Originally published on CloudOps’ blog.
Ian Rae, CEO and Founder of CloudOps, made a guest appearance on the2030.cloud podcast. He spoke with Rob Hirschfeld, CEO and co-founder of RackN, about how we can build and manage complexity. You can listen to the full podcast here.
Rob Hirschfeld: I’m your host, Rob Hirschfeld, CEO and co-founder of RackN. Welcome to another episode of the Cloud2030 podcast. This was a recording we did with Ian Rae of CloudOps, one of my long-term friends.
During this podcast, we will get into the substance of this issue: how economics are driving us faster and deeper into the cloud while colliding with the challenges of customers and operators owning and influencing how their infrastructure is built. Ian, can you give us a little background?
Ian Rae: My name’s Ian Rae. I’ve been in the tech industry since the mid-90s, where I got really obsessed with building infrastructure for the internet. That led me to get involved in tech startups, of which I participated in several large and small. I finally ended up starting a cloud services company called CloudOps, focusing on helping customers not just chart their way into the cloud but really succeed in the cloud. That has been our core focus for over a decade now.
Rob Hirschfeld: You have a unique perspective because of your experience helping companies run infrastructure. You and I met during the OpenStack heyday when there was a vision of an open alternative where a bunch of small providers could collaborate together and become a community cloud from that perspective. It’s not quite how things have worked out, but where is the cloud team right now?
Ian Rae: It’s such an exciting and interesting time. I spent the first ten years of my cloud career advocating for the cloud. Until a few years ago, folks were still saying it’s not secure and it’s too expensive. In the past year, those reasons have melted. For the first time, everyone believes the cloud is the way to go. We’re starting to have a lot more thoughtful conversations with customers about how not just to get to the cloud but how to succeed in the cloud.
Rob Hirschfeld: Is the cloud a destination or a way of doing business? If you say the cloud is Amazon or a place, then we will all just end up in a provider. I tend to think of it as a new way of computing, where we’re learning how to use infrastructure in a much more dynamic way than before.
Ian Rae: What captured me about cloud was this notion that I could grab a bunch of storage when I needed it to try some stuff, and then I could throw it away, or I could get more. Cloud gave me the right to be wrong about what my needs were. It turned out that I was always wrong, and it took me a while to realize how often I was wrong. It was eventually empowering to realize that I could be wrong, and I could iterate, and I could eventually be right.
Originally, I viewed it as a tech problem. I then started to realize it’s an operations problem, before realizing it’s about a business model being supported by an operating model. That brought me to my current state of mind, which is to view it as a culture. At the end of the day, I do think the cloud is a modern way of delivering value in this digital universe we’ve been racing into.
Rob Hirschfeld: How mature are we on this curve?
Ian Rae: It’s sometimes hard to answer that question within a contextual understanding of our history. I do think the cloud as a way of doing things is pretty mature. That said, I will quote the late, great, Canadian author, William Gibson, who said “the future is already here — it’s just not evenly distributed.” We see enormous gaps in the market between those who are doing cloud really well and those hardly doing it at all.
Rob Hirschfeld: It seems as though you’re describing an inflection point, where we have those who are ahead and those catching up. What tends to happen in inflection point markets is one of two scenarios. Either people figure out how to do it, they catch up quickly because it becomes commoditized, and their competitive advantage disappears. Or the advantage becomes an insurmountable obstacle, and the IP and the value becomes owned by industry leaders no one else can catch up to. There’s an idea that there can only be three major clouds because nobody can catch up on the investment. Amazon, for example, is so far ahead that nobody could ever compete as they would have to buy too many servers. Do you see that same dynamic?
Ian Rae: What you’re referring to is the winner-take-all dynamic. Yes, we often see economies of scale in the tech world. But there are other, significant factors, such as data gravity. There is a fascinating tug of war happening right now between the really big, multinational hyperscalers, and the jurisdictions. Across the world, there is an incredible variation in how people govern themselves, live, build, and operate.
There’s no question in my mind that the key hyperscalers have incredible momentum and advantages when it comes to the key platform capabilities. In many cases, those advantages are passed on to the market. There are some healthy dynamics in the market right now.
We’re also at the stage where folks are realizing how critical operations are in a digital world, not only to delivering the key value of your product but also to create the exhaust of the operations of scale that are leading to new generations of innovation, IP, and capability. Admittedly, I see the world through a bit of an operational lens — my company is called CloudOps, that’s what we do.
One of the things we’ve seen a lot of recently is the distinction between creating and operating. Canada is great at getting things started, especially when it comes to science and thought leadership. An enormous amount of AI research was actually generated in Canada. But we’re not in a great position as a country to operationalize and deliver those capabilities.
There’s a lot of discussion as to why that’s the case, but I think we can figure that out over time. It’s more about the missed opportunity. Once you start being able to operate at scale, you generate insight that isn’t just customer insight but also the insight that comes from operations. That’s an enormous amount of data, the data exhaust, that becomes part of a virtuous feedback loop that allows for continued momentum growth and success.
It’s not clear how new entrants can come in and become competitive at that scale, but the counterbalancing force remains the uniqueness of the jurisdictions and the increasing realization that both supply chains and the interdependence of jurisdictions are important.
There is an opportunity right now in the open computing movement that we are a part of. Currently, that journey is focused on the ability to build edge capabilities in far-out regions across a diversity of industries. The tension we’re seeing right now is because the hyperscalers are very clearly targeting all the way to the edge.
Rob Hirschfeld: What you’re describing is something I see a fair amount of hand-wringing about. This exhaust data, this inherent advantaged scale, is an intractable problem. With the increasing SaaSification of business, operating tools are being run by third parties. From a social perspective, it’s an amazingly low cost to entry. It’s also terrifying because you now rent instead of own the IP that is critical to your business’ success. Farmers, for example, often have to lease their seeds, which they can’t own because they’re genetically modified. Now, we’re all renting the seeds of whatever our business is.
The business model empowered by cloud, the shift from CapEx to OpEx, is both scary and irresistible. Especially in the US, in the past few weeks, we saw what can happen when all your data is monetized by somebody else and fed back to you. From that perspective, there’s a hefty price to all the free services we get. You don’t always know who’s monetizing your information.
We have decided it’s too hard to operate things, and the SaaSification of business is underscoring that decision. It’s almost as though someone is taking all the cars and making them intentionally harder to drive, so you can only drive if you’re licensed like a London cabbie. Do you feel that’s the direction things have been going in?
Ian Rae: We probably are heading in that direction right now with cars, as they are becoming harder and harder to maintain and take part in. Cars are becoming self-driving, with proprietary algorithms. This is absolutely the defining trend. That said, there is a backflow focused on the distinction between differentiated and undifferentiated value.
If you have a business, your undifferentiated value is what you must be able to operate in order to deliver your value proposition but that isn’t part of what makes you special. SaaS is a great way of solving for that undifferentiated value, without dealing with the hassle. That doesn’t mean you don’t have a backup plan, a way of mitigating risk in case you are no longer able to find that value with your chosen path. Organizations of different sizes will have more or less sophisticated ways of mitigating risk.
The conversation around your differentiated heavy lifting is more interesting, as it provides your just cause. Why do I need to be able to build edge compute on my own terms, where I can actually build, maintain, and even enhance the technology being used? It depends. Maybe I operate farms, and I’ve been building a system of farming that’s highly automated and requires a lot of edge technology. Maybe there’s a reason I don’t want to be dependent on some multinational company for a key aspect of my value chain. Coming back to your point, who even has enough scale or capability to make that decision and follow through successfully on building their own piece? That’s a whole other issue. Looking in 2021 towards 2030, it’s about differentiation. I don’t know what the answer is.
Rob Hirschfeld: When you say differentiation, and I hear this a lot, I can’t decide if I’m horrified or in agreement. You don’t want to do things that don’t add value directly to your business, right. I see that used a lot as justification to give up expertise. Nobody’s going to run a data center better than Amazon, so I should stop learning and building expertise. Vaccines are a good example. We’ve got all the scientists who can figure out vaccines, but the manufacturing is different (India makes about 80% of all vaccines in the world).
Ten years ago, I would have thought we would today have assembly lines of 3D printers that would bring manufacturing back onshore. We were expecting this revolution of modular manufacturing. It didn’t work out that way.
From an IT perspective, right now it looks like we will have hyper-focused people who do Ops and only Ops, and that we will give up the manufacturing. But the global systems are saying, we need to keep the skills, the crafts, and the ability to do things ourselves. What’s the consequence if we don’t understand how this stuff is built? There’s a separate problem to me, which is we’re not investing in making it easier to do.
Ian Rae: Yeah, I think this is the problem of our time. Again, there’s a lot to be learned from those supply chain dependencies. There may be reasons, for example, why we need to invest in certain capabilities, even if it is going to be harder or we’re going to move slower. Since it costs us more, there may be reasons why we have to do that. Being caught out with a lack of manufacturing capacity in a global crisis is a cost that wasn’t factored into the overall cost of manufacturing argument. This happens time and time again. We saw it in the early days of outsourcing software development. You’d make a decision based on the cost of a model unit economics.
Rob Hirschfeld: Then it would fail to factor in some really important other factors. Going back to your data center analogy, Amazon’s going to be the best there is. Alistair Croll resolved that the public cloud would own 100% of the market at some point in the future. That was the reductio ad absurdum of cloud, that it would take over everything and nobody would have their own infrastructure. I always felt that there would be a mixture, a balance. I think the more sophisticated argument asks whether it is really balanced, or if the pendulum is swinging to an extreme? That makes us a bit nervous. If you swing too far, you might lose the capability.
Not only do you compromise your ability, but you also stop asking the vendors to provide it. If you are always driving towards efficiencies (governments aren’t particularly efficient), you end up with mega-corporations that measure success in terms of what is most cost-effective.
Ian Rae: I think there’s a bit of a commoditization myth that popped up in the early days of cloud computing. The analysts were saying it’s a race to the bottom, and Amazon’s insane doing AWS. In the first five years of my cloud career, I was trying to argue against analysts who were all saying anyone would be insane to do this stuff. Now, people realize it’s the opposite of insane.
What’s interesting to me is that it’s not just the market capture potential of these business models but also the margin capture. We haven’t yet seen the economic efficiency from true competition passed on to the market. I think that’s partly because the market is really inefficient operationally. Even if I can buy it myself at one-tenth of the price, I’m bad at taking advantage of it on my own terms, so I’m just going to pay that price.
As someone who is quite comfortable relying on cloud services, the question comes down to the future and destiny. If I make this decision now, what’s the likelihood that in ten or twenty years I will continue to pay the toll booth on this roadway that was built and sponsored by a private corporation as opposed to with tax dollars.
I’m not suggesting that governments should be building cloud — they’re not exactly the most efficient and I don’t think taxpayers would get good value for their money there. But we could end up in a world where we are unable to build any of the infrastructures we need for a digital world. All we are going to do is pay rent in perpetuity. My question is whether that rent will be efficient or inefficient. Will there be an ability to achieve better value for the dollar or will we find ourselves in entrenched positions?
This is a common problem in real estate, where you own this plot of land or this building, and you can charge what the market will bear. The dynamics are a little bit different in real estate, but you will have the ability to continue to not have to compete in a commodity race to the bottom. I find it interesting how much the analysts got that so wrong. That there was going to be so much competition at such a great scale. My cost would drop over time, and I would never really have to worry about it. Yes, I have to pay rents in perpetuity but that’s far better than me trying to do it myself.
Rob Hirschfeld: This is the difference between hardware and software. There’s the idea that you wouldn’t write it yourself, you wouldn’t build it yourself. The question that keeps coming back to my mind is whether there is a turning point where the major cloud providers stop charging you per widget of work. Really, they’re selling building blocks.
Maybe there was a turning point with some of the things that happened with Parler, which they shut down. But today, they haven’t differentiated any of their customers necessarily based on the use case. They’re not charging you more for a server if you are something that competes with them or that they don’t like. It’s commodity pricing from their perspective, and I don’t see that changing.
It’s been amazingly powerful to have access, and part of the story here is the demand. Going back to what the analysts missed, the demand for computing, AI, machine learning, data storage, and bandwidth has only gone up. A big part of the story of the cloud is not that they were better at it. They were faster at it, and they were lower-friction at it. Ten years ago, companies were getting in the way of their own internal innovation cycles. People would go to the cloud and bypass resource limitations. Looking forward, that won’t change but will become more accelerated.
I want to get your ten-year impression on this, if we’re going into the edge and edge is going to be an acceleration of demand for computation. Does that only push us faster and faster into these hyperscalers?
Ian Rae: It’s such a good question. I will likely be wrong about this. That itself relates to my key point, which is that I liked the cloud because I realized how often I was wrong and therefore needed agility. I get to be involved in both how the cloud is consumed and how it’s built. I feel very lucky to see it from both of those angles. I do think there’s some convergence happening. I want to pick on something you said, which is the key difference between hardware and software.
If you look at where the different hyperscalers are coming from, they’re all coming from different directions. Amazon did build from the ground up, from the boring commodity servers as you say. But they are absolutely working their way northbound into software, and they are offering more and more platform and industry-specific solutions.
Even though that server isn’t sold at different prices to different industries, they are now starting to sell industry-specific solutions. They’re working their way up the stack because it makes sense. It’s what the customer wants and is willing to pay a much higher margin for. That’s the irrefutable trend, in my view. If you look at where Microsoft and Google are coming from, they’re coming from a much more software-centric place and are working their way down to delivering infrastructure. Yes, we could end up in a world where it’s winner takes all.
But there is a natural and healthy tension between the hyperscalers that will keep the market somewhat interesting. When I say interesting, it’s both good and bad, because the enterprise customers I work with don’t choose the cloud, the cloud chooses them. They still think their architects and procurement can select this cloud provider. But what’s actually happening is that their lines of business and the people in their organization are buying cloud from all the clouds, and then they end up with all the clouds. That is a huge dynamic that has shifted. And when we go into an enterprise to help them out, we just assume that’s already happened. It’s a good thing because it makes for a healthy market, and it’s a bad thing because the paradox of choice is bewildering. Everybody charges different business models, so it’s a problem for customers as well.
Where I’m going with that is that the clouds will continue going up the stack to software, and they will try to sell all the bits and pieces of hardware. But we could end up in a world where increasingly we’re delivering hardware in a very particular spot, let’s say at a farm. I might buy commodity infrastructure from a supply chain that is a very competitive one. Maybe it’s from Supermicro, maybe I have to build it myself, maybe I’m using your software to help bootstrap and operationalize it.
What is really hard, much harder than buying and building your own infrastructure, is actually going in, getting, and building all the functions and capabilities yourself. Increasingly, the way we look at it, especially in industrial use cases, is that we’re going to be getting these raw machines that are able to run all kinds of stuff. And then we’re going to be Terraforming them, not in the HashiCorp sense, but in making them available for stuff to happen. The really key part is an orchestration that helps us go and grab the capabilities from the cloud that we need when we need them, and to have the optionality to either execute them locally or in the cloud.
My note of optimism, I suspect we are going to continue to head into an even more complex universe. I think the cloud is going to look much more like a biological ecosystem than it does a technological ecosystem. I think we’re going to have all the things, and it’s going to be painful for some folks but also a huge opportunity for creativity.
I might, as a technologist, hope I can still access the bits and pieces I need and not just from a hyperscaler. I can reserve the right to grab and execute the functions and capabilities that a hyperscaler might afford, where I choose based upon what I think is doing the right thing, reserving the right to change what I believe is doing the right thing. Let’s just say it’s an evolving target.
Rob Hirschfeld: There are echoes of what we’ve been talking about the Cloud2030. This complexity is increasing dramatically. You added some interesting spice here with the idea that Amazon, Microsoft, or Google, becoming software companies and suddenly needing to run their software, not on their own infrastructure. That’s going to be an interesting conundrum to watch.
I appreciate the time today. It’s been amazing to have you as a guest. Where can people find you?