Overview
Performance is a top concern for architects and engineering leaders evaluating any database for production use.
SurrealDB is designed with performance in mind - supporting real-time querying, fast indexing, and scalable throughput across diverse workloads including graph traversal, vector search, document storage, and time-series ingestion.
This article highlights SurrealDB’s core performance principles, benchmark data, and best practices for tuning your deployment for speed and scale.
Core performance strengths
Fast graph traversals
Deep relationship queries that would require joins in SQL are optimised via SurrealDB’s native graph engine.
SurrealQL performance
The SurrealQL query planner and execution engine are optimised for a hybrid of relational and document-style access patterns.
Indexed search
Full-text, numeric, and geospatial indexes can dramatically speed up query performance.
Efficient vector search
Native support for similarity search with optimised vector indexing.
Streaming writes
High-throughput ingestion pipelines are supported via batch insert operations and client-driven streaming.
Scalability model
Embedded → distributed
Run SurrealDB embedded (single binary), then scale to distributed clusters without rewriting your application.
Query concurrency
Supports high concurrency with built-in ACID transaction isolation and lockless reads.
Write performance
Internally powered by RocksDB, tuned for write-heavy use cases such as IoT or logging.
Benchmarking
Interested in benchmarking SurrealDB or understanding how we test performance? Here are some resources to get you started:
Read our blog on the start of our benchmarking journey - and why benchmarking multi-model databases presents unique challenges.
Explore our public GitHub benchmarking repo - with all benchmark code, detailed instructions, and ways to contribute
Performance best practices
Model your data and schema carefully
Use SurrealDB’s flexible modeling to simplify complex joins and leverage graph relationships efficiently. Design your schema to match your most common query patterns.Use indexes strategically
Apply full-text, numeric, and geospatial indexes where needed to improve query performance on large datasets. Avoid over-indexing frequently updated fields.Tune queries for efficiency
UseLIMITto avoid unbounded result sets. Write SurrealQL queries that align with your data model - whether document-based, graph-based, or vector-based.Monitor performance actively
Use SurrealDB’s metrics endpoints and query logs to track slow queries, identify hotspots, and optimise query performance.Benchmark and iterate
Test your workload under realistic conditions before deploying to production. Profile key query paths and adjust schema, indexing, or query structure as needed.