“Just rewrite it”: What platform teams really think about modernization

Mergers, acquisitions, and business initiatives have led to an unfortunate reality: redundant infrastructure and expertise. Typically, platform engineering teams manage cloud-native and Kubernetes workloads, while traditional IT manages virtual machine (VM) workloads. Despite organizations' efforts to standardize on Kubernetes, many still maintain a significant VM footprint. This coexistence is often a permanent operating model rather than a temporary transition.

On-premises environments require separate networking, servers, and storage for each environment, which is structurally less cost-efficient than consolidation. VM-based mission-critical workloads are unlikely to disappear anytime soon, and teams are eager to modernize but find the transition challenging. The separation of these worlds leads to extra infrastructure, duplicated expertise, slower projects, mixed governance, and ballooning budgets.

The notion of "just rewrite" doesn't solve modernization issues. Finance and executives often favor this approach, unaware of the hidden costs. Rewriting doesn't necessarily deliver business value and can be financially, risk-wise, and timeline-wise detrimental. While re-platforming and re-architecting legacy applications are popular, they often result in square one and functional parity. It's more practical to keep mission-critical applications as-is while scaling out cloud-native platforms for new value.

AI can speed up coding, but the real expense lies in decision-making, verification, and data migration. Rewriting involves shifting from software writing to dealing with complex decisions, validation, and migration. Despite AI's advancements, the costs of rewriting remain human-driven. Organizations still need software engineering expertise to design target systems, a task that has become increasingly complex due to the shift towards microservices and cloud-native applications.

The gap isn't philosophical, but operational. Platform teams must deliver services at expected velocities; otherwise, developers may blame the platform. VM environments rely on UI-driven tools, while cloud-native environments require command-line interfaces (CLIs) and configuration files. Transitioning from UI to CLI can be challenging without a significant team capability shift.

Cloud-native workloads require databases, object storage, and file services, which must operate at cloud-like speeds. While containers are often perceived as stateless, most workloads have state, such as data, logs, metrics, or dependencies, that must be managed consistently.