Beyond CI/CD: Foundational Platform Delivery

Rethinking Delivery for Core Infrastructure

Ian Nowland's presentation offers a compelling argument for why traditional CI/CD practices, widely adopted and effective for product teams, fall short when applied to foundational platforms. His core thesis—that statefulness, deep infrastructure dependencies, irreversible changes, broader blast radius, and inherent complexity make testing and deployment fundamentally different for platforms—is well-supported by his extensive experience at AWS and Datadog. The identification of common pitfalls, such as relying on feature flags for complex platform components or blindly adopting DORA metrics without context, provides valuable guidance. The emphasis on building platforms to abstract complexity, and how this very abstraction creates unique CD challenges, is a nuanced and important distinction. The actionable techniques for safe progressive deployments and synthetic testing in production are particularly noteworthy, offering practical solutions beyond theoretical discussion.

However, while the presentation effectively highlights the problems, the depth of the 'solutions' could be further elaborated. The talk touches upon progressive deployments and synthetic testing, but a more detailed exploration of how to implement these safely and effectively in complex, stateful environments would be beneficial. For instance, specific strategies for managing state during rollbacks, techniques for creating realistic synthetic tests for highly distributed and stateful systems, and concrete examples of mitigating blast radius beyond general principles would enhance the practical value. The discussion on 'bad tech' and 'rewriting everything' is a common sentiment but could benefit from more nuanced advice on identifying when a rewrite is truly necessary versus when refactoring and incremental improvement are more appropriate for foundational platforms. The presentation serves as an excellent primer and a call to action for platform teams to critically re-evaluate their delivery strategies, but the implementation details of the proposed solutions warrant deeper dives.

Key Points

• Conventional CI/CD practices are ill-suited for foundational platforms due to their stateful nature, critical business impact, and inherent complexity.
• Foundational platforms abstract complexity, which paradoxically makes them harder to test and deploy safely using standard application-centric methods.
• Common misapplications include over-reliance on feature flags, naive use of DORA metrics without context, and the assumption that staging environments are universally effective.
• Key challenges for platform teams include managing statefulness during deployments, mitigating the broad blast radius of failures, and the difficulty of accurately testing complex systems outside of production.
• Actionable techniques involve safe progressive deployments, robust synthetic testing in production, and a nuanced understanding of when and how to apply metrics like DORA.

---

📖 Source: Presentation: Continuous Delivery for Foundational Platforms