Architecture AI Updates: November 14, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. Netflix Describes an Ontology-Backed Knowledge Graph That Gives Its Root-Cause Agents a Shared Model of Production
InfoQ reported on November 13 on a Netflix engineering write-up about an ontology-driven knowledge graph that links applications, incidents, deployments, alerts and infrastructure dependencies. The motivation given is that a typical application regression investigation takes about four hours, much of it spent reconstructing how systems relate, because metrics, events, logs and traces sit in separate alerting, incident, deployment and chat tools. An Observability Ontology defined with RDF, OWL, SHACL and SPARQL sets the entity types and relationships, and the triples are stored in QuipuDB, a distributed RDF database built on the Rust redb storage engine. Pipeline components called enrichers turn operational data into subject-predicate-object triples, some deterministically and some with LLMs, and the outputs are checked against validation rules for entity types, URI naming and the software catalogue. A correlation engine examines 13 signal types, including log errors, metric anomalies, deployments, configuration changes and experiments, and scores possible causal links by factors such as temporal proximity and severity to produce ranked hypotheses. The graph underpins AutoRCA, Netflix’s internal agentic root-cause analysis system, where specialist agents for traces, logs or deployments write to the same entities through the ontology instead of exchanging loosely structured findings. Automated remediation and integration with Netflix’s Unified Data Architecture are described as goals, with humans approving proposed changes. Source · Netflix post