New in the benchANT Database Ranking: YugabyteDB for CRUD Workloads
We have expanded the benchANT Database Performance Ranking by adding distributed database YugabyteDB for the first time. As a distributed SQL database, YugabyteDB offers both PostgreSQL compatibility and Cassandra Query Language (CQL) compatibility, making it a versatile choice for different application scenarios.
For this release, we benchmarked YugabyteDB via its PostgreSQL-compatible interface, focusing on our CRUD general purpose ranking workload. To provide meaningful insights into distributed deployments, we tested three out of five scaling sizes, medium, large, and xlarge, which demonstrate realistic distributed database deployments.
This addition enables a direct comparison of YugabyteDB with other distributed relational and NoSQL databases in our ranking, highlighting how a distributed SQL approach performs in terms of throughput, latency, scalability and price-performance.
Key Insights YugabyteDB Performance Results Across Cluster Sizes
Pure performance: The xlarge YugabyteDB setup reaches over 100,000 ops/s which is only achieved by three other (NoSQL) databases on AWS in our ranking
Scaling impact: YugabyteDB shows strong scalability for the CRUD general purpose workload, with the large setup reaching 106,000 ops/s, significantly higher than the medium setups.
Price-performance trade-off: with growing cluster sizes and workloads, the price-performance ratio remains on a constant level
Latency profile: Latency remains consistent across setups, with P95 read latency between 5–7 ms and P95 write latency between 16–19 ms, indicating stable behavior under varying cluster sizes.
Key Insights How YugabyteDB Stacks Up Against Other Distributed Databases
Solid throughput: With 106,000 ops/s, YugabyteDB as SQL database delivers strong absolute performance, but remains behind the NoSQL databases ScyllaDB, Couchbase, and Cassandra in raw throughput at this scale.
Latency advantage: YugabyteDB shows lower P95 read latency (6 ms) than Cassandra and Couchbase, making it attractive for read-intensive workloads, though write latency (16 ms) is higher than its peers.
Positioning: While not the leader in throughput, YugabyteDB offers a consistent cost/performance, even at a high scale, with data resiliency. Along with YugabyteDB's PostgreSQL compatibility, makes it an viable alternative for organizations prioritizing PostgreSQL ecosystem fit over raw performance.
Benchmarking: Scenarios & Methodology
For the CRUD general purpose workload of our ranking, we applied the following three scaling sizes for the benchmarks: medium, large, xlarge. A full overview on all scaling sizes is associated with the ranking.
Workload
- Benchmark: benchANT YCSB with the Yugabyte JDBC driver v42.7.3-yb-3
- Initial data size: 2.5 GB
- Data set size: 500 bytes
- Access distribution: Zipfian
- Benchmark API VM: 16 cores
- Runtime per benchmark: 30 minutes
- Repetitions per benchmark: 1 execution
- Parallel threads: 100 - 600 (depending on scaling size)
Database Deployment
- Yugabyte version: OS 2025.1.0.0
- Configuration: vanilla
- Cloud provider: AWS
Scaling: Medium
- VM size: 4 vCPUs x 16 GiB RAM
- Cluster size: 3
- Replication Factor: 3
- Workload threads: 100
Scaling: Large
- VM size: 8 vCPUs x 32 GiB RAM
- Cluster size: 3
- Replication Factor: 3
- Workload threads: 200
Scaling: xLarge
- VM size: 8 vCPUs x 32 GiB RAM
- Cluster size: 9
- Replication Factor: 3
- Workload threads: 600
The full results are available on GitHub.
Conclusion & Outlook
YugabyteDB adds an important distributed SQL perspective to the benchANT Database Ranking, offering both PostgreSQL and CQL compatibility.
Performance scales predictably with larger clusters, reaching over 100,000 ops/s in the xlarge setup.
Latency remains stable and competitive, with strong read performance compared to other distributed databases.
Price-performance stays consistent, making YugabyteDB a balanced choice, even if it lags behind NoSQL leaders in raw throughput efficiency.
For organizations already invested in PostgreSQL, YugabyteDB offers a compelling mix of ecosystem familiarity and distributed scalability.
Beyond generic CRUD workloads, it would be interesting to benchmark towards HTAP, analytical, and vector workloads to capture a broader picture of YugabyteDB’s strengths and trade-offs.
About benchANT and the Database Ranking
benchANT is an independent benchmarking and consulting company focused on database and cloud performance. For vendors, we design and run unbiased benchmarks for database and cloud services, and we help enterprise customers make informed technology choices with application-specific performance evaluations.
Our benchmarking framework is automated, transparent, and reproducible, ensuring reliable insights that go beyond marketing claims.
With the openly accessible Database Ranking, we provide a standardized comparison platform for database performance results, which is continuously updated with new systems and workloads. We also offer the DBaaS Navigator, a free technical tool to explore and compare Database-as-a-Service offerings.
Disclaimer on the Database Ranking
The performance results presented in the Database Ranking are based on standardized, generic workloads (e.g., CRUD, time-series) executed under specific configurations and resource settings. These measurements are intended to provide high-level, comparable insights into database performance, cost, and scalability.
However, they should not be interpreted as definitive guidance for all use cases. Real-world applications often differ significantly in workload characteristics, data distribution, query patterns, and infrastructure requirements. As such, the ranking results should be seen as an orientation and starting point, rather than the sole foundation for technology decisions.
For organizations where performance is a critical factor, we strongly recommend the design and execution of application-specific benchmarks. benchANT provides tailored benchmarking services to ensure that database and cloud technology choices are validated against the actual needs and workloads of your application.