Cloud Migration Reality Check

A look at cost, staffing requirements, timeframes and potential pitfalls when moving to a cloud infrastructure.

By Gene Knauer

By Gene Knauer April 15, 2022

The migration of workloads to the cloud is relentless. By 2024, says McKinsey, most enterprises expect $8 out of every $10 they spend on IT hosting to be in a public or private cloud environment. In addition, researcher Adapt predicts that the share of IT storage and processing that takes place in the public cloud will rise to 47% by 2023, an increase of 10% over current levels.

Yet companies are being choosy about what workloads they migrate and where. They’re weighing the cloud’s benefits against the nontrivial tasks and risks inherent in moving apps from three-tier data centers to cloud-enabled environments, which are increasingly multicloud in nature. There are cost, expertise, time, performance, compliance and other issues to consider when joining the Great Cloud Migration.

Workloads on the Move

Changing business goals and regulations are among the diverse factors fueling application relocation. A study in mid-2021 by U.K. researcher Vanson Bourne for the Nutanix Fourth Annual Enterprise Cloud Index report, for example, found that 91% of 1,700 global, cross-industry enterprises surveyed had moved at least one application to a new IT environment in the prior year. The top reasons were to improve their security posture and ability to meet regulatory requirements, accelerate data access speeds, and gain better control of the application.

But with a decade of cloud computing experience now behind them, enterprises have learned that a one-size-fits-all approach doesn’t necessarily deliver all the benefits they seek, as evidenced by the uptick in multicloud use. More than a third (36%) of all respondents globally said they were running multiple clouds, private or public, as their primary operating model—the dominant infrastructure in use globally. The reason is that they’re increasingly looking to match each application or workload to the IT infrastructure best suited to handle it. But what constitutes “best” can change over time, so enterprises want the agility to move resources around when conditions warrant it.

At this juncture, most enterprises are in the early stages of public cloud adoption, according to Peter Chang, cloud architect at Nutanix. He estimates that “a small percentage of their workloads, perhaps 15% to 20%, have been migrated.”

Related

Cloud Computing’s Catch-22

The lengthy amount of time to move apps might be causing hesitation to migrate more aggressively, he said. Chang sees companies migrating multiple apps in parallel because it can take one to two years per migration, depending on the size of the application.

“Most companies are underestimating the complexity of cloud migration,” he said. He added that the COVID pandemic, which resulted in high turnover rates, slowed down a number of migrations, as well.

“It takes a lot of time to hire and train new IT staff. Companies don’t want to slow their business down to do a lot of testing and application cutovers,” Chang said.

A Challenging Priority

While a 2021 study by Next Pathway found that cloud migration was a top-three priority for 65% of 1,000 IT decision-makers it surveyed, these organizations were also upfront about the challenges they faced. Among them: planning an effective migration, understanding data lineage in the applications that they want to migrate; concern about how to proceed without disrupting critical business applications; moving SQL storage procedures; optimizing code post-migration; rewriting and repointing extracts, transfers, and loads (ETLs); and testing downstream reports.

“Migrations are always hard,” said Steve McDowell, senior analyst at Moor Insights and Strategy. “And moving apps from your server to the cloud can be a terrifying proposition.”

Related

6 Core Challenges to Hybrid Cloud Implementation and How to Overcome Them

The potential for downtime is one reason, he explained. Additionally, the tools, processes and procedures developed for a data center environment must be adapted to the cloud or replaced with cloud versions. Migration teams must determine if monitoring, network security and other functions can be transferred seamlessly to the cloud or require an overhaul.

How an enterprise deals with these issues depends on its culture and staff expertise. Compliance requirements also play a role; for example, some industry and geographic rules mandate hosting data in an IT environment over which the enterprise has full control.

Another factor influencing the relative complexity and time required for cloud migration is how much data is involved. For a small application, a cloud migration might take a day. For a big enterprise resources planning (ERP) system, it can take much longer, said Chang.

He’s referring to the cutover to the cloud for an application, the last step of the migration for the specific application. 

“There will be continued testing once in production in the cloud,” he said. “Rolling back in case of issues may not be easily possible without data loss or time synchronizing changes back to the original on-prem location.”

Related

Framework for Choosing Private, Public and Hybrid Cloud IT

One time-burner is mapping all an app’s data dependencies to make sure that other apps and reporting functions remain connected when the cloud cutover happens. These requirements are being increasingly alleviated by container technologies that handle these functions behind the scenes, though enterprises require container expertise to make use of them. 

Sizing and performance are additional factors to consider; for example, determining the right sizing for a virtual server in the cloud versus a physical server in a data center.

Planning for Success

Cloud migration teams should include application developers, storage administrators, line-of-business stakeholders, financial analysts and legal personnel, Chang advised. Leaders should clearly define the business implications and anticipated benefits driving the migration, service-level agreements (SLAs) to shoot for, and a clear plan for everything from requirements gathering to testing and deployment.

Deciding what to do with existing core business applications is a strategic, critical step in any cloud migration. The primary options:

  • Retire them if they’re no longer needed or are too expensive to maintain in tandem with a cloud-based infrastructure
  • Completely rewrite or refactor (modify) them to run on a cloud platform
  • Rehome them, if possible, which involves simply “lifting and shifting” from one environment to another

Those companies that have elected to rewrite or heavily modify existing data center applications for use in the public cloud frequently encounter a few unwelcome surprises along the way. Among them are more time required than expected, complexity, cost and possible vendor lock-in issues.

Related

Cloud Smart in The Hybrid Era

While making application code changes prior to moving an application to the cloud, for example, developers should take care not to affect the external behavior of the application, according to experts. At the same time, developers should account for the CPU resources an application consumes. If it’s a high-utilization application and that attribute isn’t modified prior to migration, enterprises could wind up with an unexpectedly high cloud services bill, as public cloud billing models are based on usage.

Another conundrum is that the more “cloud-native” the application—meaning, the greater the number of the cloud provider’s services and features it integrates with—the more tightly the enterprise is beholden to that provider. That’s a benefit as long as the application stays where it is, enabling businesses to use the full richness of functionality their cloud provider has to offer.

Chang pointed out that implementing a good cost optimization practice still does give IT teams much control over public cloud providers increasing their costs.

However, in the “cloud smart” multicloud model, businesses are likely to want to move applications from time to time as goals and conditions change. Tight coupling makes it more difficult and expensive to extricate applications and data from one provider and move them to another public or private cloud platform, said Chang.

AWS now has 200 services,” he said. Integrating with all of them “represents strong and tight cloud lock-in.”

To minimize this situation, Chang advised using “lightweight” public cloud services, such as load balancing and storage, which he said are fairly generic across different cloud providers’ platforms and thus more easily portable. He also suggested “staying away from features that are unique” to a given provider, where possible.

On the benefits side, refactoring an application for the cloud can increase its resilience by decoupling applications components to take advantage of the public cloud’s highly distributed and redundant footprint. And, of course, enterprises gain the auto-scaling capability to grow or shrink their environment as business demands change and to pay only for the resources they use.

Companies that have attempted to move application workloads from the public cloud back to their data center because of costs or other factors have found that it’s far from an easy switchover. Deployments must be redesigned from scratch.

Moving from public cloud to private cloud is easier than moving on or off a legacy infrastructure. However, Chang said moving from public cloud to private cloud can be more difficult than going the other way if the migration team has not been careful to use fairly generic services that are available in private cloud environments.

“In some cases migrating back may actually be easier than going the other way if the applications were refactored into containers for example, as they are able to run anywhere quite easily,” he said.

Some issues to be mindful of in cloud migrations:

  • Security configurations moving when to a more public-facing environment

  • Application performance post-migration

  • Hidden fees with cloud for such tasks as data back-up snapshots and exporting data, given that everything is priced a la carte based on usage volume or time

Managing Expectations

One alternative to full cloud migration is available using public cloud provider-supplied infrastructure installed in the customer’s data center.

Chang said this approach can be costly and come with caveats, such as requiring continuous connectivity to the cloud provider to be fully functional.

“For example, AWS has on-site versions with its software stack on top of it,” said McDowell. “So do HP and Dell. You can buy capacity on demand without the complexity or risk of [the public] cloud.” Whether this becomes a major trend or a short blip in the Great Cloud Migration remains to be seen.

If a migration gets the green light, McDowell warns that it’s risky to overpromise because “you don’t know what you don’t know. The companies that do well are the ones with better lines of communication between IT and the business units,” he said.

The bottom line is that, although the benefits of cloud migration can be substantial, it can be complex and challenging to accomplish. For big companies, cloud providers are available to provide a lot of help. Smaller companies are challenged to find personnel and consultants with solid experience in the destination cloud environment(s) to help them migrate as efficiently and with the least disruption possible.

Large or small, however, businesses should seek a flexible IT architecture that assumes that the continual migration of applications across environments is the new normal.

Gene Knauer is a contributing writer who specializes in IT and business topics. He is also the author of Herding Goldfish: The Professional Content Marketing Writer in an Age of Digital Media and Short Attention Spans.

© 2022 Nutanix, Inc. All rights reserved. For additional legal information, please go here.