DATABASEPerformance,
with the context attached.
Operations and latency come from a versioned FZ evidence artifact. Select a database family, then choose a metric to compare every recorded database column across every workload row.
Only matched, qualified lanes show a value. Pending lanes never receive a placeholder score.
Compare one workload across its databases.
The report is fetched from the published static evidence artifact. A new qualified FZ lane changes the artifact and this table without copying performance values into the frontend.
Every model needs its own measurement.
Results enter this page when matching workloads and independent checks pass. Pending databases have no published score.
| Database | Workload | Latest qualification |
|---|---|---|
| Embedded | Qirava · SQLite | Qualified warm reads |
| Key-value | Qirava · Redis | Pending |
| SQL | Qirava · PostgreSQL | Pending |
| Document / NoSQL | Qirava · ArangoDB · MongoDB | Pending |
| Graph | Qirava · Neo4j · ArangoDB | Pending |
| Search | Qirava · OpenSearch | Pending |
| Vector / AI | Qirava · Qdrant | Pending |
Read the conditions with the number.
The current embedded lane uses 2,000 warm JSON records, matching unique indexes, five-second trials and one or eight pinned workers. Each timed request contains 16 logical queries and verifies returned keys, values and ordering. The report keeps physical and logical Qirava residency separate, with encryption disabled and enabled; SQLite is plaintext.
Durable writes, cold data, beyond-memory datasets, concurrent mixed traffic, replication and production saturation require separate qualification. A score in one family does not establish a universal ranking.