SAREF4SYST Extension Reports 90-100% Deployment Success for Edge-Fog-Cloud AI Workflows

The ontology adds classes for executable AI jobs, deployment constraints, and communication relationships to SAREF4SYST. It models both workflow structure and heterogeneous infrastructure layers in a single ETSI-aligned schema. Competency questions covering workflow deployment, execution reasoning, and workload adaptation were all answered via SPARQL and semantic rules.

Proof-of-concept tests across edge, fog, and cloud nodes produced 90-100% success rates with average decision times below 80 ms. These metrics were obtained from repeated orchestration runs on smart grid energy service scenarios. The results confirm that unified semantic descriptions remove format mismatches that previously blocked automated placement.

The model supports resource-aware reasoning required for cloud-fog-vertex coordination by exposing compute capabilities and latency constraints as first-class entities. This directly addresses gaps in prior SAREF modules that lacked explicit AI execution semantics. Operational deployments can now query the ontology to select nodes that satisfy both pipeline requirements and infrastructure limits without custom translators.

Next steps include integration with existing ETSI oneM2M and SAREF device ontologies to enable live adaptation when vertex-layer nodes join or leave the continuum.