May 2024 Summaries

In this article, we discuss the importance of backend design for Internal Developer Platforms (IDPs) and explore two common patterns in their architecture: pipeline-based and graph-based backends. The main value drivers of platform engineering are standardization and automation. A well-designed backend should promote standards, manage the entire lifecycle of a resource, be API-first, interpret abstract requests from users to enable golden paths, allow users to leave the path when needed, design for easy integration, ensure security, have a workflow engine built in, and maintainability. Pipeline-based backends use CI pipelines and IaC but are not ideal for complex business logic in the infrastructure world. They can be difficult to audit and consume with multi-interface approaches. Graph-based backends, such as Platform Orchestrators, provide a dynamic approach by describing dependencies between nodes (resources) based on their edges (dependencies). This allows for easier updates and better standardization compared to pipeline-based designs. However, they may require more time for infrastructure platform engineers to understand the logic and approach.

The Humanitec Platform Orchestrator enables enterprises to ship software faster and more securely by allowing developers to self-serve infrastructure without accessing secrets. It uses a graph-based backend for defining rules on how developers can consume resources, with Placeholders used in resource definitions instead of hard-coded secrets. The Humanitec Operator, which sits inside K8s clusters, complements the Platform Orchestrator and securely completes rendering procedures without accessing external secret stores. This ensures that secrets are never shared outside their designated locations, maintaining control over them even if other systems become compromised.

Platform engineering, which focuses on enabling developer self-service and reducing cognitive load, is transforming infrastructure and operations (I&O) teams' daily lives. It goes beyond just improving the developer experience; it also addresses I/O professionals' pain points. By using an Internal Developer Platform (IDP), platform engineering can help align these pain points to provide value for all. Infrastructure platform engineers ensure that IDPs interface well and sustainably with the work of I/O teams, while DevEx platform engineers focus on developer experience. The backend of a platform is usually a graph-based orchestrator that reads abstract requests and matches them with the correct resource packs. This approach allows for global rules to be set for how resources get created, enabling I/O teams to work proactively and streamline their internal processes.

The article discusses the four phases of the Minimum Viable Platform (MVP) framework developed by Humanitec after working with numerous enterprise organizations. These phases include Discovery, Integration, Deployment, and Prep for Adoption. The MVP is a subset of the reference architecture that focuses on key components to ensure it represents real requirements and can clearly show value to users and stakeholders. It should not be compliant with all security and policy requirements or involve advanced architectures or configurations. The Discovery phase involves aligning platform initiatives with broader business objectives, defining scope and boundaries, and identifying pain points for developers and ops teams. The Integration phase sets up the organization within Humanitec, connects cloud providers and Kubernetes clusters, and creates resources defined in phase one. In the Deployment phase, resource definitions and matching are set up, values and secrets are defined, CI/CD pipeline is established, and developer interfaces are set up. The Prep for Adoption phase measures the impact of the MVP on workflows and calculates ROI for a final leadership-level presentation. The article emphasizes that the MVP program helps organizations move quickly and avoid losing momentum on their platform engineering initiatives while ensuring they have a clear framework to prove value to stakeholders.

Adoption of containers and container orchestrators is growing rapidly, leading to a need for safer ways to handle application and infrastructure configurations. GitOps emerged as an approach to prevent configuration mistakes by storing the state of apps and their dependencies in code outside the application itself. Weaveworks open-sourced FluxCD, followed by ArgoCD, but confusion around definitions and best practices led to uncertainty in the industry. However, consolidation and adoption of platform engineering have made GitOps ready for enterprise prime time when combined with an Internal Developer Platform (IDP). GitOps is a framework that applies version control and CI/CD best practices to infrastructure automation in a declarative way. It has four key principles: Declarative, Versioned and Immutable, Pulled Automatically, and Continuously Reconciled. Combining GitOps with platform engineering can standardize app and infra configurations, establish separation of concern between devs and Ops, and improve security and compliance for large enterprises.