6 App Migration Tips for AI Modernization in the Hybrid Cloud Era -- Virtualization Review

6 App Migration Tips for AI Modernization in the Hybrid Cloud Era

By David Ramel
08/14/2025

As organizations move toward AI integration in hybrid cloud environments, modernizing application stacks often means tackling the challenge of migration. In the first session of the today's "Future-Proofing Your App Infrastructure in the AI-Ready Hybrid Cloud Era" summit, technology author and Microsoft MVP Brien Posey outlined the six main approaches to application migration and how they apply to AI workloads in hybrid deployments in his session, titled "Expert Take: Modernizing App Stacks for AI in a Hybrid Cloud World," now available for on-demand replay.

"Modernization isn't just about making your application run more quickly or adding AI to your application. It's really about future proofing."

Brien Posey, Freelance Author, 20x Microsoft MVP, Commercial Astronaut Candidate

Why Migration Matters for AI
Legacy application stacks may lack the GPU support, scalability, and orchestration capabilities required for AI workloads. Posey noted that AI integration is most effective when supported by architectures that can operate seamlessly on-premises, in public cloud, or across both. This flexibility allows workloads to run where they make the most sense -- for example, training large models in the public cloud to leverage elastic GPU resources, while running inferences locally to reduce costs and latency.

The 6 R's of Application Migration
Posey described the six common migration strategies, all starting with the letter "R." While variations exist, the general concepts are:

Rehost
Often called "lift and shift," rehosting moves an application to a new environment with minimal or no changes to its architecture or configuration. Posey explained that this is essentially "picking up an application and moving it to a new environment, making extremely minimal changes along the way." A common example is migrating a VMware virtual machine to AWS EC2 without altering the VM's contents. This approach is generally the quickest way to relocate workloads, but it does not take advantage of the capabilities of the target platform. For AI workloads, rehosting may be appropriate as an interim step, especially when the goal is to change infrastructure quickly while keeping functionality intact. However, because it preserves the original architecture, it may not address limitations in scalability, GPU access, or orchestration support.

Replatform
Sometimes referred to as "lift, tinker, and shift," replatforming is similar to rehosting but allows for small optimizations to take advantage of the target platform's features. Posey described this as making "really small optimizations along the way as a way of taking advantage of the new platform," without changing the application's core functionality. For example, an organization moving a database to the cloud might switch from hosting it in a VM to using a managed database service, offloading patching and maintenance to the provider. This approach can improve performance, resilience, and maintainability without the complexity of a full redesign. In AI contexts, replatforming can enable better integration with platform-native AI tools or improve storage and compute performance for model training.

Repurchase
Also known as "drop and shop," this strategy replaces an existing application with a new one that better meets business needs. Posey noted that this could involve moving to SaaS --for example, adopting Salesforce in place of a custom CRM --or simply replacing software that no longer fits the organization's requirements. "There might be a new application that has some features that you'd love to take advantage of," or the change might be driven by licensing shifts, such as a vendor moving from perpetual licenses to subscription-only models. For AI-readiness, repurchasing can bring in AI-enabled SaaS solutions without the effort of building AI capabilities in-house, but it requires careful planning for data migration and integration with other systems.

Refactor
Refactoring involves a full redesign of the application to leverage the capabilities of the new platform. Posey described it as "the most complex and typically the most expensive" of the six R's, but often the most relevant for AI modernization. This might mean breaking a monolithic application into microservices, enabling event-driven architectures, or restructuring data flows to better serve AI models. By decoupling components and adopting containerization and orchestration, organizations can achieve portability across hybrid environments and improve scalability for AI workloads. While refactoring demands significant time and resources, it enables integration of AI modules as independent services, simplifies updates, and supports real-time responsiveness to AI-driven predictions.

Retire
This option decommissions applications that are no longer useful, cost-effective, or strategically aligned. Posey noted it can also reduce overlap: "Let's say you've got three different applications that all kind of overlap with each other. Maybe you can retire one of those applications, thereby helping to reduce cost and complexity." Retiring unneeded systems can free up resources for AI-focused modernization, streamline the application portfolio, and reduce security risks associated with maintaining outdated software.

Retain
Sometimes labeled "revisit later," this strategy postpones migration for specific applications, keeping them in their current environment until conditions change. Posey explained that it means "don't do anything right now. Let's revisit in six months and just see if things change or if there's still a need to do something." Retaining can be useful when an application is stable, low-risk, and not currently impeding AI adoption. It allows IT teams to focus on higher-priority workloads while monitoring whether future requirements --such as compliance, scalability, or AI integration --justify action.

Refactoring for AI-Ready Hybrid Cloud
While all six R's may be used in different scenarios, Posey emphasized that AI-focused modernization often centers on refactoring. This allows AI services to be integrated as discrete modules rather than deeply embedded in monolithic code. Combined with containerization and orchestration, this modular approach supports portability between on-premises and cloud environments, improves resilience, and simplifies updates.

In hybrid cloud AI deployments, refactoring also facilitates event-driven architectures that can respond to predictions in real time, an important capability for industries like financial services where latency and compliance are critical.

Strategic Considerations
Migration strategy selection depends on factors such as business use case, application architecture, and resource availability. For AI workloads, considerations include data locality, GPU/TPU access, container orchestration, and the ability to move workloads to the environment that best matches their compute and compliance needs. As Posey noted, modernization should be driven by clear business goals, not technology trends alone.

And More
Posey had a lot more expert advice to share on this topic and offered much discussion about many other apects in his complete presentation, so an on-demand replay is definitely in order. And, although replays are fine -- this was just today, after all, so timeliness isn't an issue -- there are benefits of attending such summits and webcasts from Virtualization & Cloud Review and sister sites in person. Paramount among these is the ability to ask questions of the presenters, a rare chance to get one-on-one advice from bona fide subject matter experts (not to mention the chance to win free prizes -- in this case a $300 AMEX gift card from sponsor A10 Networks, which also presented at the summit).

With all that in mind, here are some upcoming webcasts coming up from our parent company in the next month or so: