Amazon Web Services Inc.’s essential promise has been that with its cloud services, companies no longer need data centers for just about all of their operations — but as it turns out, not quite all.

As Amazon Web Services Inc. opens its seventh annual re:Invent conference this week in Las Vegas, it’s looking to extend its dominance in the cloud computing juggernaut that’s sweeping across all of information technology by going after large enterprises, not just startups. But many of those companies need to keep running some applications on-premises because of latency, regulatory or other issues.

And now, AWS Chief Executive Andy Jassy (pictured) is at least publicly embracing the notion of these “hybrid” cloud setups with an enthusiasm that belies his previous insistence that building one’s own data centers was “kind of wacky.” I met Jassy last week for an exclusive interview at the sports bar he has set up in his Seattle home to learn more about AWS’ next stage.

In this second of a three-part series from the two-hour conversation, Jassy talked about AWS’ hybrid cloud strategy, as well as how it’s viewing the rising importance of edge computing. He also talked about how AWS is beefing up its alliances with partners and expanding its Marketplace app and services store. This interview is edited for clarity.

Look for more strategic and competitive insights from Jassy in the third part of the interview tomorrow. The first part ran Monday with hints of what’s coming at re:Invent and how AWS aims to keep growing.

Also, check out re:Invent coverage all this week by SiliconANGLE in its special report on the cloud, including insights from its market research sister company Wikibon. Not least, SiliconANGLE’s livestreaming studio theCUBE will be interviewing dozens of AWS and other cloud leaders today through Thursday from the show floor.

Hybrid helpers

Q: How is AWS approaching the reality that enterprises still need and want some operations to remain on-premises. You have a big deal with VMware, but is AWS really serious about hybrid cloud computing?

A: When we think about hybrid, we think about on-premises data centers and the cloud, and the cloud being things like AWS. We’ve been saying pretty consistently for the last several years that we believe that there’s really a really substantial transition going on that we’re in the middle of, where the majority of companies are not going to have their own data centers, and those that keep them will have much smaller footprints than they have today.

We still believe that. And I think you can see that borne out in our numbers, where we’re a pretty big business still growing 46 percent year over year. But I think that transition will take place over many years. There are going to be some workloads that [enterprises] to need to keep on-premises… because they require really low latency or need to run close to the factory.

What we hear consistently from customers is that they would like to be able to run their existing on-premises data centers that they’re not ready to retire yet as seamlessly as possible with AWS in the cloud. And that’s what we’ve been working on pretty significantly for the last seven or eight years. It started with services like Direct Connect, which is a private connection from on-premises data centers to AWS, and Storage Gateway that lets you store things locally and actually asynchronously load to the cloud, and a number of services like that.

Q: How does VMware fit in and what’s coming next?

A: A couple years ago, when we announced this partnership with VMware, that was something that we had been talking about for a while and that our joint customers really wanted. They wanted to use the same software they used to run their on-premises infrastructure, that they’d been running on VMware for a long time, on AWS. And that has had very, very strong early resonance in the market, and the momentum is very strong for that offering.

Part of what’s also happening is that the relationship between VMware and AWS is not only helping companies that are trying to make this transition to the cloud do so more easily, but I it allows VMware, as workloads move to the cloud, to have their software services remain really important and relevant and central for customers as they move to the cloud. That’s what we’re seeing in the collective offerings we have, and that’s what I expect moving forward.

AWS on-prem?

Q: Will we ever see an AWS software or services stack for the enterprises and partners interested in deploying on-premises? Is VMware’s RDS a tell sign that you guys are open to continuing to look at other options on premises when it makes sense?

A: I think we’re open to anything. The only thing that we’ll close the door on is doing something that we believe is insecure or that it’s going to be a bad customer experience. So anything is on the table for us to consider. We have looked really hard every year for the last five years in our operating cycle, which we just went through again this year, at ideas of doing more things on-premises. And one thing that we’re pretty clear on, at least at this point, is that we don’t like the customer experience of what we see today in some of these [other] on-premises experiences. They’re not working well, and they don’t have a lot of traction.

If you do something on custom hardware with a different control plane than what you use in the cloud as on-premises, with different tooling, and where the offerings on-premises and in the cloud are really different, customers get really frustrated and there’s not a lot of value. It sounds good, but it doesn’t work in practice.

We will release over 1,800 service and features in 2018 alone, and those are the customer-facing feature ones. We have thousands of other deployments that are behind the scenes. You’re never going to be able to keep that pace of change as a services model in the cloud, as you will in on-premises hardware. So the offerings very quickly get inconsistent and feel one-off, especially on-premises.

Q: If you look at what you guys have done, you’ve created a horizontally scalable infrastructure. And so does the stack even exist?

A: We don’t talk about the word or the term stack very often. We don’t think about it that way. We think of these as a set of services that you can compose together however you see fit, in whatever combination you want. If we were going to do something on-premises, where we were going to bring compute or storage or some of those capabilities on-premises, we would try to reimagine what’s there. I think we have some ideas on how to make it a better experience.

If you’re a builder and you want to use compute and you want to use block storage and then you want to combine that with a database, you want to combine that with analytics and machine learning, you don’t want different capabilities. You want to have a consistent interface, API and capability, use all the same services you’re using for AWS, for all your other applications that live in AWS.

Data centers: the new edge

Q: Is multicloud a real trend?

A: Just so we’re talking the same language, I don’t think of on-premises and the cloud as multicloud. I think of that as a hybrid. So when people talk about multicloud, I really take it as multiple cloud providers. And I think over the last number of years you’ve seen companies — most enterprises, when they begin their journey of moving — planning their move to the cloud will start off thinking, maybe I should be multicloud.

Oftentimes they think about maybe I’ll split up my workloads in some kind of even manner across two or three. Very, very few end up making the decision to do so. The vast majority predominately pick a cloud provider. And the reason they don’t split it evenly is that if they do, they have to standardize to the lowest common denominator, and these platforms [are on] wildly different capabilities and stages of development.

It also forces teams to be fluent in multiple cloud platforms. It’s very unattractive to development teams and developers and very resource-inefficient. And then all the cloud providers have buying discounts. So you really lessen your bargaining power if you’re splitting across multiple providers. Sometimes they have groups that, you know, just really want to work with a cloud provider on this workload, or sometimes that company will say, well, I want to make sure, if something goes sideways with my primary provider, I know how to run it on somebody else.

Q: How does edge computing change the dynamic of data center and the public cloud?

A: There are lots of endpoints you can call the edge. A big edge is a data center. A smaller edge is a location somewhere on the road with no connectivity. An even smaller edge is a device somewhere. When you talk 10, 15 years from now, I think a lot of the on-premises piece is going to be the edge. And a lot of them are going to be devices, too.

We have a lot of offerings in those areas today, and we think about if you want to provide people at the edge, you need a bunch of offerings. Some people theorized, with more and more edge devices, that that would somehow be bad for the cloud. In my opinion, it’s a giant opportunity for the cloud. Because most edge locations, either devices today or data centers over time, because they’re smaller footprints, will have disproportionately small amounts of compute and storage available, so all the heavy computation and analytics are going to be done up in the cloud.

You’re going to have multiple options. You’ll be able to always run on the cloud for computationally heavily or analytic workloads or the brains of what you build and your machine learning pieces. Then you’re going to have smaller data centers that are largely going to run things that, just for latency reasons, have to live there… but also seamlessly connect with the rest of the cloud.

Q: So you’re saying the edge is going to be very compatible with cloud in terms of its interactions.

A: The brains are going to be in the cloud, and the inference and the predictions and the actions that have to have low-latency will be done on the device or the edge itself.

Q: Some people wonder what AWS’ strategy is for allowing customers to move workloads out of its cloud if customers find it necessary?

A: We have a pretty broad set of migration tools that we’ve built over time. Database and virtual machines and applications. All of our tools that migrate in also migrate out. In reality, relatively few companies are going to want to be moving these things around lots of places, because there’s opportunity cost and there’s people cost, and frankly, you don’t gain very much by doing so, and the feature sets are so different. But if people want to, we have migration tools that let them, and we’ve had a handful of customers who’ve chosen to do so, especially when there have been acquisitions.

Partnership party

Q: What other big partnerships are you looking at?

A: We have thousands of partnerships today, as a lot of the ecosystem runs on top of AWS, and I think that when you’re talking about the depth of partnership we have with VMware, you can’t have too many of those. That is a very deep partnership, and organizations don’t have the bandwidth for too many of those deeply integrated, deep technical collaborations, deep going to market together.

We have a good one with Red Hat and with what they’re doing with OpenShift, that was announced at Red Hat’s conference about a year ago. We have a pretty deep collaboration with Salesforce … on the machine learning and the IOT side and some of our applications together.

I could see over time having a handful of other relationships like the VMware one.

Q: What’s the role of the system integrators?

A: I think about SIs and ISVs or SaaS providers. On the SI side, they don’t really have the type of fully featured, many-services type of infrastructure and platform as a service cloud that something like AWS has. And so we partner very deeply with SIs, and their primary role is to help enterprises be able to migrate to AWS. They’re selling their own capabilities to help enterprises make that transition to the cloud, but they’re also building those applications and helping them implement them on top of AWS.

There are some SIs that are replacing their outsourcing business on top of AWS. DXC Technology is a good example, where they’re moving a lot of those customers to hosting solutions on top of AWS.

And then there’s ISVs and SaaS providers, new ones like Acquia, are from the get-go completely on top of AWS. And lots of existing ones, like Salesforce and Workday and Infor and Informatics and Tibco and Splunk, those are all evolving their existing applications, which started around the same time as AWS or earlier, and moving more and more of those workloads on top of AWS.

Marketplace madness

Q: You don’t have to be a rocket scientist to figure out that with SaaS, with sets of services, marketkplaces will become more common. How are you evolving the AWS vision around Marketplace?

A: We’ve made a huge amount of progress, and we’ve launched a lot of things that have helped our customers on both sides of our Marketplace, both our software and ISV partners who are trying to sell their software as well as people that want to consume it.

One of the things we hear from both of those segments is that they hate the current process of selling and procuring software. It’s arduous. It’s slow. It’s all this contract work. It’s hard to actually get to deal with the payment piece of it. it’s hard to deal with the deployment piece of it. So a lot of what we’ve been building in the Marketplace is to make it as simple as possible for a software seller and a buyer to work together.

Q: Examples?

A: If you look at what we’ve launched with ISVs and what ISVs have been able to sell, and what SaaS providers have been able to sell in the marketplace over the last year, it’s completely changed what’s possible for buyers as well as those ISP and SaaS providers. We want to make it easier for those two parties to contract, starting with a starting point that’s 80 to 90 percent done. And then they can iterate on the last piece of it.

I think you should expect us not only to continue to make it easier for both parties, but to add selection in areas where historically there hasn’t been much. Look at the Lambda functionality. Look at how many people are building their own Lambda functions. Those should be shared.

Q: Do you think Marketplace could be better? Sexier, more glam? People have said, “Marketplace feels old; it feels like a catalog.” You’re in the retail business.

A: I feel really happy with Marketplace. That business is booming. It’s grown really fast. It’s a really significant business for us. Both sellers and buyers getting a lot of value. A lot of sellers have turned to it as a primary channel that they want to sell through. And I think we’re at the beginning of what’s possible with Marketplace. I think you’ll see even at re:Invent a number of additional categories that we’ll add to Marketplace that will change what people can buy.

Photo: Robert Hof/SiliconANGLE