Oracle Cloud Infrastructure SVP shares competitive strategy

SAN FRANCISCO -- Oracle's late entry to the public cloud space has been met with skepticism, but the company has a few strategies, particularly around autonomous cloud, unexpected partnerships and shared responsibility, that executives expect to make up for its tardiness.

Oracle Cloud Infrastructure (OCI), the company's second attempt at a public IaaS, was built with the help of former AWS technical employees like Clay Magouyrk, who joined Oracle in 2014.

Now senior vice president of engineering for OCI, Magouyrk spoke with TechTarget during Oracle's OpenWorld conference about the company's market position, its efforts to attract customers and what the future may hold.

A common refrain is that Oracle is very late -- maybe too late -- to arrive at a viable IaaS strategy. What's your response to this?

Clay Magouyrk: I think you have to understand where we're at in our evolution of the cloud. I remember when everybody had a kind of, you know, a dumb phone. And I remember it took two years to get to a place where everyone had a smartphone. The switching speed was amazing.

So then the question is, [the industry has] been doing cloud infrastructure for 15 years. Why isn't it all [migrated over from on-premises] yet? It's still 90% greenfield. One of the things I worked on at Amazon before I left was the Fire Phone, and the Fire Phone was too late, in the same way that the Windows Phone reboot was too late. Once Apple and Android had 90%-plus market share it was impossible [to catch up].

The reason I spend my day here every day working really hard is because we're still at the 10% penetration [level]. People act like we're 90% of the way into the cloud infrastructure transition, but we're not anywhere close to it.

One of the biggest OCI announcements at OpenWorld was Maximum Security , which you positioned as a response to data breaches at IaaS providers caused by misconfigured systems. But it doesn't fully reject the shared responsibility stance around security forwarded by AWS and others.

Magouyrk: I think, fundamentally, in the long term, you always end up with some shared responsibility. What we're trying to say here at Oracle is that we are going to push that boundary.

If you look at cloud in general, cloud providers have the responsibility of patching the hypervisor. And then, okay, your job is to patch your OS. Cool, well, with Autonomous Linux, now it's our job to patch the OS. So we're bringing that up the stack.

We're going to create a construct where you can't misconfigure it. We're taking the stance that this is not just your responsibility, we are going to work with you on it.

If you look at the vast majority of security incidents these days, they're not [committed by] massively sophisticated hackers. It's misconfiguration. When you moved to the cloud, we gave you all these tools. And now you've got a million tools. And you have this programmable infrastructure. It's so easy to do stuff, right?

We're going to create a construct where you can't misconfigure it. We're taking the stance that this is not just your responsibility, we are going to work with you on it.

Think about it from our perspective on SaaS. We take on a ton of responsibility for that. There's no way for you to mess it up. As we can make infrastructure to where you can't mess it up, I don't see any reason why we can't take on more responsibility the same way we do in SaaS. The problem is that if people still need that control, to be able to mess it up, then it has to be on them.

You to dramatically expand OCI's global footprint, bringing it to 36 regions by the end of 2020. Tell us about Oracle's process here.

Magouyrk: There are things under the hood that people don't see. One of those things is that we've massively optimized how we build regions, both from an infrastructure perspective, as well as a software perspective.

When I joined Oracle in 2014, we had zero OCI regions. I knew that if we were going to compete, we had to build regions that are way faster than our competitors. I knew how [AWS] built them, because I had worked there. And we hired people that also worked there. What we did is we took a much more aggressive approach.

The way most companies build these regions is you have 200 teams, and they all do it by hand. So the physical hardware gets rolled in, and the actual software teams spend all their time installing the stuff.

When you bought a Windows CD back in the day, Microsoft didn't send an engineer to install it for you, right? They had an installer. Well, you can do that for the cloud; it's a bunch of work. But we've made that investment from a software perspective.

If you look back to when Amazon got into this, they didn't know where they were going. It was very early days. And they were like, we're going to build these big honkin' regions, and they're going to cost a jillion dollars and there is going to be so much stuff there.

We realized, by the time I came to Oracle, we were going to need to put regions everywhere, but they're going to start small and be able to grow. We've engineered for that. From a technical perspective, that gives us a lot more flexibility than some of our competitors.

It's my understanding that Oracle is doing this buildout largely through co-location agreements and not actually building its own data centers, in the interest of speed. Can you elaborate?

Magouyrk: I think the term [co-location] has changed. If you go back 10 or 15 years, co-lo meant you get one rack in a small data center, that kind of a thing. But as you actually see these large cloud providers -- the way they're doing these worldwide rollouts -- co-lo providers have actually changed their model. They still do that kind of retail stuff. But they also have a wholesale model. And that's what has enabled a lot of this rapid global expansion.

Every major cloud provider uses a ton of colocation facilities. There's just no way that it makes sense otherwise. As a co-lo provider, you can do things where you put stuff in a campus. You buy a bunch of , and then you start small. But then, as that fills up, you build buildings and you can deeply interconnect that with fiber. As a [co-lo] customer, you have this expansion in that same , but you don't have to build a giant data center [yourself].

One OCI announcement this week at OpenWorld concerned a "always free" tier, which seemed clearly aimed at attracting developers to the platform. Tell us more about your efforts in this .

Magouyrk: A lot of it is social proof. While we would like to think that everyone's making a fully informed, highly educated decision, the reality is that most humans do close approximations based on what everyone else does. The key is to create a flywheel that gets enough going, and then it becomes oh, well, they all chose Oracle, why can't we?

We have dedicated startup investments where we go to startups and give them a bunch of free credits and incentives. Around developers in general, we put a lot of energy into attracting people out of college. We work with universities to give them access to cloud computing resources. They use that in their classrooms to get people familiar with it.

I view [the free tier] as the start of something … Because it's not just making it free, it's making the signup easy, the support experience easy. Do you have the collateral and the ecosystem around it? Do you have the right forums for people to ask questions? You do all of those things in a row, and it builds.

The initial use case you put forward with your interoperability partnership between Microsoft Azure and OCI was to split an application up, with the logic and presentation tiers on Azure tied back to an Oracle database running on Exadata inside OCI. Why would a customer want to do that?

Magouyrk: When [Oracle CTO and chairman] Larry [Ellison] gets up there, and he's like, look, let me show you the performance difference between what you can get from this versus an Exadata, I don't think people actually believe it, but it's true. There are all these amazing workloads that need giant relational databases that just can't move anywhere.

My job is to get you in OCI and then just keep pulling more. Maybe you add some Analytics Cloud on top of it to analyze that data. You just get people hooked that way.

What about going further and bringing Exadata boxes right into Azure data centers? That would eliminate the need for the high-speed interconnect between OCI sites and Microsoft's.

Magouyrk: As you can imagine, those types of conversations do happen in the abstract. I'm sure Microsoft would love that. The thing you have to understand is that we did a lot of work in Oracle Cloud infrastructure to make Exadata run well. We offer bare-metal; we offer off-the-box virtualized networking. There's a whole bunch of features.

Let's say that we were to make a deal and I gave Microsoft a bunch of Exadatas. It would be off in a little tiny part of the network. It's not actually integrated into their experience. They wouldn't have a database service wrapped around it. The experience would be terrible. For them, it's not important [to host Exadatas].

For us, it's so valuable that we make it really, really good. What we're not going to do is take the thing that we think is incredibly valuable and then have other people do it badly.

What is next for this deal? Will we see others like it, say with Google? Going even further, is a détente between Oracle and AWS possible?

Magouyrk: Right now, we have great ideas, and we have good buzz, and we have very interesting customers. What we have to do over the next six months is convert those into a bunch of very happy, very public reference customers. That's the next level of uptick in the process. Until we have that play out, I don't think anyone's going to know.

In terms of where we're taking this with Microsoft, I think it's about us working much better together. We're making Oracle Linux work better in [Microsoft's] cloud. A big part of the reason we chose them is because we have such customer overlap. It might be interesting, technically, for us to do it with Google. But it's not like Google is already in every single one of our customer accounts. We're doing this from a customer-driven perspective.