All Products
Document Center

Performance tests on backup restoration to MySQL

Last Updated: May 30, 2020

Performance tests on backup restoration

The following performance test results are for reference only and cannot be used to evaluate SLA performance.

Logical backup

Test environment

Test tool: SysBench

Test item Test value
Database size 4 GB
Number of tables 16
Total number of records 20,000,000
Single record size 0.2 KB

Restoration tests

Restoration performance

Instance type (number of restored threads) 4core 16G 9000IOPS 8core 32G 18000IOPS
xlarge(16) 97151 (19.4 MB) 120861 (24 MB/s)
large(8) 75763 (15.1 MB) 81873 (16.4 MB/s)
medium(6) 63851 (12.8 MB) 65924 (13.2 MB/s)
small(4) 44333 (8.8 MB) 46357(9.3 MB/s)
micro(2) 34531 (6.9 MB) 35052(7.1 MB/s)

Restoration optimization

The challenge in logical restoration lies within the CPU, IOPS, and memory of database instances. If you want to restore backup data to ApsaraDB for RDS, we recommend that you select the High-availability Edition. You can also restore backup data to Apsara PolarDB to improve the restoration speed.

You can optimize configurations of the following parameters to improve database restoration performance:

  • max_allowed_packet
  • innodb_buffer_pool_size
  • innodb_doublewrite
  • innodb_flush_log_at_trx_commit
  • innodb_flush_method
  • innodb_io_capacity
  • innodb_io_capacity_max