For Kenny Mullican, the decision about whether to use a hybrid cloud was once cut-and-dried. “It was a binary choice between your own servers and going with a public cloud,” said the chief information officer at Paragon Films Inc.

Fear of losing control over mission-critical systems is one reason the company, which makes stretch films used in packaging, has hedged about moving large parts of its processing workload to the cloud. “Can you afford to have Microsoft choosing when your outage is going to be?” he said, referring to Microsoft Corp.’s Azure cloud service.

But with the advent of Azure Stack, a version of the cloud provider’s platform that runs in on-premises data centers, Mullican is taking a harder look at hybrid cloud options. His information technology organization doesn’t have the resources to install and maintain a full version of Azure, but he thinks nearby “co-location” providers that rent businesses space for servers and other hardware could do the job. That could give Paragon the benefits of cloud virtualization without the latency and reliability penalties of a long-distance internet connection and with more personalized service as well.

“Now there seems to be a spectrum of options,” Mullican said. “It doesn’t have to be one way or the other anymore.”

That pretty much sums up the state of hybrid cloud, a concept hatched in the early days of public cloud infrastructure to describe a combination of private, on-premises computing with one or more public clouds. Hybrid cloud quickly captured the industry’s imagination. International Data Corp. dubbed 2014 the Year of Hybrid Cloud and Gartner Inc. forecast that half of enterprises would adopt a hybrid strategy by 2017.

Now more relevant than ever, the concept is poised to shake up the established order in public cloud computing. Indeed, hybrid computing has become a key nexus of competition in the cloud, embraced to varying degrees by just about everyone: No. 1 cloud infrastructure provider Amazon Web Services Inc., hard-charging Microsoft, challenger Google LLC and a range of traditional IT providers such as Oracle Corp., IBM Corp. and Dell Technologies Inc.’s majority-owned VMware Inc. All of them offer, or are planning to offer, various flavors of hybrid cloud offerings that collectively will redefine the multitrillion-dollar IT industry.

Changing definitions

The evolution of hybrid cloud will be on the minds of the thousands of IT professionals, engineers and suppliers who will flock to San Francisco this coming week for Google’s Cloud Next conference. Those early predictions about hybrid cloud adoption mostly came true, but a recent cross-pollination of technology and business developments has considerably grown the varieties of hybrid clouds.

David Floyer, founder and chief analyst at research firm Wikibon, a sister company of SiliconANGLE, recently proposed that five different types of hybrid clouds are emerging, ranging from autonomous clouds connected by networks at their most simple to complex “autonomous standalone clouds,” in which processing pools of various size and function exist independently, connect sporadically and operate in real time.

Gartner Inc. even all but stopped using the term last year, deeming it too limited in scope. “I don’t think hybrid cloud itself is a useful term,” said Craig Lowery, research director in Gartner’s technology and service provider group. Instead, Gartner uses the term “hybrid IT” to describe the mélange of virtualized servers, containers, microservices and edge devices that IT organizations are figuring out how to manage.

Does the definition really matter or is it merely a semantic debate? Experts agree that having a common frame of reference when discussing IT architecture is important, particularly in areas such as cybersecurity, where people charged with protecting data need to know precisely where it is. “When you’re moving data back and forth [definition is] very relevant, particularly in regulated industries,” said Rohinee Mohindroo, co-founder of Dyjit LLC, who built her first hybrid cloud in 2010 as the CIO at a subsidiary of Rakuten Marketing LLC.

Beyond bursting

Early definitions of hybrid cloud largely referred to cloud-based extensions of on-premises applications that could be deployed when needed, a tactic called “cloud bursting.” But today’s reference models are much more sophisticated.

Hybrid cloud is evolving to encompass complex scenarios such as networks of autonomous devices that orchestrate large fleets of self-driving vehicles, each connected to multiple clouds. “People now have powerful options about where to place things,” said Brian Gracely, director of product strategy at Red Hat Inc.

Most hybrid clouds initially consisted of a bursting arrangement with a single public cloud provider. But today, said Jack Norris, senior vice president of data and applications at MapR Technologies Inc., “we see organizations wanting to treat the cloud as a service that’s part of an overall infrastructure strategy, but not with one vendor. That’s particularly true of global companies whose needs aren’t defined by one of the big three” public cloud providers: AWS, Microsoft and Google.

Evolutionary drivers

Reference architectures for hybrid clouds matter because on-premises computing isn’t going anywhere anytime soon. Wikibon expects enterprise spending on private clouds – which are on-premises computing stacks that mimic public cloud infrastructure — will grow nearly 32 percent annually through 2027. That’s more than double the growth rate of public cloud, although from a much smaller base.

A recent study by managed hosting provider Cogeco Peer 1 Inc. found that 92 percent of respondents run on-premises servers and two-thirds consider them highly important to their businesses, according to Cindy Jordan-Ford, general manager of the company’s U.S. and Latin American operations. “Hybrid IT is optimal for preserving existing solutions while implementing new technologies,” she said. Most businesses use cloud providers to access services and tools they can’t build themselves rather than migrating existing services to public clouds, she added.

Among the factors driving hybrid cloud evolution are recent technology advances that have made it easier for users to shift processing loads between platforms and to tap into best-of-breed offerings from both public cloud infrastructure and software-as-a-service providers.

“Moore’s Law is happening at the compute power level, [application program interface] level, software development level and analytics level,” said Jeff Kaplan, managing director of THINKstrategies Inc., a consulting firm that specializes in on-demand computing services. “As each becomes more powerful, they become more easily integrated.”

API economy

The explosive growth of APIs — the number of published APIs has roughly quadrupled since 2012, according to Programmable Web Inc. — has made it easier for developers to stitch together sophisticated applications by tapping into services exposed over the internet. That means applications and data can be spread across multiple cloud and on-premises environments, reducing the need to consolidate processing in a single cloud.

“You can put parts of your application in different places and communicate using remote procedure calls and APIs,” said Gartner’s Lowery. “It’s a cloud-spanning architecture.” The emergence of the Representational State Transfer or REST architecture as a consensus choice for exposing web services has fueled the trend, he said.

APIs are cloud-independent, which reduces the need for users to align themselves with particular cloud vendors. They’re even giving software-as-a-service providers a foot in the door with customers that are building hybrid clouds. For example, Salesforce.com Inc. uses its army of 4 million developers “to strengthen itself by allowing third parties to integrate with its platform, Kaplan said.

In fact, SaaS could be a wild card in the hybrid cloud game. Providers can use APIs to grab a piece of customers’ business that might otherwise flow to public cloud providers, particularly customers that aren’t interested in developing their own applications. Information Services Group Inc. recently surveyed 300 companies and found that 61 percent expect themselves to be “majority SaaS” by the end of 2021, up from just 5 percent today, ISG analyst Blair Hanley Frank said in an interview on theCUBE, SiliconANGLE’s streaming media platform.

However, API-laden applications have drawbacks that make them unsuitable for many enterprise purposes, said Arpan Shah, general manager of Azure marketing at Microsoft, which has a large SaaS business. “The SaaS model won’t be suitable for situations where a hybrid cloud model is needed to meet compliance and data sovereignty requirements, address network latency issues and enable edge and disconnected scenarios,” he said.

Kubernetes and the edge

Another technology driver for hybrid cloud is Kubernetes, the open-source software for orchestrating networks of the software containers, which in turn enable applications to run across private and public cloud platforms. Although still in its early days, Kubernetes promises a vast simplification of the process of managing complex multicloud environments.

“I think before Kubernetes, hybrid cloud wasn’t very feasible,” said Brian Grant, principal software engineer at Google and one of Kubernetes’ primary architects. Kubernetes gives organizations the ability to unify such essential plumbing as caching systems, messaging buses and systems monitors in a way that wasn’t feasible before, Grant said. “It used to take months or years to create that environment but with Kubernetes you can do that in days.”

The combination of containers, agile development methodologies and platform-as-a-service “are starting to fulfill the long-struggling goal of ‘bursting’ that virtual machines are ill-suited for,” said Microsoft’s Shah.

A third technology force is the rise of “edge computing” epitomized by the “internet of things” and the shift to move processing power closer to point at which data is generated. Examples of edge devices are sensors that need to make split-second decisions about whether to shut down equipment on the factory floor or point-of-sale terminals that recommend discounts for customer based upon the contents of their shopping baskets. Autonomous vehicles will have no time to shuttle data from a car sensor to a cloud application for a decision when the vehicle is bearing down on a tree. By the time a response is received, the assembly line is crippled, the customer has left the store or the car is totaled.

“If you want to have data-driven applications that do things in real-time or close to real-time, you have to process data where it’s created,” said Wikibon’s Floyer. “You can’t move it all to one place.” Wikibon has estimated that the cost of pushing computing to the edge of the network is also one-sixth that of pulling data into the cloud.

Edge computing, noted Wikibon analyst James Kobielus, “is driving more of the architecture of hybridized multiclouds toward radically distributed architectures. There will be less and less need for round-tripping.”

Indeed, said Red Hat’s Gracely, “As companies have become more comfortable with public cloud, they’re realizing they don’t have to think so much about being all-in on the data center.”

In a 2017 blog post entitled “The Edge Will Eat the Cloud,” Gartner analyst Tom Bittman suggested that edge computing will transform the way enterprises think about cloud computing. Data-intensive edge applications like machine learning and augmented reality look nothing like the workloads being processed in the cloud today. Organizations that move the quickest to retool at the edge will create competitive advantage.

“The edge will create some serious winners and losers,” Bittman wrote. “We need to expand our thinking beyond centralization and cloud and toward location and distributed processing for low-latency and real-time processing.”

Business drivers

The changing face of hybrid cloud isn’t just about technology. The current digital transformation craze has alerted organizations to the value of data – and the difficulty of managing and securing it as it moves around the globe.

In the seminal days of cloud computing, infrastructure-as-a-service providers persuaded customers to move large amounts of data to cloud platforms, reasoning that data should be located close to where it’s processed. Cloud providers dangled low-cost storage and free bandwidth as incentives, but once they had the data, they often made it inconvenient or expensive for customers to get it back. “The vendors are always driving for lock-in,” said MapR’s Norris. Owning data was the best way to achieve that goal.

That frustrated customers that wanted to arbitrage their cloud investments or shift workloads to providers whose tools more closely matched their needs. “You’re not going to take exabytes of data and move it to another cloud,” said Gartner’s Lowery. Many users are still smarting over the experience of being bound to software providers that held their mission-critical data ransom while raising prices and charging large maintenance fees. They’ve become savvier about maintaining control over their own data, and they see hybrid cloud as a form of investment protection.

This notion of “data gravity” holds that data is central to decisions about how it’s processed, a mantra often repeated by proponents of digital transformation. Having let cloud vendors take possession of large amounts of their data, some users are now questioning the wisdom of that strategy. They want to own the data and choose which services they want to manage it.

“Customers want to take more control of the software stack,” said Rhett Dillingham, vice president and senior analyst at Moor Insights & Strategy. “I expect a few years down the road organizations will have a mix of applications that aren’t attached to sticky surfaces and can be deployed across multiple platforms.”

Paragon Films CIO Mullican agreed. “Maybe it’s a pipe dream, but I think it’s where this has to go,” he said.

Users set the agenda

As users have grown more confident about public cloud, they’re putting cloud providers on notice that they need more flexibility to federate services. Their choice of a hybrid cloud model increasingly depends less on capacity and more on the needs of the business.

Gracely cited the example of two automotive companies that had very different visions for their markets. One wanted to optimize the in-vehicle experience, so it chose to use a central private cloud to connect to an entertainment and information system in the vehicle and to collect data about how the options were used. The other focused on optimizing for a web shopping and buying experience to mitigate customer unease about working with dealers. It went with a public cloud approach, using APIs to tap into a variety of video and social media channels along with ties back to its legacy configuration system.

“You can now start from the experience and use the cloud or third parties to do things you could never do before,” Gracely said.

The big public cloud companies have gotten the message and responded with on-premises offerings that duplicate their cloud software stacks. Microsoft has Azure Stack, Amazon has the as-yet-unshipped Outposts and Google has its Cloud Services Platform. For their part, traditional IT firms turned cloud hopefuls do too — IBM with  Cloud Private, Oracle with Cloud at Customer and VMware with major partnerships bridging the cloud and data centers. “Most large cloud providers have come to terms with the fact that they need to coexist,” Mohindroo said.

But while all the big three’s offerings expand customers’ hybrid options within a particular cloud vendor’s sphere of influence, they don’t make it easy for customers to shift data and workloads as they wish.

Cloud providers say they’re simply making things simpler for their customers. “We can’t be building a different application for every type of edge device and running two different environments, like [one for] cloud and [one for] edge,” Shah said. “Such a model will not be efficient, secure, manageable and cost-effective.”

Over time, IaaS vendors say, they’ll make boundaries transparent. “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,” AWS Chief Executive Andy Jassy said in late November at its annual re:Invent conference.

Although the hybrid service it announced then, Outposts, won’t offer all AWS cloud services on-premises when it’s available late this year, “in the fullness of time this suite will deliver the same experience on-premises, at the edge and in the cloud,” said Eron Kelly, general manager of enterprise services marketing at AWS.

Local cloud stacks are convenient but are also meant to entice customers to stay within a particular cloud vendor’s orbit, a fact that makes some users skeptical. “You have to ask if the vendors are providing the on-prem solutions because they think it’s a better way, or are they using it to keep people on their public cloud infrastructure?” asked Colin Bodell, chief technology officer at Groupon Inc. “My jaundiced view is that it’s a stepping-stone to the public cloud.”

Continuum of choice

Cloud providers disagree. “To do [hybrid cloud] in a durable way, it needs to run like a single environment, which requires both consistency and comprehensiveness,” said Microsoft’s Shah. That means having a consistent way to build applications, a common set of database choices, a reliable network and unified identity and access management, among other things. Microsoft has introduced several enhancements to Azure Stack that address such issues as security, cost accounting and monitoring, but is adamant that customers get more value from working with a single cloud provider than from juggling a heterogeneous set of services.

“Multicloud requires skill sets in understanding different clouds, which becomes challenging when companies are facing a shortage of cloud talent,” Shah said. “It also adds operational complexity.”

AWS’ Kelly agreed. Many CIOs “start off believing that they’re going to split their workloads in the cloud relatively evenly among two or three providers, but when they get into the practicality and the rigor of assessing it, very few end up going that route,” he said. Among the downsides of the multicloud approach he listed are the need to settle for “lowest common denominator” functionality, the difficulty of transitioning to hybrid and multicloud models simultaneously and the loss of discounts and other incentives users gain by working with a single providers.

There’s no question that federating multiple public and private cloud services creates complexity, but is choice worth that tradeoff? Dillingham said customers appear to be willing to accept some pain in exchange for flexibility. “Pushing more workloads onto a single cloud platform can give you benefits in areas like training, tools and discounts, but most users seem to prefer the flexibility of a multicloud path,” he said.

Could the growing preference for choice change the balance of power in the public cloud market? “Yes,” said Chris Wolf, global field chief technology officer at VMware Inc. “As you move more capability close to the customer, it creates the potential for disruption.”

VMware, IBM and a host of others are counting on customers’ demand for choice to fuel the need for services that can broker multiple cloud services. VMware has a close partnership with AWS that’s focused on making it easy for customers to move workloads between their own infrastructure and the public cloud giant, but the deal isn’t exclusive, and VMware is betting that customers will want more flexibility. “The work we’re doing with cloud providers is to let their customers run where it makes the most sense,” Wolf said. “We can provide a platform with common security, networking, management and whatever the business needs.”

Cloud providers tout the consistency and simplicity virtues of staying within a single platform, and all are moving to address customer demands for a wide range of development tools and stronger support for edge devices. AWS intends to do everything from “integrating networking, security and access control to powering automated workload migrations to extending AWS services to on-premises and edge locations so our customers can easily use AWS as a seamless extension to their existing infrastructure investments,” Kelly said.

Their message resonates with users such as Groupon’s Bodell, who’s currently involved in his fourth migration from on-premises infrastructure to public cloud. “My preference is to pick a single cloud vendor,” he said. “I don’t want to do the maintenance work myself.”

As the continuum continues to expand between one-size-fits-all and fully autonomous options, users will have more choices than ever to pick and choose their preferred hybrid models. Ultimately, the definition of hybrid cloud will be what each organization chooses to make it.

Image: Gellinger/Pixabay