July 2026 Summaries

The text discusses the limitations of traditional Network Source of Truth (NSoT) databases in managing modern, complex, and dynamic network infrastructures, arguing that they often become static records rather than proactive tools for network management. It advocates for a shift towards "Intent-Driven" infrastructure, where infrastructure management is based on a source of intent that defines desired future states using declarative code, similar to modern software engineering practices. This approach allows for proactive management, testing, and validation of infrastructure changes, enabling more efficient and error-free deployment. The text highlights Infrahub as an example of a platform designed to implement infrastructure intent, offering a version-controlled, schema-first architecture that adapts to business needs and facilitates automation and compliance without the limitations of traditional NSoT systems. A case study of a Dutch fintech is mentioned, emphasizing the practical benefits of using such a platform to save time and reduce errors, especially in achieving complex AI and automation goals.

Opsmill argues that traditional Network Sources of Truth often function mainly as inventory or historical record systems, documenting what infrastructure exists or what changed rather than defining the desired future state of a network. It distinguishes sources of record, inventory, and intent, presenting infrastructure intent as a version-controlled, declarative approach that allows teams to model, validate, test, and approve network designs before deployment. The text contends that this approach is particularly useful for greenfield data centers, where no physical infrastructure yet exists, and for managing complex environments such as AI clusters, cloud overlays, and edge fabrics. It positions Infrahub as a schema-first, graph-based source-of-intent platform with native branching and automated validation, intended to support collaborative automation, compliance checks, adaptable data models, and more reliable infrastructure change management.

An AI data center fabric is a complex network design comprising numerous switches, interfaces, and topological layers, which requires consistent addressing across pods and racks. The goal for AI infrastructure providers is to bring the fabric online quickly to accelerate revenue, relying on a repeatable delivery model to expand capacity without scaling engineering efforts proportionally. Design-driven automation, facilitated by Infrahub, manages the fabric through its lifecycle by storing design and implementation as structured, versioned data, which allows for generating, validating, and merging changes while keeping the context of the generated infrastructure. Infrahub's open-source AI Data Center reference solution provides a practical approach to model and deploy an AI data center fabric using structured design intent and modular Generators, which translate design into technical implementation. This approach supports initial builds and day-two changes without requiring separate processes for expansion, enabling multi-site delivery with consistent standards across sites. The reference solution offers pre-built schemas, modular Generators, and configuration templates for Cisco, Arista, and Dell, allowing users to evaluate or adapt the project to their specific needs, thereby demonstrating how structured design, modular generation, and controlled change can work together in AI data center workflows.

Design-driven automation for AI data center fabrics aims to manage not only initial switch configurations but also the topology, addressing, standards, resource allocations, and change history behind large, evolving networks. Infrahub supports this model by storing design intent and generated infrastructure together as versioned, structured data, enabling teams to generate configurations, validate and review changes, coordinate shared IP and ASN resources, and deploy through existing tools such as Ansible or Nornir. Its open-source AI Data Center reference design demonstrates a modular approach for five-stage Clos fabrics, using Fabric, Pod, Rack, and overlay Generators to create physical infrastructure, EVPN/VXLAN services, configurations, and cabling plans for Cisco, Arista, and Dell environments. The same design model supports Day Two operations such as adding racks, pods, tenants, or updated standards through branch-based changes, CI validation, diff review, and controlled merging. The reference implementation includes schemas, generators, resource pools, templates, sample data, and setup tooling, offering both an evaluable demonstration and a customizable foundation rather than a prescriptive or certified network design.

Large, complex networks and data center infrastructures present significant challenges for automation, necessitating advanced data management solutions that encompass the intricate mesh of relationships among components, business logic, and design intent. Traditional data models like relational databases and document stores fall short in capturing these complex interdependencies, which are crucial for maintaining automation efficacy. Knowledge graphs offer a promising solution by representing real-world entities and their relationships in a graph format, providing semantic context, flexible schemas, and interconnected data that facilitate logical deduction and reasoning. This approach allows AI systems to interpret and manage infrastructure data effectively, supporting intelligent automation and decision-making. Infrahub, built on a knowledge graph model, exemplifies this paradigm by offering a flexible, reliable, and scalable platform for infrastructure data management, enabling teams to model infrastructure as a dynamic system with accurate, queryable, and metadata-rich data that integrates seamlessly with automation workflows. By adopting knowledge graphs, organizations can transform complex data into a manageable and understandable model, paving the way for AI-assisted operations and enhanced automation capabilities.

Large AI data centers and other complex infrastructure environments contain vast numbers of devices, connections, configurations, and layered dependencies, making automation difficult when data management captures only individual components rather than their technical, business, and design-intent relationships. Traditional relational databases, document stores, and file-based repositories can struggle to preserve consistency and model flexible, interconnected infrastructure data across vendors, legacy hardware, physical and virtual layers, and changing services. Knowledge graphs address this by representing entities as nodes and their relationships as contextual, metadata-rich edges, enabling users and systems to query dependencies, assess the impact of failures or changes, and support inference about infrastructure intent. The text argues that graph-based models are especially important for AI-driven automation because agents require structured context about ownership, dependencies, and service impact to make reliable decisions. It presents Infrahub as a graph-native platform that combines flexible schemas, relationship mapping, immutable versioning, data synchronization and lineage, and multiple access interfaces to create an integrated foundation for scalable automation and AI-assisted operations.

Over the last decade, IT leadership has focused on automating network operations through a Network Source of Truth (NSoT) to enhance agility and reduce human error, but as enterprise environments grow and diversify across multi-cloud fabrics, the traditional NSoT struggles to keep pace. Initially, network data was manually recorded in spreadsheets, leading to the development of centralized databases that could manage IP addresses and assets, but these systems have since evolved to include staging workflows and validation extensions to support modern demands. However, the limitations of SQL-based databases have become apparent in the face of high-velocity Infrastructure-as-Code workflows, prompting a shift towards programmable platforms like Infrahub, which offer flexible data models and graph-native architectures to accommodate complex infrastructure relationships. This shift is exemplified by Eurofiber, which reduced service deployment times significantly by adopting Infrahub, highlighting the need for a data model that represents the desired state rather than merely inventorying existing assets. As enterprises seek fully automated infrastructures, the distinction between traditional applications and native data platforms becomes crucial, with the latter providing a more adaptable and dynamic foundation that aligns with business needs and supports AI integration, ultimately setting a new standard for what a source of truth should be.

Infrahub's new Graph Traversal capability enhances network management by conducting automatic reachability checks, which ensure that changes within a network are assessed in real-time before being implemented. Unlike traditional methods that rely on hard-coded scripts and can miss critical updates, Infrahub offers a user-friendly interface allowing operators, rather than just programmers, to define network rules and enforce them automatically. This system uses a live network graph to immediately reflect any proposed changes, enabling engineers to identify potential violations before they affect the production environment. It also incorporates role-based access control, ensuring that only authorized personnel can modify reachability rules while providing visibility to all engineers about potential impacts on network paths, dependencies, security compliance, capacity management, and path redundancy. Infrahub's approach facilitates a more dynamic and secure network management process, significantly reducing the risk of network outages and compliance issues.

Infrahub 1.10 introduces Graph Traversal, a new capability that enhances infrastructure management by allowing both engineers and AI agents to navigate and analyze the complex relationships between infrastructure components. This feature is built on Infrahub's existing graph-based data storage system, which treats relationships as first-class entities. Graph Traversal offers two modes: Path Traversal, which identifies connections between objects, and Dependency Mode, which determines what depends on a particular object, aiding in impact analysis and troubleshooting. These modes are accessible through the UI, GraphQL API, Python SDK, and AI assistants via the MCP server, providing a consistent experience across the platform. The integration of Graph Traversal supports AI-driven operations by enabling agents to reason about infrastructure relationships and propose network changes based on governed data, facilitating quicker impact analysis and more efficient network management.

Infrahub 1.10 introduces Graph Traversal, a capability that lets infrastructure teams explore the relationships stored in its knowledge graph through an interactive UI, GraphQL API, Python SDK, and AI assistants connected through its MCP server. Path Traversal identifies all connections between two selected objects, helping users investigate connectivity, validate designs, and troubleshoot issues, while Dependency Mode identifies objects affected by a selected component up to a configurable depth to support impact and blast-radius analysis. The feature supports AI-driven operations by allowing agents to access governed, relationship-aware infrastructure data, generate impact reports, and propose changes through existing review and approval workflows. Traversal results respect both branch context and user permissions, enabling engineers, automation systems, and agents to work from an accurate and controlled model of infrastructure dependencies.