Benchmarking SurrealDB 3.x vs. Postgres, Mongo, Neo4j and Redis (With Fsync)

3 days ago

Benchmarks were conducted using identical hardware and an open-source harness to ensure fairness, with all databases configured for production-grade durability.
SurrealDB 3.x shows significant performance improvements over previous versions, including a 31% faster CRUD throughput and massive gains in full-table scan performance.
In comparisons, SurrealDB outperforms PostgreSQL and MySQL in write operations, competes closely with MongoDB in document workloads, and surpasses Neo4j in graph database metrics.
For key-value stores, SurrealDB's in-memory engine significantly outperforms Redis and KeyDB on write operations while offering additional features like transactions and a full query language.
In embedded mode, SurrealDB greatly exceeds SQLite's performance on most operations, including being up to 110× faster on updates.
Future development focuses on closing performance gaps in batch operations, query planning, and storage layer optimizations, aiming for competitiveness across all workloads with a single multi-model engine.
The multi-model capability of SurrealDB is particularly relevant for AI agents, which require diverse data types in a single, transactional store that can run locally or distributed.

Hasty Briefsbeta