Performance and Scaling¶

Guidance for sizing OntoLogos workloads. OntoLogos v0.4 uses batch fixed-point engines (not incremental). Benchmark numbers are indicative — run cargo bench on your hardware for production decisions.

Default limits¶

JSON deserialization¶

Limit	Default	Configure via
`max_json_bytes`	16 MiB	`Limits::max_json_bytes`
`max_entities`	1,000,000	`Limits::max_entities`
`max_axioms`	10,000,000	`Limits::max_axioms`
`max_iri_len`	8,192	`Limits::max_iri_len`

See Security and JSON snapshot v2.

OWL file parsing¶

Limit	Default	Configure via
`max_file_bytes`	64 MiB	`ParseLimits::max_file_bytes`
`max_axioms`	10,000,000	`ParseLimits::max_axioms`

See ParseLimits.

Engine behavior¶

RDFS (`RdfsEngine`)¶

Complexity: Depends on taxonomy depth and property hierarchy size; runs until TBox rules saturate.
Memory: In-place — inferred axioms are added to the same Ontology.
Parallelism: Sequential only in v0.4.

OWL RL (`RlEngine`)¶

Parallelism: RlEngine::new(n) with n in 1..=64. Parallelism affects ABox type-rule candidate expansion only; use 1 for fully sequential execution.
Pipeline: Always runs RDFS materialization first, then RL rules to fixed point.

let report = RlEngine::try_new(4)?.saturate(&mut ontology)?;

For reproducible debugging, use RlEngine::new(1).

Reference corpora (v0.4)¶

Corpus	Mapped axioms (approx.)	Profile	Notes
Family	~57	RL	Vendored; good RL smoke test
Pizza	~658	DL	Requires `download.sh`; stress parser + profile

Run locally:

./benchmarks/scripts/download.sh
cargo test -p ontologos-parser --test manifest_integration
cargo test -p ontologos-rl --test corpus

Optional Criterion bench:

cargo bench -p ontologos-core

Results under target/criterion/.

External comparison (RL)¶

For OWL RL materialization baselines, compare against reasonable using the optional harness:

./benchmarks/scripts/compare-reasonable.sh benchmarks/data/family.owl

See Conformance coverage.

Scaling recommendations¶

Workload	Recommendation
Small ontologies (< 10k mapped axioms)	Default limits; `RlEngine::new(1)`
Medium batch jobs	Tune `ParseLimits`; snapshot to JSON v2 after saturation
Untrusted uploads	`load_ontology_in(base, path)` + reduced limits — Production integration
Large DL corpora (GALEN, SNOMED)	Not production-ready in v0.4; optional `#[ignore]` stress tests only
Incremental updates	Not supported until v0.7 — reload and re-saturate