February 2026 Summaries

Initially trained in electronics, the author reflects on the evolution of programming languages and compilers, highlighting the shift from low-level assembly language to high-level languages like Python, facilitated by compilers that abstract complex machine instructions. The text draws parallels between historical advancements in software abstraction and the emerging role of AI as a compiler, suggesting that AI can transform natural language into code, further raising the abstraction level. While traditional compilers optimize for human readability and ergonomics, the author notes that future languages might prioritize compilation speed, portability, and performance due to AI's involvement in code generation. Despite the potential to bypass human-readable languages altogether, the author argues for maintaining abstraction layers, as they provide valuable benefits like portability and debuggability. The text concludes with a prediction that future developers will focus on outcomes and user experiences while AI compilers manage the technical details, continuing the tradition of trusting underlying systems to accurately translate human intent into machine execution.

Eurofiber Cloud Infra faced a significant challenge when rebuilding their cloud infrastructure, needing to swiftly inventory components from network backbones to virtual machines while establishing a foundation for consistent service automation. Initially considering a combination of NetBox, legacy DCIM solutions, Terraform, and Ansible, they encountered issues with inflexibility and integration fragility. Instead, they opted for Infrahub, which offered a more flexible schema and Git-native architecture, allowing for seamless automation and inventory management. Infrahub became the central source of truth, enabling rapid service deployment and simplifying infrastructure management, significantly reducing deployment times from days to minutes. The implementation not only streamlined provisioning but also enhanced lifecycle management, including upgrades and decommissioning, with integrations that automate configuration backups. The GitLab CI/CD pipeline further facilitated the process, proving the efficiency and transformational impact of Infrahub's automated workflows on Eurofiber's operational efficiency.

Infrahub Profiles offer a novel solution to the persistent challenge of managing infrastructure and network standards at scale by maintaining a dynamic connection between standards and the objects that inherit them, eliminating configuration drift and the need for manual updates. Unlike traditional methods that treat standards as static snapshots, Infrahub Profiles allow for the automatic propagation of updates, separating the static physical layer from the evolving logical layer, which facilitates the seamless evolution of infrastructure designs. This separation enables different contexts to be applied to the same physical device template, and exception management is treated as a first-class concept, making deviations visible and governable. Infrahub Profiles support a variety of modern, multi-vendor environments by working at the schema level, thus providing flexibility in rendering configurations for diverse platforms without requiring multiple definitions of the same standard. Additionally, the Profiles enable structured change management and audit trails, ensuring infrastructure standards are maintained consistently and can evolve gracefully over time, thus turning complex, error-prone processes into streamlined, automated workflows.

Infrastructure teams often struggle with traditional data management tools due to schemas that are either too generic or too rigid, leading to issues with validation and adaptability. Generic schemas attempt to accommodate all possible use cases, resulting in optional fields and unreliable data, while rigid schemas, deeply integrated into database systems like NetBox, make necessary updates risky and cumbersome. Infrahub addresses these challenges by offering a schema system that is both well-defined and flexible, allowing for strict validation and adaptability without breaking integrations. Infrahub utilizes a graph database, decoupling the schema from the core platform, which enables schema evolution without heavy migration operations. It employs polymorphism to define base types and specialized types, ensuring accurate data modeling and reliable automation. By using YAML for schema definition and integrating version control, Infrahub facilitates safe schema modifications and collaboration, allowing infrastructure teams to start small and expand as needed without disrupting existing systems.

Infrahub Generators are designed to automate repetitive network deployment tasks by allowing users to define logic once and reuse it for efficient resource provisioning. This system eliminates manual steps by automatically creating necessary objects, connecting them, and allocating resources based on predefined logic captured in Python code. An Infrahub Generator interprets a high-level request into a technical implementation, such as provisioning a dedicated Internet circuit or rolling out new branch offices, ensuring idempotency by updating existing objects and avoiding duplicates. The process involves creating a generator definition, targeting specific object groups, and utilizing GraphQL queries to fetch input data. A practical example demonstrates the automated provisioning of network sites, where a Generator allocates IP prefixes, creates subnets and firewalls, and assigns management IPs without manual intervention. This approach streamlines network deployments, as showcased by Otter Bank's efficient rollout of new branches, and can be extended to more complex scenarios with additional logic for interfaces and routing configurations.

Infrahub streamlines network automation by using GraphQL, a schema language developed at Facebook, to efficiently handle interconnected data in infrastructure settings. Unlike traditional REST APIs that require multiple calls to gather data, GraphQL allows users to specify exactly what data they need in a single request, reducing latency and unnecessary data retrieval. Infrahub's infrastructure data is stored in a graph database, which suits GraphQL's design for querying complex, interconnected data. This approach allows automatic generation of GraphQL queries and mutations based on user-defined schemas, eliminating the need for manual API code maintenance. The dynamic nature of GraphQL provides built-in documentation and type safety, facilitating error detection before deployment. It is particularly advantageous for AI applications, as it offers machine-readable documentation and enables AI agents to automatically discover and query the necessary data. Infrahub's GraphQL engine supports efficient configuration generation, topology mapping, compliance validation, CI/CD integration, and audit and troubleshooting tasks, all while maintaining a single endpoint for all queries, which simplifies the process and enhances security through authentication and network access controls.

A network source of truth (NSoT) is crucial for maintaining order and clarity in automation, but when its architecture lacks healthy boundaries, it can lead to significant operational issues. Symptoms of an unhealthy relationship with an NSoT include the interdependence of components leading to widespread disruption from minor changes, the need for excessive testing and reassurance during updates, difficulty in supporting new functionalities without risking system integrity, stagnation due to fear of upgrading, and a loss of trust in data integrity. These issues stem from a lack of separation between schema and application logic, reliance on undocumented schema internals, and a fragile customization of the schema. A healthy NSoT should feature decoupled architecture, composable extensions, stable upgrade paths, schema independence, and reliable data, which allows for growth and adaptability without compromising system stability. Choosing platforms designed for change rather than control is essential for building a robust infrastructure.

Databases serve as the essential storage backbone for implementing schemas in network automation, with the choice of database significantly impacting performance, scalability, and flexibility. Various types of databases optimize for different aspects: relational databases emphasize strict data integrity and consistency but can struggle with schema flexibility and complex queries; key-value store databases prioritize speed, sacrificing relationships and query capabilities; document store databases offer a flexible schema with moderate enforcement, suitable for applications needing more structure than key-value but more flexibility than relational databases; graph databases excel in handling highly relational data with efficient relationship traversal, offering linear performance and schema flexibility ideal for network topology; and time-series databases are optimized for timestamped data, excelling in metrics storage and aggregation but lacking in schema enforcement and relational capabilities. The choice between stateful schemas, which are tightly coupled with specific databases, and stateless schemas, which offer flexibility across different storage solutions, further influences data management strategies. This decision is crucial in network automation, where graph databases are increasingly recognized for their ability to efficiently manage complex, evolving network structures.

Understanding data schemas is crucial for network automation as they define the organization, interpretation, and constraints of data, ensuring consistency and facilitating communication between systems. Various schema types, such as SQL, JSON Schema, YANG, GraphQL, and Infrahub, offer different strengths suited to specific use cases, from traditional relational databases to flexible APIs and network device configurations. While flexibility in schemas allows for rapid adaptation and development, especially in small projects or evolving data models, it can lead to data inconsistencies as systems scale. Conversely, enforced schemas provide data integrity and consistency, essential for reliable network automation in production environments. Schemas universally consist of three components: structure, relationships, and constraints, and understanding these is vital for selecting the right schema approach based on specific requirements rather than trends. Moreover, as systems mature, incorporating schema enforcement is necessary to prevent data inconsistencies, underscoring the importance of aligning schema choices with the nature of the data and the intended use within network automation.

High upfront costs aren't the only factor to consider in technology decisions, as platforms with quick deployment but high maintenance needs can hinder innovation and become burdensome over time. A UK-based financial services company discovered that 80% of their engineering efforts were occupied by maintaining tools rather than building new automation capabilities, a result of choosing platforms for immediate benefits rather than long-term viability. This issue is compounded by architectural coupling, which ties a platform's schema to its feature set, making upgrades risky and resource-intensive. This maintenance trap is evident in organizations like OpsMill's clients in aerospace and retail, who are stuck on outdated versions of heavily customized platforms due to the high testing burden required for upgrades. Sustainable Network Source of Truth (NSoT) platforms should feature decoupled architectures where the data model is independent of platform versions, allowing for upgrades without breaking integrations or creating bottlenecks. This approach preserves engineering capacity for innovation rather than maintenance, and Infrahub is presented as a solution that offers a decoupled architecture to reduce maintenance burdens and facilitate growth without technical debt.