Architecting Scalable Intent-Based Networking Platforms for Enterprise Automation

Abstract

Enterprise intent-based networking requires more than translating high-level objectives into device configurations. A platform must coordinate planning, execution, state management and assurance across heterogeneous domains while preserving policy constraints and traceability. This paper proposes a conceptual architecture that separates intent admission, semantic normalization, hierarchical planning, distributed execution, state management and continuous assurance, and governance. An authoritative, versioned intent record defines the desired outcome; bounded domain contracts delegate planning and execution to controllers with local knowledge. Stable ownership, asynchronous work queues, idempotent operations, incremental recomputation and local reconciliation are proposed to reduce unnecessary global coordination. The architecture distinguishes globally significant invariants, which require appropriately coordinated decisions, from telemetry and intermediate operational observations that may tolerate explicitly bounded staleness. The contribution is a structured synthesis of intent-based networking, model-driven management and distributed-systems principles into a testable enterprise control-plane design. A proposed evaluation framework compares centralized planning and execution, centralized planning with distributed execution, and hierarchical planning with local reconciliation under controlled workloads. It identifies admission latency, planning cost, execution throughput, convergence, telemetry overhead, conflict detection, invariant violations and recovery behavior as evaluation measures. The paper presents architectural propositions and an experimental protocol; it does not report an implemented prototype, measured performance gains or completed experimental validation.