ABSTRACT: This blog offers five guiding principles to help CIOs, CDOs, and team leaders optimize hybrid data environments.

Read time: 4 mins.

Sponsored by Exasol 

While most companies embrace the cloud for agility, they also get value from residual systems on premises. And the rising challenges of cloud computing—cost, security, compliance, sovereignty and performance—now force them to rethink their options. The result is a balancing act in which innovative companies continually reassess how to optimize their hybrid environments.

This blog, the third and final in our series, explores how data leaders can optimize hybrid environments. The first blog assessed the tradeoffs of migration and repatriation, and the second blog applied these tradeoffs to AI projects. Now we define how chief information officers, chief data officers and team leaders should decide when and where to move workloads. We offer five principles for success: take a holistic view, internalize stakeholder needs, consider the economics of change, keep it simple and always ask why.

First let’s review the tradeoffs of migrating, repatriating and maintaining the status quo. Data and AI leaders can:

• Migrate to the cloud to increase flexibility of infrastructure, finances and AI tooling
• Repatriate back on premises to increase control of data, cost and compliance
• Maintain the status quo by keeping data and workloads in place to avoid the risks of change

Migrate, Repatriate or Maintain?

They must consider the familiar dimensions of people, process and technology. Significant changes in any of these areas should prompt you to revisit your options.

Dimensions to Consider

• People. When you change key people, you might need to change the cloud/on prem mix to fit their skills. This includes business owners, end users and technical specialists such as ITOps engineers, data scientists, ML engineers, data engineers or developers.
• Process. When these stakeholders start, stop or change projects, they might need to spin up a new cloud sandbox or deploy on-prem servers to meet specific requirements. Company partnerships, mergers or acquisitions also can force teams to accommodate a new entity’s infrastructure.
• Technology. Applications, tools, models and datasets tend to work well on some platforms but not others. When you introduce new elements like these, give careful thought to the right supporting platform.

Principles

So when something changes, how do you evaluate your cloud/on prem options? These principles can guide your decisions.

Take a holistic view. Consider all dimensions in a holistic fashion. If your smart new data scientists demand a certain cloud-based AI/ML platform, consider how your data engineers would prepare on-prem data to create model features for that platform. If your newly acquired business unit maintains its proprietary data center for regulatory reasons, consider how to integrate that data with the rest of the environment. For example, your data team might need new APIs, virtualized views or pipelines that export only obfuscated data.

Internalize stakeholder needs. Be sure to internalize the skills and preferences of your stakeholders. With each migration or repatriation, you must recruit the right people, speak to their needs and convert them into internal champions. You also must train them on the necessary skills, for example with centers of excellence that showcase and propagate best practices across the enterprise. By winning the hearts and minds of talented people you can make team efforts successful, for example by driving adoption of new tools, techniques or platforms.

Consider the economics of change. The bottom-line reason to migrate, repatriate or maintain the status quo is price-performance. For example, your ITOps and FinOps engineers might determine that they can meet latency, throughput and uptime SLAs for a stable application more cost-effectively on premises. But those savings must exceed the cost of repatriating that application and underlying data back to on-premises servers and storage. To confirm this, your data and DevOps engineers should help calculate the level of effort required to refactor the application code, transfer the data and build a new test environment on premises. And the FinOps engineer should take a hard look at any egress fees from the cloud service provider.

Keep it simple. Any migration or repatriation effort involves significant complexities and interdependencies. This makes the status quo option pretty compelling because it minimizes risk. Given this reality, data and IT teams should be vigilant but careful. They should monitor all the dimensions described here to identify opportunities for change—but only act when the opportunity is compelling. You should only migrate or repatriate when the price-performance benefits are high and the risks are low. The most successful teams tend to make a few high-impact, low-risk decisions.

Keep asking why. Nearly 20 years after the launch of AWS, cloud computing continues to spark enthusiasm. And no wonder: cloud platforms improve agility and cultivate powerful AI innovations. But during a time of hype, data and AI teams need to question conventional wisdom. Why might you really need a cloud AI platform? Perhaps your customer service team can get what it needs for now by installing a GenAI language model inside its own firewall and manually prompting it for advice. Why might you need a cloud lakehouse? Perhaps your data warehouse on premises offers better price-performance for your basic BI projects. Whatever the situation or opportunity, be sure to ask why and consider how best to leverage your existing resources.

Look to the cloud… but keep your feet on the ground

This final blog in a three-part series offers five guiding principles to help CIOs, CDOs, and team leaders optimize hybrid data environments as they assess the tradeoffs of cloud migration, repatriation, and maintaining the status quo. Leaders should take a holistic view of people, process and technology, internalize stakeholder needs, evaluate the true economics of change, simplify decisions, and continually question assumptions. By applying these principles, organizations can make deliberate, well-informed decisions that align infrastructure with evolving business and AI requirements. To learn more about this topic, be sure to check out this blog by Exasol.