MariaDB-5.3.4 performance testing MariaDB
BitsCN.com
Last weekend, Vadim from Percona released a MariaDB 5.3.4 benchmark results. as a new member of Monty Program, I would like to add some of my own test results.
In the Vadim test results, a comment said: Take MariaDB 5.3.4 and MySQL 5.5 to compare whether it is fair, it should be compared with MySQL 5.1, because MariaDB 5.3.4 is based on MySQL 5.1. I would like to say this is not a big problem, because the OLTP performance test results of MySQL 5.5 and MySQL 5.1 in Sysbench are not much different.
Therefore, I created a Sysbench test environment similar to Vadim, including the following test versions:
MariaDB-5.3.4-latest version of Monty Program, including XtraDB and InnoDB plug-ins
Percona-Server 5.1.61 is also based on the same XtraDB version as MariaDB 5.3.
Percona-Server 5.5.20-latest Percona Server version
MySQL-5.5.20-latest MySQL version
Is the test result:
We can see that the performance of the MariaDB-5.3.4 is much worse than MySQL 5.5.20, and the difference between Percona Server 5.1 and 5.5 is not big, but if MariaDB 5.3.4 uses the InnoDB storage engine instead of the default XtraDB engine, mariaDB has more performance than Percona Server 5.5.
Therefore, we will always provide InnoDB storage engine support in later MariaDB versions. we are also studying why MySQL 5.1 has better performance than MariaDB 5.3, they even use the same storage engine.
In addition, Sysbench is a comprehensive benchmark testing tool. loads in real environments are affected by various factors. In addition, SSD is used in this test, and its saturation is much earlier than that of a general disk, which may also affect some test results.
If I can draw a conclusion, I would like to say that if you need the highest concurrency level, please select MySQL 5.5 or wait for MariaDB 5.5.
If your application contains very complex queries and subqueries, you will benefit from MariaDB's new optimization features. in addition, if you need the database replication function, MariaDB 5.3's group-commit will be stronger than any MySQL version.
Test environment:
- Sysbench OLTP multi-table.
- 16 tables with 250 K rows each.
- 20 GB InnoDB buffer pool.
- Data and logs on SSD.
- Dual socket hexacore Xeon with HT enabled (24 logical cores ).
You can download the script, configuration and test results for this test: sysbench-XL-2012-02-21.tar.gz
BitsCN.com