ReasoNet: Inferring Network Policies Using Ontologies

Abstract

Abstract-Modern SDN control stacks consist of multiple abstraction and virtualization layers to enable flexibility in the development of new control features. Rich data modeling frameworks are essential when sharing information across control layers. Unfortunately, existing NOS data modeling capabilities are limited to simple type-checking and code templating. We present an exploration of a more extreme point on SDN data modeling: ReasoNet. Developers can use semantic web technologies to enrich their data models with reasoning rules and integrity/consistency constraints, and automate state inference across layers. We demonstrate the ability of ReasoNet to automate state verification and cross-layer debugging, through the implementation of two popular control applications, a learning switch and a QoS policy engine.

An agent-based approach to disintegrate and modularise Software Defined Networks controller

“…There has also been some recent work on the knowledge representation front for SDN. Authors of [41], [42] propose detailed ontology for SDN. Zhou [42] recursively fetches network information to build a knowledge graph and employs the SPARQL query engine to make inferences.…”

“…Zhou [42] recursively fetches network information to build a knowledge graph and employs the SPARQL query engine to make inferences. [41] proposes services to update the knowledge graph and employs SPARQL to query knowledge. In the proposed work, we build a knowledge base to detect inconsistencies in ontology and network abnormalities such as link-down events and congestion events while leaving functions like pathfinding and resolving network abnormalities to agents.…”

Towards a Self-Driving Management System for the Automated Realization of Intents

“…The lack of formalism and heterogeneity in intent and policy processing in [48]- [50] incur sophisticated mapping functions and analysis to map the attributes among policy abstraction types. This additionally introduces the challenge of inconsistency and synchronization across layers of abstraction [51]. The languages proposed in [52]- [54] are specifically tailored to model QoS and access control policies.…”

“…There are two main requirements from our autonomic management system to be addressed by a policy model. First, the policy model should facilitate mapping from one layer to another leading to strong synchronization and consistencies among various policies, especially at run-time [98]. Hence, we require a unified policy model that models policies across various layers of abstraction within our management system.…”

Software-Defined Multi-domain Tactical Networks: Foundations and Future Directions

“…Semantic ontology incorporates various reasoning rules and integrity constraints that helps in automating state inference across the SDN layers. Additionally, it simplifies the remote configurations of data plane nodes and allow controllers to define complex data relationships [62]. An illustration of semantic ontology-based operations in SDN domain is depicted in Fig.…”