Maybe you’re old enough to remember the dot-com boom when big companies, green with startup envy, wrapped themselves in webbyness as they watched, say, Webvan receive a valuation of $4.8 billion.
Something similar is happening now with digital transformation, as a fresh wave of high-flying digital natives such as Lyft, Snap, Spotify, and dozens of others crank up corporate fears of missing out. And in many cases, enterprises are emulating those startups’ decision to build on the public cloud.
Last week, I interviewed Google Cloud Platform Vice President Brian Stevens, who highlighted the degree to which dreams of cloud transformation have inspired the C-suite: “The CEO, in many cases, is driving cloud strategy.”
That can sound a little terrifying to IT. A wholesale lift and shift of legacy workloads to the cloud is not realistic. Transformation never occurs all at once -- at least not without disaster. In reality, says Stevens, such migrations don’t happen overnight.
Our conversation began by addressing the large gap between legacy enterprise technology and all the cool new stuff in the cloud, from containers to machine learning APIs. (Note that this interview has been edited for length and clarity.)
InfoWorld:You worked at Red Hat, so you have a good sense of what legacy enterprise infrastructure looks like. What’s Google’s message to companies that want to cross the chasm to the cloud? How do you help them?
Brian Stevens:It hasn’t been a one size fits all. In some cases there are different groups that are not even on the same page in the CIO’s office. I’ve never seen two [customers] take the same path.
InfoWorld:You can’t even talk about common patterns?
Stevens:No. I’d say a year ago, it was more workload-based. Now what’s happening is it might start out workload-based, but I’m [also] going to do DR or I’m going to do some analytics or I’m going to do some HPC kind of thing. It’s augmented now by the CEO being more involved in it and more aggressive-looking shifts.
InfoWorld:Yeah. The CEO says, “Move to the cloud!” And then IT goes, “Oh my God ...”
Stevens:"... we don’t know how to do it!” Exactly. That’s what I was saying. When people visit us in our briefing center, 90 percent of the time the CEO is with them, even at the largest enterprises. The CEO, in many cases, is driving cloud strategy. This is the future of this business.
It’s around their security, their reliability, their cost, and what kind of future businesses they’re going to create from a digital business perspective. That’s the CEO’s job. It’s no longer the point group’s job.
In some cases, we might be helping people deal with machine learning for fraud detection for credit cards. OK, we can do that in eight weeks. In other cases it might be: We have eight data centers, we’re going to go down to two. The rest are going to shift to cloud. Everyone is kind of different.
InfoWorld:Are you actually seeing the migration of legacy applications?
Stevens:Yeah. They don’t happen fast. First, you have to earn trust. Are we committed? What’s our stack? What’s our roadmap? What’s our vision? Once you get past the trust phase, all of a sudden you have a proof of concept. And after that, agreements.
The third piece is all the containerization that’s happening. That’s happening fast. That’s also been a driver for cloud because they’re redoing their packaging for on-premise, because it’s useful to have. As part of that, it’s letting them move their whole toolchain and test environment to cloud, which is often 50 percent [of usage]. Greater than 50 percent of VMware’s usage is just test and dev. All of a sudden ... we’re moving to Docker for that because it’s lightweight. Containerized test and dev is often one of the first journeys -- and that’s great for enterprises because that’s not mission-critical.
InfoWorld:Virtualization may have been the most quickly adopted enterprise technologies ever.
Stevens:Absolutely. It just solved such a huge problem. It was a no-brainer.
InfoWorld:One of the ways VMware did that was to provide a very clear path in terms of tools.
Stevens:Diane [Greene, senior vice president at Google and former VMware CEO,] and I talk about that all the time. She sees that, the whole physical-to-virtual aspect of that. Just set up the clear value proposition with tools to actually allow you to migrate.
InfoWorld:What’s the analog with the cloud?
Stevens:The analog is a little more complex. We’re building a partner ecosystem, because there are lots of mini-experts around various areas of migration tools. We have a strong partnership with a company called CloudEndure -- they handle a lot of the transcoding aspects from VM to the cloud and then set that up in DR configurations. There are other companies that are good at providing database replication -- some people want to still keep the primary database on-prem and want to have a replica in the cloud. Others want virtualized data because they want active/active on both sides. There’s no one silver bullet.
InfoWorld:Google has been reaching out more and more to the enterprise. How far does that go? As far as providing on-prem tools beyond Kubernetes to create a hybrid cloud between enterprise customers and Google?
Stevens:Right now all the on-premise stuff is through partnerships. There are a lot of companies out there whose whole business model is on-premise. They’re the experts.
InfoWorld:Companies like the one you used to work for, Red Hat?
Stevens:Yeah. Why wouldn’t we work with them? We [work together] as opposed to saying we have to own the on-premise world as well. I’m not sure I want to go back to that world of managing software on-premise.
InfoWorld:Well, you have Kubernetes to work with.
Stevens:That’s why that strategy works so well. We don’t want to monetize but we do go in and teach. If we had to own Kubernetes, we’d be competing with all of our partners. [When I was at Red Hat] they weren’t talking to Red Hat about cloud. They were talking to Red Hat about on-premise. It’s the same way now. It’s the same set of customers. Now what I’m trying to do is join those conversations together with a number of companies where we actually go in together and realize that we both win.
My developer relationship team actually goes inside of enterprise and just teaches them Kubernetes. We’ll go in and spend a day with 30 developers, and by the end of that day they become experts in deploying containers in Kubernetes whether they want to use it internally or on-premise.
InfoWorld:Do you see containers as complementary to VMs?
Stevens:They’re entirely complementary. For us, as an example, our container platform runs on top of our VM platform. You could argue that down the road, that when you’re using containers, you probably don’t need VMs.
InfoWorld:What do you see as the future of VMs, and do you care?
Stevens:I think they’re probably going to be around forever. Go into Citigroup and you’ll see a bunch of mainframes [as well as] distributed systems and x86 ... you’ll see both. VMs are always going to be around but I think that when you try to use them as the methodology for portability, like at a cross-cloud level, they fail. Here’s an example: Amazon uses VMs and they transcode the VMs from VMware with vCenter and then move them to the Amazon Cloud. That’s kind of a one-way street. You can’t come back. It’s kind of wasteful.
InfoWorld:How realistic is workload portability? Applications tend to stay where you built them, because you have all these integrations with special services and so on.
Stevens:Even when you start at the Docker level, even then [portability] is not perfect, it’s better. I’d argue that even with the early work in containers, it’s far easier to be version-independent of the OS and the underlying platform. In some cases you’re 100 percent there. In other cases you’re 80 percent there. That’s far better than the world that I used to live in.
The side effect, though, is that a lot more dependencies find their way inside of the containers themselves -- so you still get software stuck inside a container, very different dependencies. You have to manage those. It’s not like this perfect nirvana where it’s just the app stack. It’s still the top part of the OS that’s living along with them. I call it a 100x improvement, even though it’s not perfect.
I really feel like the industry is not great at understanding application dependencies and services. If we can get to a lingua franca about how we represent things: Here’s what an enterprise application looks like, here are the components, here’s how we specify those, here’s how I wire them together, here are the configuration parameters.
I think you sort of represent that in ... call it a JSON file combined with a bunch of microservice components. Then all of a sudden you get to a point where even for IT, even if they’re not doing portability, they’re doing repeated deployments that are predictable because all of a sudden it’s not a manual process. You’ve got it cold.
InfoWorld:When you put applications on the cloud, particularly a cloud that’s container-native, you want to break it down into a microservices architecture, don’t you?
Stevens:I’d say a minority are doing that, though. Even though Google Container Engine is our fastest-growing platform right now, a minority of a large enterprises coming in [are using it]. Predominately they’re going to bring their existing database, but they’re going to bridge their database. In some Oracle cases, they’re bringing it. In some cases they’re getting rid of Oracle and going to another SQL compliant database, with the IT stacks that they have on top. A lot of lift-and-shift stuff is more traditional.
It really becomes a network configuration issue, and because this is how we manage on-premise, we want the same sort of controls on how we manage the firewall and access control. We have far more conversations around whether they have all the controls and knobs to manage the cloud than around whether I build that app stack in the cloud. The app stack is kind of easy.
InfoWorld:One last question. When you start to move huge portions of enterprises into the cloud, then you really start running into enterprise architecture. You start running into some of the same problems from the SOA era.
Stevens:It’s a whole new world. The future architects are going to be experts at this. They’re just starting to exist.
InfoWorld:You’ll need new tools to manage it.
Stevens:What’s amazing is that because there’s such an opportunity, you’re going to see all that stuff coming online. It’s creating a whole new set of companies around that tech. It’s fascinating.