Advanced Computing in the Age of AI | Friday, May 20, 2022

Workload Placement: Into the Cloud and back out Again 

Application migration is the process of redeploying an application, whether to newer infrastructure or to a private, public or hybrid cloud.

The history of cloud migration has not always been kind to IT groups charged with executing it. Often, the move has started as a “Cloud First” mandate from the CIO or CTO to move applications en masse to the cloud, and the lack of tools to justify or automate it left migration as a trial-and-error process. But I think we’re in a better place today, as the cloud becomes part of the fabric of nearly every enterprise.

Unclouding Takes Hold

What first took the name “reverse cloud migration” – moving applications from a cloud back to the corporate data center – has more recently been dubbed “cloud repatriation” or “unclouding.”

It's easy for IT to impulsively jump on the cloud bandwagon, which touts scalability, high availability and a more hands-off approach. But many enterprises have discovered that the public cloud wasn’t the best fit for every workload.  Gartner takes a tempered approach to the matter: “The cloud is not the end state, but rather an effective means of delivering some, but not all, application workloads.”  And Datalink recently reported that nearly 40 percent of organizations with public cloud experience have migrated systems from the cloud back to on premises.

Clearly, unclouding has become part of the everyday toolset of IT.

Not All Applications are Monolithic

But unclouding is more complex than it looks. Today, most enterprises deploy a wide range of applications that are customer-facing. Those applications interact with customers dynamically, and the code for these apps is built for public clouds.

At the same time those apps, because they wrap an element of business process around the customer, need to integrate with backend assets. So whether it is about customer records or financials, a lot of data resides in applications hosted on private, on-premises servers. This means many applications are essentially hybrid by nature, consisting of cloud and on-premises components.

What Drives Migration or Reverse Migration? 

Application migration involves many complex tasks, such as determining the right type of deployment model for each workload, performing the actual migration process and validating the migration to ensure that connections, service levels, security and performance are all maintained or enhanced. The matter is complex and dynamic, requiring identifying the suitability of each application for migration, moving applications from their sources to their target destinations and executing the move cost-efficiently and securely.

But what are the prime drivers for moving either to the cloud or back from the cloud? After cost-effectiveness, I would argue, comes both performance and availability; in fact, based on my experience with customers, those are typically their top two priorities.

Of course, this varies by organization. In a recent study by Gartner, performance and availability were three of the top five factors to consider in determining where workloads belong. For Gartner, top priorities were:

  • Performance
  • Latency
  • Service continuity
  • Availability
  • Reputation

Although Gartner did not specifically call out SLAs, any decision on cloud migration must consider SLAs for performance and availability. Why? In part because some applications are mission-critical, requiring both a performance SLA and an availability SLA. Others are not mission-critical, but may still require either a performance or an availability SLA.

Plus, any decision about SLAs must take into consideration the fact that shifting apps to the cloud also shifts the SLAs to the cloud provider. That, of course, means that SLA enforcement shifts to the provider, too.

Workload Placement

Whereas making decisions on whether to place applications on- or off-premises used to be a trial-and-error process, today it has become systematized in a methodology called “workload placement.” Workload modeling and simulation allows IT to “try out” an existing workload in a new environment without having to make the physical move to that environment. Workload placement provides the validation – through testing and simulation – to justify moving an app to a public, private or hybrid cloud, or leave it on-premises. Just as important, it can be the driving factor and justification for unclouding.

The point of workload placement is that IT no longer must leave costly decisions about migration to a best guess. It’s a next-generation approach to validating, if not proving, where each workload should be hosted.

Unclouding is an emerging trend, but perhaps there’s a better way to look at it. Workloads have to be evaluated, and periodically reevaluated, for their suitability as a cloud migration candidate. And the same is true for workloads currently running in the cloud. Periodic and systematic evaluation of workload placement is part of the business of IT, and a means of keeping enterprises running smoothly.

Len Rosenthal is CMO of Virtual Instruments.

Add a Comment