The average number of active clouds in enterprises -- already at a half dozen -- will continue to grow. Controlling these sprawling resources and services is crucial for maintaining efficient operation and preventing runaway spending. What's a CIO to do?
"A business might have multiple private clouds, multiple public clouds and multiple managed clouds, including managed services and service providers," said Jim Ganthier, vice president and general manager of Dell's engineered solutions and cloud business unit. "The question is, strategically, how do you execute that?"
It's not just the number of clouds in use that is soaring, according to RightScale, a maker of universal cloud management tools. In the newly published 2016 edition of its annual State of the Cloud report, the average number of clouds in service is six -- three each of public and private. But tally the number of virtual machines running in those environments and the scope of the management challenge widens significantly. In a public cloud environment, 28% of respondents operate between 100 and 1,000 VMs, with 17% surpassing 1,000. For private cloud, nearly half (48%) are operating in excess of 1,000 VMs.
Carl Lehmannresearch manager at 451 Research
For Ganthier, it's easier to gather the various cloud types, locations and services into an all-encompassing catch-all: multicloud computing. "When I say multicloud, it shouldn't be a public or a private cloud. I might have multiple private clouds, multiple public clouds and multiple managed clouds, including managed services or service providers."
Getting a grip on a multicloud computing environment isn't just a matter of oversight from an omniscient central control point. Choosing where workloads run and ensuring that they operate in a cost-effective manner are essential to effectively managing a hybrid cloud architecture. It is also crucial to make sure that cloud sprawl does not lead to runaway siloing of workloads or data.
"The first problem in hybrid cloud management is that you have to decide which workloads run on which infrastructure," said Carl Lehmann, research manager for enterprise architecture, integration and business process management at 451 Research. "This is tied to performance elasticity, service-level agreements and economics."
What's in a name matters
Once decisions are made where particular workloads or processes should be run, or where data should be housed, administration quickly grows convoluted by the lack of terminology standardization, according to Brian Adler, RightScale's director of enterprise architecture. Where Amazon Web Services (AWS) uses a management console, Google Cloud Platform uses a portal. Accounts in AWS are called projects in Google. "Keeping the nomenclature straight is a problem, let alone juggling all the different portals into the data," Adler said. "What you want is a single pane of glass, the ability to view everything in one unified, consistent way." Adler said RightScale's customers continually complain that this lack of consistency is a key factor that makes centralized management difficult.
Lehmann agreed with Adler's assessment of the quandary, noting that as workloads get distributed across ever-widening hybrid cloud architecture, application performance monitoring is essential to maintain an overall view. Lehmann also noted the search for value pricing raises its own issues. If workloads are performing fine on their respective cloud execution venues, will those performance levels be maintained or improved if those loads are consolidated to a single platform, such as AWS or Microsoft Azure? Determining the answer can require significant testing and performance management, he said.
Cloud sprawl leads to silos
In the race to retire traditional consolidated data centers and replace legacy applications with modern cloud-based counterparts or software as a service subscriptions, fragmentation is inevitable. Even for clean-sheet startups that sign up various cloud providers of storage, analytics, backup and other services, there is a high likelihood of creating silos that cloud computing was intended to eliminate.
"Business leaders have been aware of the challenge tied to silos across the board," Lehmann said. "With hybrid cloud management, it gets worse, because there are so many more options to choose from as to where to run a workload and where to put the data." Breaking down silos requires integration of applications, data and services, complicating matters for developers. According to Roberto Medrano, executive vice president at API services vendor Akana Inc., based in Los Angeles, the number of integrations enterprises must deal with is typically several hundred.
Though Adler warned cloud sprawl is likely to expand, he said it is possible to maintain control through the use of tools for managing a hybrid cloud architecture that are universal in nature, and support the concept of deploying anywhere and managing everywhere.
"It's more than identifying the workloads you want to run and matching them with the right resource pool," he said. "I don't even call it a cloud, because sometimes, the right resource pool is an on-premises data center." The type of project also is a factor in choosing the pool. "For a project that might not be in production for the long term, a new niche cloud resource pool might work fine. But if it's something that's going to be in place for a long time, one of the established players is likely to be your best option."
Be aware of cloud providers' services
An often overlooked factor is portability, according to Adler. "If you want to survive in this multicloud world and have the ability to move from vendor A to vendor B, for a technological or financial reason, you need to beware of each cloud provider's services," he said. "They can take care of backup and scaling for you, and be very enticing and brain-dead simple to use. The danger is the more you use them, the more the handcuffs tighten."
Lehmann also is concerned about portability. "There are plenty of hybrid cloud management platforms; all the major cloud providers have their own version to migrate workloads." Characterizing these tools as "lift and shift," after their function of moving applications from an on-premises data center into a particular cloud, he said their functionality leaves much to be desired.
"All of these tools claim to do some degree of hybrid cloud management, but I do not personally believe they do a very good job with the complex challenge of dealing with multiple clouds, multiple software services, multiple infrastructure services or platforms as a service." In his view, companies are, in some cases, inventing their own technology for hybrid cloud management.
Ganthier and Adler both said they believe that's unnecessary, and each, not surprisingly, offered tools to manage the ever-expanding hybrid, multicloud computing universe. Dell Cloud Manager, currently in release 11, is based on technology developed by Enstratius, which Dell acquired in 2013. RightScale's universal cloud manager supports public, private and virtualized cloud environments.
Managing a hybrid cloud architecture will continue to grow increasingly vital, as corporate clouds multiply and the number of applications and services running on them swells. It can be done, but getting there requires careful analysis of current infrastructure, along with a clear strategy for future workloads and multicloud infrastructure.
Perhaps Dell's Ganthier put it best. "Don't try to be a pioneer," he said. "Learn from our experience; frankly, learn from our scar tissue."
Valerie Silverthorne contributed to this article.
Break down silos to improve efficiency
Refactor Windows applications to keep them relevant
Don't wait, detect vendor lock-in now
Plan carefully for multicloud computing