SurrealDB 3.x Benchmarks Detail Production-Grade Durability and Performance Claims

This review analyzes SurrealDB's 3.x performance claims, focusing on its detailed, open-source benchmarking methodology and the explicit shift to full disk durability settings.

The Answer Up Front

SurrealDB 3.x presents itself as a compelling option for developers seeking a multi-model database with strong performance guarantees under production-grade durability settings. Its transparent benchmarking approach, emphasizing fsync on and optimized configurations, makes its performance claims more credible than many vendor-published numbers. Builders prioritizing a single database for diverse workloads (graph, document, key-value, relational) and those who value verifiable, durable writes should consider it. Skip it if your application demands extreme low-latency reads where in-memory caching is acceptable without immediate disk persistence, or if you are already deeply invested in a specialized database ecosystem.

Methodology

This v0 review draws on the founder's published claims at dev.to/surrealdb/surrealdb-3x-by-the-numbers-39ao and the linked public artifacts: surrealdb.com/benchmarks (for full results and methodology) and github.com/surrealdb/crud-bench (for the open-source harness and configuration files). Independent benchmarks are pending. Update cadence: re-tested when claims diverge from observed behavior.

SurrealDB 3.x, as observed on May 29, 2026, was benchmarked by its founder, Tobie Morgan Hitchcock, against other database engines. The testing environment was standardized: an AMD Ryzen Threadripper 9970X (32C/64T) machine with 128 GiB DDR5 RAM, NVMe storage, and Ubuntu 24.04. All databases ran on this identical hardware.

The crud-bench open-source harness was used, with each workload translated into the respective engine's native query language to ensure fair comparison. Crucially, all engines were configured for production-grade durability, meaning fsync was enabled, WAL flushed on every commit, and no buffered writes were hidden behind the page cache. The configuration files for each engine are available in the crud-bench repository for audit. Furthermore, each database was run with optimized configurations, not out-of-the-box defaults. This involved tuning connection and worker pool limits, sizing buffer pools and caches, enabling parallel query execution, and setting WAL/checkpoint intervals according to each project's performance guides. Workloads involved 128 clients issuing 48 concurrent queries against datasets of 5-15 million rows, comprising mixed-type records.

What's covered in this review are the founder's claims regarding the benchmarking methodology, the explicit commitment to full durability, and the architectural changes in SurrealDB 3.x. What's not covered are independent performance verification, long-term workflow integration, or edge-case behavior under specific failure conditions.

What It Does

SurrealDB is a multi-model database designed to handle diverse data types and access patterns within a single engine. The 3.x release focuses on significant internal architectural overhauls to improve performance and stability, particularly under durable write conditions.

Rebuilt Query and Parser Layers

Across three major releases, SurrealDB has undergone a fundamental redesign of its query and parser layers. This aims to enhance efficiency in processing complex queries and improve overall query execution speed. While specific performance numbers are on the surrealdb.com/benchmarks site, the blog post emphasizes the internal improvements that underpin these gains.

Enhanced Storage Engine

The storage layer has also been fundamentally rebuilt. This is critical for a multi-model database, as it must efficiently manage and retrieve various data structures (documents, graphs, key-value pairs, relational tables). The rebuild is intended to optimize disk I/O and data persistence, especially when fsync is enabled for full durability.

Full Durability by Default

A key aspect of the 3.x release, and the focus of the new benchmarks, is the explicit commitment to full disk durability. Unlike previous benchmarks where fsync was disabled, SurrealDB 3.x is tested and configured with fsync on, ensuring that every committed transaction is written to disk before an acknowledgment is sent to the client. This aligns with standard production requirements for data integrity and resilience against power outages.

What's Interesting / What's Not

What's genuinely interesting about SurrealDB 3.x is the founder's explicit commitment to a transparent and rigorous benchmarking methodology. Database benchmarks are often opaque, making performance claims difficult to verify. Tobie Morgan Hitchcock's decision to run all databases on identical hardware, use an open-source harness (crud-bench), and, most importantly, enable full disk durability (fsync on) for all comparisons, sets a high bar. The public availability of the crud-bench repository, including the optimized configuration files for each database, allows for auditing and potential reproduction, which is rare and commendable.

The focus on optimized configurations, rather than out-of-the-box defaults, is also a pragmatic and valuable approach. Production deployments rarely use default settings; tuning is essential. By optimizing all competing engines, the benchmarks aim to reflect real-world performance potential, not just initial setup ease. The acknowledgment of previous benchmarks running with fsync disabled and the subsequent correction demonstrates a commitment to accuracy over inflated numbers.

What's not as interesting, or rather, what remains to be seen, is the independent verification of these performance claims. While the methodology is robust, the numbers themselves are still founder-reported. The blog post itself doesn't present specific performance metrics, instead directing readers to the benchmarks site. This requires an additional step for the reader to evaluate the actual performance gains. Furthermore, while the multi-model approach is a strong selling point, the specific trade-offs and performance characteristics for each model (e.g., graph traversal speed versus document query latency) are not detailed in the signal, requiring deeper analysis of the benchmark results.

Pricing

The provided source signal does not include pricing information for SurrealDB 3.x. Pricing details would typically be found on the official SurrealDB website or through direct inquiry.

Verdict

SurrealDB 3.x is a strong contender for developers building applications that require a flexible, multi-model database with a non-negotiable need for data durability. The founder's transparent and methodologically sound benchmarking, particularly the emphasis on fsync enabled and optimized configurations across all tested engines, provides a credible foundation for its performance claims. While the specific performance numbers require review of the linked surrealdb.com/benchmarks site, the approach to benchmarking itself is a significant differentiator. We recommend SurrealDB 3.x for teams looking to consolidate their data storage needs into a single, performant, and durable solution, provided they are willing to engage with a newer database ecosystem.

What We'd Test Next

For a v2 review, we would independently reproduce the benchmarks using the crud-bench harness and the provided configuration files to verify the founder's performance claims. We would focus on specific workload types relevant to common production scenarios, such as high-volume mixed read/write operations, complex graph traversals, and geospatial queries. Additionally, we would investigate SurrealDB's operational overhead, including backup/restore processes, scaling behavior under sustained load, and resource consumption in a containerized environment. Long-term stability and developer experience, including client library maturity for various languages, would also be key areas of exploration.

The investor read

The database market continues its trend towards converged, multi-model solutions, and SurrealDB's 3.x release, with its focus on transparent, durable benchmarking, signals a maturation in this segment. The explicit fsync on methodology addresses a critical concern for enterprise adoption, where data integrity often trumps raw, non-durable speed. This positions SurrealDB against established players like MongoDB (document), Neo4j (graph), and even Postgres (with extensions), by offering a unified API and storage layer. An investor would look for sustained adoption, a growing ecosystem of tools and integrations, and independent validation of these performance claims. The open-source crud-bench harness is a strong positive, fostering community trust and auditability, which could accelerate adoption among developers wary of vendor lock-in or opaque benchmarks. This could be an investable play if it captures significant market share from specialized databases by offering a 'good enough' multi-model solution with strong durability guarantees.