Coming Home – Key Considerations for Bringing Your Applications Back On-Premise

By Kristina LeBlanc, Staff Writer

“Cloud First” is no longer the assumed school of thought in IT circles. After years of migrating applications to the cloud in a mob mentality, enterprises are beginning to swing toward what analysts are calling “cloud repatriation,” where applications and workloads that had been dispatched to the cloud are coming back to on-premise datacenters.

“If an organization is gathering and processing consumers’ personal information, they’ll likely want to keep this workload on-premise, or at least in a private cloud,” says Ali Zaidi, research director at IDC.

In their 2018 survey, International Data Corp. (IDC) found that over the preceding 12 months 80 percent of respondents had moved public cloud workloads back on premise or to a private cloud. Survey respondents also noted that they planned to migrate an additional 50 percent of their current public workloads back on-premise or to a private cloud by 2020 (1).

There are several factors driving this trend, according to Jeffrey Zarse, Product Manager at Atos North America.

“Factors driving the shift back to on-premise or private cloud include cost concerns, as well as several high-profile cloud service provider outages that are leading some enterprises to conclude that on-premise infrastructure provides more direct visibility and control over performance management,” he says. “Compliance and regulatory concerns for sensitive data are also key drivers.”

Zarse adds there is a time and a place for application repatriation, but like any migration there are substantial risks and costs involved, including unplanned downtime, that are worth assessing.

So when is the right time to make a move?

Thoroughly Audit Your Applications

In retrospect, a fundamental flaw of the Cloud First approach was that it emphasized platforms over customer and application needs. When considering a return home, it is important to analyze your applications portfolio. What is the ideal platform to support a given application while keeping costs down and enabling cost-savings to be passed on to customers while meeting the application’s speed, availability, and other unique performance requirements?

As a rule of thumb, if an application demands a high level of security and reliability and is competitively differentiating for the business, run it on premise, according to Zarse. Some enterprises believe the public cloud doesn’t offer the same standard of security and compliance controls that an on-premise or private cloud infrastructure does, especially as cloud infrastructures become high-value targets for hackers.

Additionally, applications that have steady workloads, require fast speeds and transfer large amounts of data often run more cost-effectively on premise than in the cloud, since elastic scalability may not be needed and because network-related costs can eat up a significant portion of a cloud bill. “An example would be a mobile banking app that allows smartphone users to make a deposit by snapping a photo of their check,” Zarse says.

On the other hand, some applications are optimally suited to the cloud. These generally are undifferentiated applications such as HR, which don’t carry the same security and performance requirements, have unsteady workloads and can be easily managed through a single portal. Such applications also often require re-provisioning to accommodate new workloads.

Plan and Execute Your Migration Carefully

The journey back from the cloud can resemble a maze. Most migration tools and services—including those offered by cloud service providers—are designed for one-way travel, as opposed to a round trip. There are options, however, according to Zarse. Perhaps the easiest is to restore through back-up by using disaster recovery (DR) replication tools that sync cloud-based virtual machines to the internal datacenter.

“Once an app has been determined as a good fit for a return trip, organizations still must conduct a careful cost-benefit analysis,” he says. “Getting rid of monthly cloud bills and the associated sticker shock can be a great feeling, and many organizations are relieved to replace these with more predictable controlled costs based on assets under management which are paid in advance.”

However, Cassandra Mooshian, senior analyst in TBR’s cloud practice cautions, “Don’t let an initial sticker-shock lead to an automatic knee-jerk reaction of pulling applications out of the cloud. Sometimes, there are ways to better configure workloads in order to bring costs down.”

It’s also important to remember that pulling applications back from the cloud will require additional costs, including manpower. Cloud adoption often results in IT staff reductions, so a reverse move may require organizations to re-staff or pull existing staff off other projects, as well as resume ownership of software licensing costs.

“Before pulling applications back on-premise, it’s critical to make sure you, or any outsourcers you may be using to manage these workloads, have the necessary skills and expertise,” continues Zaidi. “Then, of course, there are the migration tools themselves, which come at a price.”

Realize the Decision is Not a Zero-Sum Game

It’s highly unlikely that any one platform will offer the full span of functionality you need to support all the various applications and components in your portfolio,” concludes Zaidi. The decision to stay in the cloud or return to an on-premise datacenter isn’t a black-or-white, either/or proposition. Both platform options offer unique attributes that can be ideally suited to support multi-platform, componentized applications. An example of this is an eCommerce app that requires a highly scalable, customer-facing front-end system of engagement (the cloud), coupled with a secure and reliable back-end system of record (on-premise).

This type of hybrid approach is gaining significant traction, and while it enables organizations to match various application components with the most ideal, cost-effective and appropriately performing platform resources, there can be risks, Zarse adds. An end-to-end application is only as strong as its weakest link. So while the cloud-based front-end may have superior scalability for higher than normal traffic loads, the holistic application will still break if the infrastructure is constrained on the back-end.

“This is likely the reason why we continue to see major online retailers experiencing crashes on peak traffic days, even in the age of the cloud and instant scalability,” Zarse says. “Ideally, end-to-end infrastructures should be designed to deliver relatively uniform levels of capacity throughout.”

When it comes to keeping applications in the cloud or bringing them back on premise, the decision ultimately should boil down to what makes the most cost-effective sense for individual applications from a performance perspective. The government sector aptly describes this attitude shift as “Cloud Smart,” an evolution of the Cloud First strategy.

Across industries, adopting a Cloud Smart strategy will inevitably involve organizations realizing that some applications they enthusiastically ported to the cloud a few years ago weren’t an ideal fit after all. Bringing these applications and workloads back on-premise can present many benefits and is achievable, as long as organizations are strategic in their planning and methodical in their execution.

  1. IDC Survey, Cloud Repatriation Accelerates in a Multicloud World, Doc # US44185818, August 2018