Hi BoneBunny,
ext4 and xfs are both suggested for MySQL/Percona Server and shouldn’t be the issue
Possible reasons for performance differences:
- Different configuration. Do check you are using same settings on both MySQL servers. You can also check how to flush innodb write activity. You can check how to tune MySQL with the following Tuning MySQL/InnoDB Flushing for a Write-Intensive Workload - Percona Database Performance Blog
- Difference in MySQL major version. Default value do change between major versions and that can impact performance, but there are also other architectural changes. Are you comparing against same version?
- Disk scheduler: using cfq has worse performance compared to noop and deadline: Linux schedulers in tpcc like benchmark - Percona Database Performance Blog . Newer kernels have other disk schedulers. Do check which one you are using as can be a non suggested
- Are you comparing throughput against similar disks specs? I.e if you are comparing performance between spinning and SSD, comparison won’t make any difference. Do check iostat and check that both disks have similar write/read activity
- How are you comparing inserts? You should test same load against both servers, using either sysbench or tools like pt-upgrade — Percona Toolkit Documentation . Depending on table size and row size, number of transactions might differ , but overall performance might be similar
- Are mount options the same on both systems?
Please confirm if after having everything the same, performance differences do still exist