扩展2：数据库性能测试m6g vs. m5
We uses HammerDB to benchmark and compare between different database engines. We have got their results of RDS MySQL m6g and m5 instances. Here is a summary of results and findings from them.
HammerDB Benchmark Results of RDS MySQL m6g vs. m5
DBS Benchmarking team uses HammerDB to benchmark and compare between different database engines. We have got their results of RDS MySQL m6g and m5 instances. Here is a summary of results and findings from them.
Overall Throughput
According to the benchmark results from DBS Benchmarking team:
- For peak throughput, m6g is up to 20% better than m5 (excluding 24xl).
- Note: this also include certain level of performance improvements due to OS/kernel version upgrade, from 4.4 to 4.14.
- The advantage of m6g vs. m5 at 16xlarge is also partially due to single CPU socket vs. 2x CPU sockets.
- As the biggest single CPU socket instance type of m5, at 12xlarge, m6g is about 12% better than m5.
- From large to 12xlarge, the advantage of m6g at peak throughput is around 12% better than according m5 instance type.
Throughput by Concurrency
-
For **large, xlarge and 4xlarge instance sizes, m6g shows consistent higher throughput compare to m5, in some cases up to about 50%. **
-
The benchmark results of 12xlarge and above are not as consistent as smaller instance sizes. This should also be looked into.
large
xlarge
4xlarge
12xlarge
16xlarge and above
Appendix
Benchmark configurations
- HammerDB version: 3.1
- Driver instance type: “m5.4xlarge”
- Number of warehouses
| Instance Size | Num of Wareshoues |
|---|---|
| large | 8 |
| xlarge | 16 |
| 2xlarge | Not done yet |
| 4xlarge | 50 |
| 8xlarge | Not done yet |
| 12xlarge | 150 |
| 16xlarge | 150 |
| 24xlarge | 300 |
- Custom DB parameters
| Engine | Version | DB Param | Value |
|---|---|---|---|
| MySQL | 8.0.17 | innodb_buffer_pool_instances | 32 |
| MySQL | 8.0.17 | innodb_io_capacity | 40000 |
| MySQL | 8.0.17 | innodb_io_capacity_max | 70000 |
| MySQL | 8.0.17 | innodb_log_file_size | 21474836480 |
| MySQL | 8.0.17 | innodb_read_io_threads | 24 |
| MySQL | 8.0.17 | innodb_sync_array_size | 1024 |
| MySQL | 8.0.17 | innodb_write_io_threads | 64 |
- DB instances configurations
| Engine | Instance Type | OS Version | Kernel Version | Storage |
|---|---|---|---|---|
| MySQL | m6g.large | Amazon Linux 2 | 4.14.177-139.254 | io1 40k 1TB |
| MySQL | m6g.xlarge | Amazon Linux 2 | 4.14.177-139.254 | io1 40k 1TB |
| MySQL | m6g.4xlarge | Amazon Linux 2 | 4.14.177-139.254 | io1 40k 1TB |
| MySQL | m6g.12xlarge | Amazon Linux 2 | 4.14.177-139.254 | io1 40k 1TB |
| MySQL | m6g.16xlarge | Amazon Linux 2 | 4.14.177-139.254 | io1 40k 1TB |
| MySQL | m5.large | Amazon Linux 2016.09 | 4.4.23-31.54 | io1 40k 1TB |
| MySQL | m5.xlarge | Amazon Linux 2016.09 | 4.4.23-31.54 | io1 40k 1TB |
| MySQL | m5.4xlarge | Amazon Linux 2016.09 | 4.4.23-31.54 | io1 40k 1TB |
| MySQL | m5.12xlarge | Amazon Linux 2016.09 | 4.4.23-31.54 | io1 40k 1TB |
| MySQL | m5.16xlarge | Amazon Linux 2016.09 | 4.4.23-31.54 | io1 40k 1TB |
| MySQL | m5.24xlarge | Amazon Linux 2016.09 | 4.4.23-31.54 | io1 40k 1TB |