This topic describes the online transactional processing (OLTP) performance test results for PolarDB for MySQL 8.0.1 Cluster Edition.
For more information, see OLTP performance testing.
Dedicated specifications
The performance test data in this topic was collected from a PolarDB cluster that consists of a single read/write node and a single read-only node. For large-scale testing, we adjusted endpoints and optimized key parameters to minimize link latency and fully utilize the resources of the PolarDB cluster.
Parameter adjustments
Parameter adjustments for large-scale specifications
For the 120-core and 920 GB specification, we adjusted the following parameters:
DBNodeClassCPU specifies the number of CPU cores in the current compute node.
Parameter | Before modification | After modification | Optimization description |
| 1 | 120 | Adjusts the number of lock system partitions to match the number of CPU cores. This reduces lock contention in high-concurrency scenarios and improves concurrency performance. |
| {DBNodeClassCPU × 2} | {DBNodeClassCPU × 1} | Dynamically adjusts the thread pool size based on the number of CPU cores to optimize resource allocation and prevent excessive thread competition. |
| OFF | ON | Enables the lock-free Commit Sequence Number (CSN) mechanism to improve transaction commit efficiency and reduce lock overhead. |
Parameter optimization for other specifications
For other specifications, we configured the following parameters to ensure optimal cluster performance:
Parameter | Value | Description |
|
| Adjusts the number of lock system partitions to match the number of CPU cores. This reduces lock contention in high-concurrency scenarios and improves concurrency performance. |
Performance test results
This performance test uses the sysbench tool to stress the cluster by incrementally increasing the number of concurrent threads. The test starts with a low number of concurrent threads, such as 1, and then increases the number in increments, such as 8, 16, 32, 64, and 128. At each concurrency level, the test runs until key performance indicators, such as queries per second (QPS) and transactions per second (TPS), stabilize. To ensure data reliability, the test continues to run for a period after the performance curve stabilizes. The average value from this stable period is recorded as the performance data for that concurrency level. The test stops when the average QPS and TPS values no longer increase as more concurrent threads are added. The final performance result is the highest peak QPS and TPS values recorded across all concurrency test rounds.
Peak read-only performance
The following figure and table show the performance test results for each specification.
In the read-only scenario, the parameter --range-selects=0 is used. This configuration disables range queries and effectively tests the oltp_point_selects scenario, which focuses on point-select queries.
Specifications | 2 cores, 8 GB | 2 cores, 16 GB | 4 cores, 16 GB | 4 cores, 32 GB | 8 cores, 32 GB | 8 cores, 64 GB | 16 cores, 32 GB | 16 cores, 64 GB | 16 cores, 128 GB | 32 cores, 128 GB | 32 cores, 256 GB | 64 cores, 512 GB | 120 cores, 920 GB |
Read-only QPS | 123234.31 | 128407.45 | 261992.95 | 263557.53 | 518849.56 | 514733 | 919903.31 | 968769 | 922508.84 | 1435099.19 | 1431018.25 | 1975797.41 | 3805887 |
Read-only TPS | 12323.43 | 12840.74 | 26199.3 | 26355.75 | 51884.96 | 51473.3 | 91990.33 | 96876.9 | 92250.88 | 143509.91 | 143101.84 | 197579.74 | 380588.7 |
Peak read-write performance
The following figure and table show the performance test results for each specification.
Specifications | 2 cores, 8 GB | 2 cores, 16 GB | 4 cores, 16 GB | 4 cores, 32 GB | 8 cores, 32 GB | 8 cores, 64 GB | 16 cores, 32 GB | 16 cores, 64 GB | 16 cores, 128 GB | 32 cores, 128 GB | 32 cores, 256 GB | 64 cores, 512 GB | 120 cores, 920 GB |
Read-write QPS | 34659.68 | 38909.52 | 90971.5 | 95615.19 | 197093.04 | 198704.02 | 387869.61 | 390841.8 | 380663.09 | 687255 | 663323.67 | 984014.28 | 2195884 |
Read-write TPS | 1732.98 | 1945.48 | 4548.57 | 4780.76 | 9854.65 | 9935.2 | 19393.48 | 19542.09 | 19033.15 | 34362.74 | 33166.17 | 49200.72 | 109794.20 |
Peak write performance
The following figure and table show the performance test results for each specification.
Specifications | 2 cores, 8 GB | 2 cores, 16 GB | 4 cores, 16 GB | 4 cores, 32 GB | 8 cores, 32 GB | 8 cores, 64 GB | 16 cores, 32 GB | 16 cores, 64 GB | 16 cores, 128 GB | 32 cores, 128 GB | 32 cores, 256 GB | 64 cores, 512 GB | 120 cores, 920 GB |
Write QPS | 23914.56 | 25353.92 | 65187.37 | 63711.64 | 135037.31 | 127477.78 | 231777.32 | 237859.68 | 232331.81 | 381489.46 | 377614.51 | 608817.22 | 879742.76 |
Write TPS | 3985.76 | 4225.65 | 10864.56 | 10618.61 | 22506.22 | 21246.3 | 38629.56 | 39643.28 | 38721.97 | 63581.58 | 62935.76 | 101469.57 | 146623.8 |