メインコンテンツまでスキップ

Sysbench ベンチマーク結果

This section presents database cluster performance benchmark results performed using the Sysbench tool, aiming to provide a reference for the processing capabilities of database engines across different compute configurations (flavors).

The results are provided for reference purposes only and do not represent guaranteed performance.

1. Benchmark sysbench overview

Purpose:

  • Evaluate OLTP (Read/Write) performance of supported database engines.
  • Observe performance scaling behavior when increasing CPU and memory resources.
  • Provide reference data to assist customers in selecting appropriate instance sizes.

Metrics:

  • Sysbench Read : Total number of read queries executed during the test.
  • Sysbench Write : Total number of write queries executed during the test.
  • QPS (Queries Per Second) : Average number of queries processed per second.
  • TPS (Transactions Per Second) : Average number of completed transactions per second.

Higher QPS and TPS values indicate higher throughput under the tested workload.

2. Benchmark results

Test environment:

ParameterValue
Benchmark toolSysbench (OLTP Read/Write)
Number of documents1.000.000
Number of operations1.000.000
Workload typeRead/Write
Thread count (YCSB threads)Cấu hình tương ứng với từng flavor
MongoDB versionMongoDB Enterprise 8.0.19

Result:

Flavor (vCPU/RAM)Thread countSysbench ReadSysbench WriteQPSTPS
2C4G85001954998053372.363372.36
2C8G85000224999784004.934004.93
4C8G164997725002285023.815023.81
8C16G325002934997076417.546417.54
8C32G325003724996285921.645921.64
16C32G644998655001358532.798532.79
16C64G6449998950001112972.0212972.02

3. Analysis & Recommendations

  • Increasing CPU and memory generally improves throughput.
  • Performance gains may diminish at higher configurations depending on workload and system limits.

Important:

  • Benchmark results are workload-specific and provided for reference only.
  • Actual performance may vary depending on:
    • Application workload characteristics.
    • Database schema and indexing.
    • Read/write ratio.
    • Storage and network configuration.

Customers are strongly encouraged to test with their own workloads before deploying to production.

Recommendations:

Use these benchmark results as guidance when selecting database engines and instance sizes. For optimal performance, validate configuration choices through application-specific performance testing.