Cloud Adoption and Infrastructure

As large organizations move to the cloud, what are key infrastructure issues facing Chief Information Officers and information technology organizations? To learn more, we speak with Clay Magouyrk, the Executive Vice President of Engineering for Oracle Cloud Infrastructure.


Sep 01, 2020

As large organizations move to the cloud, what are key infrastructure issues facing Chief Information Officers and information technology organizations? To learn more, we speak with Clay Magouyrk, the Executive Vice President of Engineering for Oracle Cloud Infrastructure.

In this conversation, he discusses Oracle Cloud@Customer, the Oracle Autonomous Database, cloud regions, and the Oracle Cloud Platform.

Clay Magouyrk is responsible for engineering and technical operations for Oracle’s enterprise-focused infrastructure-as-as-service (IaaS) platform and focuses on service delivery for an expanding product portfolio including compute, storage, networking, edge, and containers. He leads a team of thousands of cloud engineers focused on supporting customers’ most mission-critical workloads via best-of-breed products and solutions.


This transcript was lightly edited for length and clarity.

Clay Magouyrk: Cloud infrastructure, we're plumbers. No one wants to say, "Oh, my gosh. This is such amazing plumbing," but they're very unhappy when it doesn't work well.


Michael Krigsman: We're speaking about cloud adoption and infrastructure with Clay Magouyrk. He's the executive vice president of Oracle Cloud Infrastructure. Clay, tell us about your role at Oracle and the things that you're focused on.

Clay Magouyrk: I joined six years ago in the middle of 2014. The entire time I've been at Oracle, I've been very focused on solving some very fundamental problems for customers. Before I was at Oracle, I spent a bunch of time at Amazon, part of that time at Amazon Web Services.

I've always been an infrastructure person at heart. What's very clear to me is that cloud infrastructure is a massive enabler for the overall world, the industry, for developers specifically. But I also recognize, as many of your listeners recognize, as you recognize yourself, that while we have this amazing ability in cloud infrastructure, the vast majority of computing still happens on-premise in customer-owned data centers or colos. A lot of those advantages aren't available to most of the workloads customers use today. The big focus that I have when I came to Oracle was, how do we bring all of that amazing benefit that cloud infrastructure offers to everyone?

Cloud adoption in 2020

Michael Krigsman: Clay, you spoke about cloud adoption. You're in touch with a lot of Oracle customers, very large ones. Can you give us a snapshot of where does cloud adoption stand today among large companies?

Clay Magouyrk: Even as recent as seven or eight years ago, many companies weren't sure about the cloud. They weren't sure about maybe the security of the cloud. They didn't necessarily have a cloud strategy.

When I talk to pretty much every company today, everyone has a cloud strategy. They have a plan. But I think, at this point, customers have tasted and experienced the benefits of the self-service productivity you get from using cloud infrastructure, from being able to outsource not just the software but the operations of the service to where you can really focus on building your application rather than all of this other gunk that doesn't really want to be involved in.

While everyone that I talk to, and I think this is true throughout pretty much all industries, everyone wants to make the transition to public cloud. Everyone is struggling with that transition. You see that in the overall market penetration.

If you listen to different analysts, somewhere in the 10% to 15% range of all server-side computing has moved to public cloud. What that means is that 85% or 90% of the things that run all of these businesses are still happening in a customer's data center or they're happening in a customer's co-lo. That's why you see, I think, the major investments from all cloud providers in hybrid cloud, in edge computing, and a myriad of options that accept the reality that, hey, it's really hard to just pick that stuff up and move it somewhere else, so how do we attack that problem?

Obstacles to cloud adoption and migration

Michael Krigsman: Are these adoption challenges primarily technical in nature, is it a matter of economics, or is it a bundle together?

Clay Magouyrk: The single biggest one from my perspective is that the transition from existing customers premises to the cloud is not an easy transition, technologically. To put that in perspective, a lot of applications were not architected for cloud-native principles.

The cloud tells you what you should do is take your application, break it up into services, whether it's regular services or microservices, and so one of the ways that we're trying to attack that problem at Oracle is, well, how do I make the infrastructure more reliable? How do I make the services have the feature set that customers expect so that there's not a big gap between what they have on-premise and what's offered in the cloud? That's first the technological problem.

The second thing that customers really struggle with is the tool and in the operational practices. This is where some of the economics comes in because if you just take your applications and you move them to the cloud, that's a really good first step but it's not the end result.

A lot of customers, it's not just about reducing the cost. A lot of what they're trying to do is they're trying to increase their velocity. To increase their velocity, they have to change how they develop software, how they deploy software, how they operate that software.

I advise people, look, take your applications. Move them to the target environment. Don't think of this, like, when you move it you're going to get all the benefits day one.

Now what you do is you incrementally refactor the application. Maybe you move the front-end. Maybe you change the database layer. You change your operational practices, how you do patching.

You retrain your operations staff. You move some of that responsibility from one group to the next. If you take that as a much more incremental approach, I find that a customer's transition to the cloud is much, much easier.

That's also, by the way, what we found within Oracle. OCI is an infrastructure platform that underpins everything we do at Oracle. The whole company is transitioning and, over the next 12 to 18 months, we'll have everything within Oracle transitioned to OCI.

We're taking that exact perspective. Move the stuff and then incrementally refactor. It's going much better than trying to do it as a big-bang migration all at once.

Michael Krigsman: That measured transition also makes it easier from the cultural mindset and people change aspect as well.

Clay Magouyrk: Correct. If you have a large organization, you can't just expect to reinvent everything the next day. It's a big culture shift. It's a big technological shift.

What I find is the companies that do well at it have a significant business advantage because they have the higher feature velocity. They actually lower their costs. And their employees are happier because they're working on new, relevant technologies, which give them confidence that they're building their skillset and they have more markable kind of value in the future.

The economics of cloud computing

Michael Krigsman: Okay, but drill in for a moment on this economics question. How should organizations think about cloud economics?

Clay Magouyrk: The kind of state of the art cloud infrastructure, prior to what we're pushing with Oracle is, well, compute is going to be expensive, so what you should do is you should rewrite your application so it scales up when you need it and scales down very rapidly. You should put everything into functions or you should containerize everything. Yeah, we know that compute is expensive and also we know that that bandwidth is really expensive, but we're going to give you amazing new tools to optimize all of that.

The consequence of that is that when customers are able to write new applications and take advantage of that technology, they do realize cost savings. But when they move existing applications, they end up with bills that are significantly higher than they expected.

The approach that we take is, well, as an example, why can't you have a virtual machine that you can scale up or scale down while you're using it? Why can't you have a block storage volume that you can dynamically change the performance of while it's online? Why can't you have completely flexible infrastructure from a load balancing perspective where you don't have to prewarm your load balancers but you also can get up to super-high capacity or very low capacity all with very, very cheap costs? Why is it that bandwidth in the cloud seems to cost 10 or 20 times more than it did when you were on-premise?

We're building in these fundamental technologies. We have what we call flexible compute where you can pick the exact number of cores that you want. Later on, in a couple of months, we're launching the ability to decouple the amount of RAM from your compute cores. You can pick arbitrary sized DMs.

Then later, early next year, we're launching the ability to auto-scale those compute offerings dynamically. What that does is it gives a customer with an existing application the ability to put it on a VM. If it needs to scale up, it can. If you want to scale it down, you can. You can turn that on automatically.

Now, to get the benefits of the cloud, you don't have to rewrite. You can just bring your existing applications.

Now, does that mean that that's the way you'd write a new application? Maybe not. But for a company who has all of this existing investment, as an industry, as cloud providers, we have to help them with those things. What we'd been doing at Oracle is, as we continue to bring out new compute offerings, as we continue to make advances in storage and networking technology, as we're making significant advances in our database, we pass on those savings to customers.

Michael Krigsman: You're using architecture, software, and hardware design to drive the economics down, ultimately.

Clay Magouyrk: Completely. Absolutely, you drive it. It's not just the work that we've done in our public cloud. If you look at, for example, some of the things that we're able to do with our hybrid cloud offerings with Dedicated Region Cloud@Customer, we're able to offer that at a price point.

You can see this is the industry's response to this. We have a couple of big customers and we launched it last month. For a low commitment of about $6 million a year, you can get a full region of OCI in your data center.

People then ask me, "Well, what's the price of the services?" Well, it's the same as the public cloud. When they find that out, they're shocked.

Well, it turns out that when you take a step back and you actually design this in from the beginning, you hire the best engineers in the world, you get amazing results. That's what we've been doing at Oracle for the last six years.

Cloud strategy: public cloud, hybrid cloud, and multi-cloud

Michael Krigsman: Let's talk about the strategy for moving to cloud, public cloud, hybrid cloud, one cloud versus multi-cloud. What should customers do?

Clay Magouyrk: If you are a new startup from scratch, the reality is you should pick a cloud provider, a single one. You should invest heavily in that. You should pick the cloud that closely aligns with your overall goals and where you can hire employees. I would never advise those startups to go multi-cloud.

If you are a large corporation and you have a massive estate of different technologies from different on-premise IT vendors, the reality is, and I think you see this born out in the industry; you're not going to be able to work with just one cloud provider. The problem is, based on what your customer profile is, that changes your strategy.

Now, what I can say is that, for the vast majority of customers that I talk to, traditional kind of Oracle customers, most enterprises, the reality is you're going to be multi-cloud and you're going to have to have a hybrid cloud strategy. The easiest way to do that is to pick which workloads are in which applications are going to which cloud provider and make sure you have highspeed interconnects between those cloud providers. Make sure that you don't have things that are too chatty across them. You move centers of gravity to each one.

You're going to have some things that keep staying on-premise. Make sure you have high bandwidth connections from on-premise to your cloud providers. Then you're oftentimes going to want some evolutionary way to get there, which is where things like the hybrid solutions by cloud providers whether it's something like AWS Outpost, Azure Stack, or at Oracle we have this thing we call Cloud@Customer, which is an umbrella brand for everything we do around bringing the cloud to the customer premises.

We offer a couple of things like, for example, Exadata Database Cloud@Customer, which brings you our managed database cloud service, including autonomous database into your data center. You have to pick based on the different functionality each cloud provider offers. What is the best fit for each workload?

Oracle Cloud@Customer strategy

Michael Krigsman: You mentioned Oracle Cloud@Customer. What's the strategy behind that? Weave in regions as well.

Clay Magouyrk: Oracle has been a very large on-premise IT provider for a long time. We work with our many, many hundreds of thousands of customers. What they tell us is, they all want to go to the cloud but they also have a lot of stuff on-premise.

How can they bring some of the cloud awesomeness to on-premise? That's where we really have focused on our Exadata Database Cloud@Customer, which brings a managed database service into your premise and it builds on our massive success with our Exadata hardware and software integrated stack on-premise. Customers use that and they can manage their cloud database. Their database is on-premise, but from the cloud.

The nice thing about Exadata Database Cloud@Customer is you get to keep all of your data, including your backups, on-premise if you like. You still retain full control of all of your data.

We also, as Oracle, are not just a database provider. As you know, we do applications. We have been a large middleware player for a long time, and so customers are like, "Hey, I want more of your services available behind my firewall."

What we offer to them is what we call Dedicated Region Cloud@Customer. The easiest way to think about this is it's an entire public cloud region on your premise. All of the services are there. As we launch new services, the new services show up in your region.

Where does your data live? In your data center. Where does the control being run? In your data center.

How much does it cost? Is it a pay as you go perspective? It's the same prices that you pay in the public cloud.

We find that customers are able to get a bunch of the benefits of the cloud on their premises. The fact that, for example, these APIs are 100% compatible with the cloud stuff, as they rewrite applications and as they modernize, they can then take those things and move them to the public cloud in the future as their business continues to change.

Michael Krigsman: What are one or two quick use cases or, in other words, why is this important for customers?

Clay Magouyrk: For example, with Exadata Database Cloud@Customer, a lot of what we see is customers who need super-high performance and high availability Oracle databases but maybe they don't have the expertise to run them by putting it together themselves, buying the hardware, software, doing it all together. They can just get an Exadata Database Cloud@Customer. We ship it to them and it installs in a few days.

Then they can provision Oracle databases, back them up. Patching and all that stuff is handled for them and you get the benefit of autonomous database, which removes a whole lot of the functionality DBAs previously had to do around tuning your database, optimizing it for you. It makes an Oracle database just massively easier to use.

Around Dedicated Region Cloud@Customer, typically this is companies that maybe they're a telco and they have applications that they can't move to the cloud because they need to be in specific locations through low latency to their customers, or perhaps you're in banking or healthcare industry that's heavily regulated. You've signed contracts about what level of control and access you're going to have.

A lot of those customers want to move to the cloud but they need that control and security that only they can provide. Well, that's where Dedicated Region Cloud@Customer really comes in and saves the day.

When we built OCI, we knew that we were going to have hundreds or thousands of regions, not 10 or 20. Because of that fundamental engineering, we did upfront and the ongoing work we do to make sure that we have the architecture that scales up and scales down, that enables us to offer these services at a small footprint and then also be able to incrementally expand them for customers to as large as they need.

Oracle Cloud@Customer: Product engineering and design challenges

Michael Krigsman: From an engineering perspective, what are some of the challenges or the considerations that go into designing a system that has parity between what's in the data center and what's in the public cloud?

Clay Magouyrk: For example, part of it starts with your actual rack design, how you design your fundamental hardware footprint. You want to be able to have a system where the services themselves can run on as few different hardware combinations as possible.

Second is the network design from a physical network perspective. Can you have a network that you can start small or start large and then incrementally expand? That involves, how do you actually design the core spine and leaf switches? How do you actually have your routing protocols? How do you have a scalable, both up and down, edge network so that you don't have to plot down $100 million of equipment to be able to just spin up one of these regions?

When people talk about scalability in our industry, they always think about scaling up. Even when you just think about it, no one ever talks about scaling down. If you just take that as a goal, we knew, "Hey, I don't want to tell customers that I can't give it to them."

We took it as a goal to be able to scale down. That's what's enabled us to roll out regions like public cloud regions faster than any other cloud provider. It's what enables us to do things like offer Dedicated Region Cloud@Customer and it's also what enables us and you'll continue to see us just introduce massive numbers of regions around the world.

Advice for cloud migration

Michael Krigsman: Finally, what advice do you have for organizations who want to move to the cloud but are facing the kind of technical, economic, cultural challenges, talent challenges that you described earlier? What advice do you have for those folks?

Clay Magouyrk: Look, cloud infrastructure, we're plumbers. We want to build things that just work and people just accept. No one wants to say, "Oh, my gosh. This is such amazing plumbing," but they're very unhappy when it doesn't work well.

A lot of what the transition to the cloud is, it's not fancy rocket ships and amazing ideas. It's taking applications. It's moving them. It's incrementally refactoring them. It's every day waking up and operating, operationalizing, and finding that efficiency.

Yes, you have to create a cloud center of excellence. You have to hire people and train people to know how to use the cloud. But at the same time, being very deliberate about making choices, picking cloud providers, and understanding what are their unique differentiators and what do they have to offer and take advantage of those things, I think that's all you have to do.

Michael Krigsman: Clay Magouyrk, Executive Vice President of Oracle Cloud Infrastructure, thank you for sharing your expertise with us today.

Clay Magouyrk: Thanks for having me, Michael. This was a lot of fun.

Published Date: Sep 01, 2020

Author: Michael Krigsman

Episode ID: 668