Adventures in Technology Pt. 2
In part two of our chat with the global editor of DCD, Peter Judge, we continue our discussion from taking conferences online, open-source hardware, to how Jevon’s Paradox applies to data center growth – and why we aren’t worried about the idea of the world needing fewer data centers.
Announcer: Welcome to “Not Your Father’s Data Center” podcast, brought to you by Compass Datacenters. We build for what’s next. Now here’s your host, Raymond Hawkins.
Raymond: Welcome back. Reminder, we are still recording here on Friday, August 7th, with DCD’s global editor Peter Judge. We welcome you back to the rest of the conversation with Peter. Enjoy. Yeah, I think what DCD writes about and there’s a target market for it’s such a special view of the world of hey, what is Datacenters? What are we doing in the Datacenters? How’s it changing? How’s it providing a foundation for all the technological changes in our world and it’s such a significant business that there’s a core readership there.
But you’re right, all these other print papers and I think your point about newsstands, there’s nobody walking by newsstands these days, right? Or I say no and you’re right 5% of the people that used to walk by a new stance walking by today and what a tough spot to be as those changes that were coming get accelerated all around us. I think you’re 100% right there. All right, well, I’m gonna drive us back a little bit more to Datacenters and technology into DCD’s. So as I understand DCD, there are three or four legs of the business, right? I know, you have a print magazine, which we just talked about, there are events, you have an online publication, is there a training arm as well? Could you give me a few minutes of…I know what your role there but all of the DCD?
Peter: Yeah. Okay. I mean, I tend to talk about readers because I’m biased, but these same people are like a community who interact with us in various ways. The company historically is as much to do with the events as anything else. I believe way back something like 20 years ago, the people that founded it started out with more with the newsletter and some get-togethers. So the two things have been going together all this time and the training has kind of built on the side of that.
But so March this year, we were building up to one of our flagship events, we have conferences around the world, several of them in the States, usually three or four in a year in the North American continent, and half a dozen or more in APAC and several in Europe. So we have these global series of conferences and people come to them. So but we literally set of three or four weeks before the New York event was due to open, we realized that it wasn’t gonna happen. It was just before lockdown got announced here and in states.
And it was clear which way the wind was blowing but so DCD did what it calls this pivot to the digital, which meant that sort of people came into the office one morning and they were selling passes to an event, they were selling space in adverts of the event and maybe even standstill for an event that was happening one month way in New York. And that when they went home in the afternoon, they were working on a digital-only event.
Raymond: It was that fast, yeah, yeah.
Peter: And I think a team had been getting that project together and it sort of at first this was, we’re having to postpone the event and on the days where the event would have happened our delegates have all booked the time, we’ll give them something online to occupy them. So that was the event and it was also do the job of the event to an extent by putting out together the issues and subjects that we thought were front of our mind and would be front of mind for the delegates. We could still sort of push those discussions along online and make something happen and it worked brilliantly.
We had actually sort of more delegates, it went into the thousands the number, it’s quite hard to get more than 1000 people together in any one place for a physical data center event. But for an online program of webinars, we’ve had thousands of delegates to these and sort of after that happened, we knew we could do it, the strategy then sort of emerged piece by piece, and it’s kind of heuristic way. So okay all the events we’re doing in the summer part of our program will have an online version and we will postpone the physical event. And so later in the year because once we got the hang of the fact that we can do these online events successfully, these events became actual replacements for the physical events.
And, you know, the community of sort of vendors and users and customers and everybody that sort of in the sort of the DCD caravan kind of stuck with us and went along with it. And so we’ve now had since the beginning of lockdown, we’ve now had eight online events, which have by and large gone pretty swimmingly. And it’s sort of this week, we’ve just closed the virtual doors on the last one of those in the early part of the summer. We’re sort of gearing up for a program that restarts in September until Christmas and that’s very clear, these will be online events.
And we’re able to sort of pull out sort of track specific events that focus on a particular area, so it’s much clearer who the people are that will be wanting to come to them. There’s a lot you can do with a virtual event that’s sort of flexible and trackable in a way that a physical event isn’t. So, yeah, it’s been good. On my part, I’m looking at my colleagues in the events business sort of breathing a sigh of relief from collapsing, after all that, and thinking I’m lucky because my bit of the company is more or less unchanged. We still do the same news on the website, we still do the same magazine and we’re working from home more than we were but we’re still doing the majority of our newsgathering in conversations like this.
Raymond: Yeah, yeah, I got it. So DCD, your summer sort of for lack of a better word, summer break here in August and September, and you guys the rest of your conferences for the rest of 2020 will be online any eye towards and I know so much of it has to do with COVID have been in the eye towards what 2021 looks like and I know you’ve described that you guys have had great success in the online events but any idea where we’ll go back to meeting together in 2021?
Peter: We’ll have to be guided by what’s there, by what’s allowed and what’s practical. We’ve not got anything published on the site about any physical events next year, we’ve not got anything published yet on the site about any events next year. Judging by the success of what we’ve done, the virtual ones will happen again, and the physical ones will restart when it’s practical.
Raymond: I’ll just weigh in as somebody who likes to meet people in person, I hope we get back to a safe environment where we are comfortable being around each other sooner rather than later. Just, to me, nothing beats getting to meet people in person and shake hands and get to know them and build relationships in-person, to me is superior but we’ve got to work around, which is working today and sounds like doing great for DCD?
Peter: Yeah, and certainly, I mean, that my colleagues in the in DCD agree with that 100% and I can be very sure that just as soon as it’s possible to have a physical event, it will happen. Whether we do a sort of a very focused small event that’s very local someplace with a certain amount of social distancing, that might be a possibility, whether we have events in maybe a central location with virtual links to them bolted on, I mean, what if webinars were broadcast versions of a small audience or something like that? There’s a lot good to be done and where people at DCD are having a good look at all those possibilities.
Raymond: Gotcha. All right. Well, I’m gonna put you on the spot just like I did, tell me the technology trends you missed and then the big surprises that were things that you didn’t see coming. Do the same for me in the Datacenter space itself as someone who’s been in DCD five-plus years and been around technology for most of your professional writing career, Datacenter-specific, what’s been your biggest wow, that’s an incredible change, wow, that’s an incredible development, Datacenter specific?
Peter: Datacenter specific?
Raymond: And in the joy podcast land, you can take a minute to think about it.
Peter: Absolutely. I was thinking service providers, I’ve gotten a good one from sort of further back. I’ll do this while I’m thinking about the datacenter one because when you mentioned the .com boom, it’s brilliant. In the year 2000, I just finished doing my Y2K stories, and I got an offer to become a telecoms analyst, so I did that. I started working for a company in San Jose. I went to the Bay Area to work for [inaudible 00:09:51]. Well, to be honest, I was still working where I was. I was their European head, in fact, their only European employee working from home in many ways a little Coronavirus-like event and working from home for this company out there and then just occasionally visiting the mothership. Within a year, I was back out and the .com crash was happening.
Raymond: Yes. Right after Y2K, I mean, everybody was so excited, and everybody had a great year and then early in the 2000s, it all fell apart again?
Peter: And I think there’s a lot of things that happened then. And since then, in telecoms companies that have fed into the datacenter world, you know, those telecoms companies once said that they’d invested heavily in making the backbones which were over provided and that enabled the, when we were really ready for it, the internet to really take off. We had all these empty pipes that could be filled, bandwidth was effectively free and off we went.
Raymond: Sorry, I’m gonna do one quick another book plug since you mentioned the changes in telecom, there’s a very informative book called “Tubes” about the beginning of the internet and to your point, the change in the telecom industry and why did all of this bandwidth become so relatively inexpensive? Andrew Blum wrote it and it’s a very interesting take, a very interesting history lesson and about the early days of the internet, so exactly what you’re talking about?
Peter: Yeah, that’s a lovely, long-form, journalistic journey, isn’t it?
Raymond: Yeah, it’s a good one, I enjoyed it.
Peter: Yeah, so datacenters, I think the excursion I went out on was about open source in that four or five years ago Facebook open-sourced its hardware, the open compute project, published the white label designs that they required, they gave to the white label producers that were making their hardware instead of buying a server from IBM or HP which would have a lot of things they didn’t need nor sort of keyboard ports and a label on the front of plastic bezel was things that weren’t required datacenter, instead of having all those, they would strip it down, have a much more reduced design, and give that design to a third party to make a Taiwanese manufacturer.
And whereas some companies might have thought we’re really proud of this non-low vanity server that we’ve designed for our own use, we gonna hang on to it, they kind of thought, that’s not our core business, we don’t really want to have to keep designing these ourselves let’s share that and then with other input, we’ll have even better low-cost white-label service. So open-source hardware became a thing along with the open-source firmware and software, the network switches that go along with it.
So at the moment, if you’re building a cloud application or a cloud service inside a datacenter, you can use much cheaper hardware with designs that are shared with a big community of people. The open compute project has been one of the exciting things in the last few years. It’s a story which kind of maybe doesn’t run along as quickly and excitingly as you’d hope it would in a way like the whole open source project. We know that open source is great, it goes up to a level of hype and you think it’s gonna take over the world. A few years later, it hasn’t taken over any really obvious parts of the world but it’s there in the underneath thing in the underlying layers at the technology, just making everything work as it should be.
Raymond: So Peter, I like the way you described it, that it’s running along slowly and hasn’t quite taken over the world, I will take that same concept. So I’m completely with you on open compute and publishing their infrastructure standards and saying, hey, at the end of the day, we don’t think there’s… I think what that means is we don’t see a ton of value. In that design, we happily let other smart people tweak it and we view it more as a commodity and so whether it’s open-source software, open-source computing, I think that it’s saying, “Hey, the standard is the standard, and there’s no value meaning, we don’t have some secret sauce that we get to keep from everybody else in charge for the secret sauce. We just want it to be robust and well-done and the more smart people to look at it and improve it, the better. But we don’t need any proprietary hidden value that we keep from the market and charge for it.” And that I would completely agree.
I’ll back up you said it ran along slowly and hadn’t taken over the world, I feel that way a little bit about just the notion of cloud computing. Now to help inform this comment, I spent most of my career selling technology and selling technology from one OEM or one manufacturer to an individual corporation user enterprise that used it. And this notion of cloud computing which was it’s gonna take over the world and now I asked you the things that we either laughed that didn’t pan out or the things that did, do you remember when we used to call, we have this term in the ’90s, we called it thin-client computing. Do you remember that term?
Peter: Oh, yes.
Raymond: Yeah, the thin clients were gonna change the world that we were gonna do this thin-client computing and what were the other guys down in Florida that had, what was this, was it Citrix? I’m trying to remember the early thin-client operating system. And everything we were gonna do away with computers and everything was gonna be a thin client and that didn’t quite change the world like we thought it would, but yet we’ve ended up there, right? Clients continue to get thinner and more and more of the compute happens inside the network.
Even to the point, I enjoy the Cisco network and I promise there’s a point in this eventually. But I think back to my early compute days to what we called service bureaus, right? There was a mainframe and you drove to the service bureau and you hand it in your application, and they ran it and then they handed you back your report. And even though we don’t access remote computing like we did the old service bureaus that to me is what cloud computing is, right? I’m accessing compute function, thankfully, I get to do it through a wireless connection today but I’m accessing somebody’s compute resource somewhere else.
And what has surprised me is, as you said, it’s been a bit of a journey, and it’s sort of wandered along and hasn’t overthrown the world. We all think about Microsoft, and we think about AWS and we think about Azure and GCP and we think they’re taking over the world. But today, only about a third of the global compute function sits in Azure, AWS, or GCP, that’s not taking over the world. I still think it’s coming but I’ve been amazed at how slow the pace of change has been is where I was going with that long winding convent is that I thought the cloud would be a larger percentage of compute faster.
Peter: Yeah, I mean, especially as, you know, before the word cloud came out, we had application service providers.
Raymond: You are right. What we call the XXPs, the ASPs, right?
Peter: Yeah. That’s right, which is really just the cloud but we hadn’t yet got that name for it.
Raymond: Right, right.
Peter: But when you say only a third, when you think that’s a third of a quantity that is growing massively fast and then within that vastly growing thing, Amazon Web Services, let’s forget about the others for a moment, that’s grown to be that much of a part of a thing, in a really short space of time.
Raymond: No, absolutely. No, no. I agree, no question. It has been rapid but I think that this at times when I go to conferences or seminars and you hear this is where we’re going and it’s like, yes, but still fully two-thirds of global compute resides in a privately owned on-premise corporate own datacenter that the people who own the facility also on all the computers and all the staff touching it, it’s still… I say that to say the upside of the cloud is tremendous. I get asked this one a lot and I’d love to get your thoughts on this.
Peter, I’ll occasionally get asked, “Well, Raymond, you’re in the datacenter development business, aren’t you worried about as compute functions get miniaturized as they shrink, as they get smaller, that the people and they get more efficient, and they use less power, that the need for data centers is gonna shrink?” And I love getting asked that question. And Peter, for me, I always ask people, I’m like, look, let’s think about the three biggest changes in the last three decades around compute that have allowed us to shrink the technology footprint on a datacenter floor. Those three things are, they have to be virtualization, right? They have to be the fact that we could put multiple cores on a single chip, multiple cores on a single processor, and then the change from spinning platters on disks to flash storage.
Those three changes had a massive impact on what physically got put on a data center floor and all of those have happened and none of those three changes are new. I mean, you know, I’m not a good enough student of technology history but all of those have been started more than a decade ago. What’s the next I mean, those have massive changes, people don’t remember that prior to virtualization and really virtualization because we were virtualizing mainframes in the ’60s.
But prior to virtualization in a generally available computer, computers averaged 17% utilization, averaged 17%. Today, we virtualized them, and in the well-run enterprise data centers, they average 92%, 93% utilization. We fundamentally changed the footprint and the workload in the compute with virtualization, then we started cramming multiple processors or making the boxes smaller, then we took all the storage and said you don’t have to put it on big heavy spindles, you can put it on flash drives. I mean, those of all already happened.
Peter: Well, that’s another storage one hasn’t happened yet.
Raymond: What’s that one?
Peter: The storage one. I think there’s still a majority on the spinning disk, but it’s obviously going to, isn’t it?
Raymond: It’s going that way, that’s my point. Yes, yes, you are just right, it’s not replaced spinning drives everywhere but we have the technology now and that migration is happening now.
Peter: Yeah, that’s absolutely. And it’s the same really as the rate at which the cloud is taking over or not taking over, isn’t it? I mean, the cloud does what it does because of virtualization, that’s the fundamental thing under it. And so that, you know, automatically, you can spin up your new server in the cloud, you can spin up as many servers as you like, and in a way, if they were physical servers in your own data center and you’d be way below that 17% of utilization but because they’re in a centralized cloud resource, that cloud resource is managing utilization for you and making sure that you’re not wasting space. So all that happens and it does take over slowly and again without us realizing.
Raymond: Yeah, exactly right. Yes.
Peter: I wonder if the process of moving from a physical data centers into the cloud might have a slight hiccup this year because of the pandemic because decommissioning of physical data center is a process that involves people going somewhere and doing physical work. I think a lot of physical migration projects may well get put off, the cloud is doing fine, it’s growing hugely because of the increasing use of its services. But the specific case of migrating things from enterprise datacenters into the cloud, there’s probably a lot of projects there that have been delayed.
Raymond: Well, I think you’re 100% right that the ability to physically go touch things and migrate and do the services that someone has to go pick it up and move it has slowed. But the tsunami of momentum behind the cloud and I know I said it’s only a third, but to your point, it’s a third of an incredibly big business and an incredibly big business that’s growing at lightning speed. And that’s the thing that I think when I get asked when people say, “Well, Raymond as we miniaturize compute function, and as we get them more efficient, aren’t there gonna be fewer datacenters?” And I ask people I say another analogy I give people Peter and see if it resonates with you.
I’m in the warehouse business for ones and zeros, I warehouse ones and zeros, tell me what warehousing business has the total global inventory of the product they store double every year, globally, double every year? But that’s my business, right? My business, the number of ones and zeros on the planet is doubling about every 12 to 15 months and I’m in the business of warehousing those ones and zeros and that tidal wave, tsunami of momentum behind the cloud and behind our friends at Google and Amazon and Microsoft, that’s what’s fueling the rapid growth for them even is there only at a third of global compute, that backdrop of global computing rising it’s such an incredible pace because the expansion of ones and zeros when nobody is deleting any other data and the digital revolution is giving us new ways to create data, generate data, analyze data, copy data, think about data, all of that stuff. I mean, we not even get off into the buzzwords of AI, machine learning and all that. I mean, just the pace of digital change is accelerating at this point in my life, which I’m surprised now three decades into being a part of it.
Peter: Absolutely. You say that people come to you and they say, datacenters are getting more efficient, why does that not mean we need fewer of them? The answer to that question is exactly the reverse, we will use vastly more data centers because they’re getting more efficient.
Raymond: Exactly right, Peter. That’s right. It’s sort of counterintuitive, you’re right.
Peter: It’s been known about for like, oh, nearly 250 years. William Stanley Jevons was this Victorian economist. I don’t know if you’ve heard of the Jevons paradox.
Raymond: I’m not familiar, this will be a whole another episode, the Jevons paradox.
Peter: People said he did a mathematical theory of economics and people said to him, “If we make our steam engines more efficient, that will save coal.” And he said, no, it won’t because if you make steam engines more efficient, steam power will become so much more cheap that people will use it for more things.
Peter: And that’s the way things happen in the datacenters because things might become cheaper and cheaper and you get more and more computing for a given amount of kilowatt-hours of electrical energy because you get more computing from it you can effectively make all your services free and the amount of capacity you’ve got can however much it is to almost infinite level can be blotted out by things people are doing for farm, by kitten videos by anything that it pleases them to do.
Raymond: The Jevons paradox, did I get that right?
Raymond: Jevons paradox. Okay, folks at the top of the show…
Raymond: Yeah, J-E-V-O-N-S. At the top of the show, I told you that we were gonna have Cambridge-educated physicists and that, you know, distinguished alumni such as John Maynard Keynes, and Peter pulls out the Jevons paradox. So guys, if you needed any demonstration of how much smarter Peter is than I am there, you just had it you just had to dial in from listen for almost an hour. We appreciate that. Yeah, you’re exactly right. You say it much better it yes, the Jevons paradox is the fact that we’re getting so much more efficient means we’re gonna figure out more things to do in the cloud, that’s exactly right, that’s exactly right.
Peter: If you build another lane on your highway it doesn’t make the traffic go faster, it makes more people travel to town.
Raymond: That’s right, it makes more folks come to town. Amen. That’s exactly right, that’s exactly what I’m saying. And I just didn’t even know what the Jevons paradox was but I agree with it and that’s when folks talk to us about the future of this business, I go, guys, it’s growing at a pace that we just don’t even understand it. We can’t comprehend the new things that will be done and the pace at which people will convert their workloads from one location to another. I mean, our industry and I’m blessed to be in an industry that in a global pandemic is continuing to grow but I think that growth is only going to accelerate as what we do helps solve the world’s problems like how do we produce a vaccine and how do we track the virus and how do we ensure that we know how to keep people safe, those things are gonna get done in the cloud. And that’s what you write about every day and we try to provide every day.
Peter: Yeah, that’s the point we missed earlier when we were talking about these other good things of collaborative research, the fact that we’re getting a vaccine for the Coronavirus in maybe one year compared with the normal time in eight years that’s all down to accelerated communications.
Raymond: That’s right. The ability to communicate, collaborate and computate, right? To calculate faster than we’ve ever been able to before so just feel fortunate I get asked a lot by my degrees in finance and I get asked a lot, “Well, Raymond, did you study technology or computer science or IT or IS in school?” I was like, no, I stumbled into this business. It was a pure accident and been lucky to hang around it for three decades.
So well, Peter, I’ve thoroughly enjoyed our time together, we have been able to record over an hour, and I’m grateful for folks that have been patient enough to listen, I think we’ll cut this into two different episodes. But just been a joy to talk to you and grateful to get to meet you virtually and hope to meet you in person someday. London, I will just tell you is my favorite city in the world, every chance I get to go to London, I go and I look forward to getting to meet you at some point in person. This has just been a joy to get to hear about you, your history, your experience, your wisdom around the technology business, and how it impacts the data center business and how you talk about the data center business. So thank you for being willing to join me on this podcast.
Peter: Well, thank you, Raymond. It’s been a real pleasure talking to you.
Raymond: Peter, thank you so much. Have a lovely weekend and enjoy more fresh berries from the garden.
Peter: I certainly will.
Raymond: Thank you, my friend. Bye now.
Peter: Thank you. Bye.