of data center model that existed for 20 or 30 years. And so to continue to innovate, you absolutely have to be able to answer the question why. Why do we believe - back to your air conditioning example - why do we care how cold it is and, and let's constantly readdress that question and ask that question again so that as the technologies change and the talent pool changes too that we're doing the right things.
Mark: Excellent. Excellent points. And I couldn't agree more so I've kept you for quite a while already. and I appreciate the time. Thank you very much. As we wrap up, one more shot at Jay, five years from now, are you still going to be fixing pinball machines on the weekend and starting or managing new companies, or are you going to be playIng golf?
Jay: I could guarantee I won't be playing golf. I played that game twice. My goal honestly over the next three, four or five years is to find excellent teams that, that need me on at a strategic level. I love serving on boards and I'm always looking for really good organizations that really could benefit from my experience. That gets me, that keeps me frosty, it keeps me in the business enough to sort of see what's going on. But yeah, and then also doing a pinball machine or two on the side.
Mark: Yeah. Sounds good. One of these days I'm going to have to find a way to share a drink and play on one of your pinball machines. I am not a pinball wizard, but I do enjoy the game.
Jay: Yeah. You don't have to be good at it to really enjoy it. That's what's really great about it.
Mark: Especially when you don't have to keep dropping quarters into it. Well with that, to the audience, please join us next time where we're going to have a Derek Collison of Initially Cloud Foundry and most recently Apcera and now he has his own startup that I'll let him talk more about, on the cast. And lastly, if you'd like to nominate a guest for our podcast, email us at email@example.com. Thank you very much to Jay and thanks for listening and we'll see you next time.
Jay: Thank you so much for having me.
Mark: Thanks Jay. Take care.
Share this episode:
June 13, 2018
A conversation with Mehdi Paryavi
Mehdi is the Founder & Chairman of IDCA. He is the inventor of Application Ecosystem concepts and methodologies and the author of Infinity Paradigm AE360 Standards Framework. He actively promotes genuine specialization across industry verticals suited to the precise application needs. His initiatives and notions have formed strong pillars for the information technology industry and reshaped the way our industry foresees its future.
July 4, 2018
A conversation with Ryan Fay
Ryan is the Global CIO, currently leading ACI Specialty Benefits global business and technology strategy across 170 countries in over 180 languages via multidisciplinary technology teams radically disrupting the corporate benefits experience leveraging decentralized applications, blockchain, and smart contracts.