Improving MariaDB/MySQL performance tuning

Published on August 13, 2021 at 10:00:08 AM GMT+8 by Administrator

Tips and tricks to improvise your database search engine

Improving MariaDB/MySQL performance tuning

To improvise your system performance developer must always make sure that their system query the database will not take a long time to fetch results.

Recent studies show that 2 second is the new threshold in terms of average waiting time for normal people to expect the web page loaded. Furthermore, the latest examination, led by Forrester Consulting, proposes that two seconds is the new edge as far as a normal online customer assumption for a Web page to stack, while 40% of customers will stand by close to three seconds prior to leaving a retail or travel website.

Hence, in this topic we will discuss a few tips that developer can take note to make sure that the database fine tuned for efficient and best performance.

Hardware and System

It is recommended that you improve your hardware in the following order:


Memory is the main factor for database as it permits you to change the Server System Variables. More memory implies bigger key and table reserves, which are put away in memory so that circles can get to, a significant degree more slow, is along these lines decreased. 

Remember however, essentially adding more memory may not bring about intense enhancements if the worker factors are not set to utilize the extra accessible memory. 

Utilizing more RAM openings on the motherboard builds the transport recurrence, and there will be more dormancy between the RAM and the CPU. This implies that utilizing the most noteworthy RAM size per opening is ideal.

Disk (I/O)

A fast disk access is basic in every computer performance related issues as it is the task where we are read and write the data, as at last it's the place where the information dwells. The key figure is the disk look for time (an estimation of how quick the actual disk can seek to get to the information).

Fast Ethernet

With the appropriate requirements for your internet bandwidth, fast ethernet means it can have faster response to clients requests, replication response time to read binary logs across the slaves, faster response times is also very important especially on Galera-based clusters.


Despite the fact that equipment bottlenecks regularly fall somewhere else, quicker processors permit computations to be performed all the more rapidly, and the outcomes sent back to the customer all the more rapidly. Other than processor speed, the processor's transport speed and reserve size are additionally significant elements to consider.

Optimize MariaDB/MySQL Configuration

It is recommended that you change the default configuration of MariaDB/MySQL depends on the amount of resources that you currently have. Below are the list to look out for.


The essential worth to change on a data set worker with entirely/primarily XtraDB/InnoDB tables, can be set up to 80% of the all out memory in these conditions. Whenever set to 2 GB or more, you will presumably need to change innodb_buffer_pool_instances too. You can set this progressively in case you are utilizing MariaDB >= 10.2.2 rendition. Else, it requires a worker restart.


For tmp_memory_table_size (tmp_table_size), in case you're managing huge impermanent tables, setting this higher gives execution gains as it will be put away in the memory. This is normal on questions that are intensely utilizing GROUP BY, UNION, or sub-inquiries. Despite the fact that in case max_heap_table_size is more modest, as far as possible will apply. On the off chance that a table surpasses the breaking point, MariaDB changes it over to a MyISAM or Aria table. You can check whether it's important to increment by contrasting the status factors Created_tmp_disk_tables and Created_tmp_tables to perceive the number of transitory tables out of the all out made should have been changed over to plate. Regularly mind boggling GROUP BY questions are answerable for surpassing the cutoff. 

While max_heap_table_size, this is the most extreme size for client made MEMORY tables. The worth set on this variable is just relevant for the recently made or re-made tables and not the current ones. The more modest of max_heap_table_size and tmp_table_size additionally restricts inner in-memory tables. At the point when the most extreme size is reached, any further endeavors to embed information will get a "table ... is full" blunder. Impermanent tables made with CREATE TEMPORARY won't be changed over to Aria, as happens with inward transitory tables, yet will likewise get a table full blunder.


Enormous recollections with rapid handling and quick I/O circle aren't new and has its sensible cost as it suggests. In case you are favoring more execution gains particularly during and dealing with your InnoDB exchanges, setting the variable innodb_log_file_size to a bigger worth like 5GB or even 10GB is sensible. Expanding implies that the bigger exchanges can run without expecting to perform disk I/O prior to submitting.

Sample of additional configuration which improve the DB query significantly:

## You can set .._buffer_pool_size up to 50 - 80 %. Assuming here we have total RAM of 8GB
## of RAM but beware of setting memory usage too high
## Set .._log_file_size to 25 % of buffer pool size