What database does Facebook use is one of the most common questions asked when folks start taking about what database is the most scalable for large scale web applications.   In fact, it is usually a person who is an open source proponent, and knows very well that Facebook uses MySQL as their core database engine.  Because of this fact, this is often the single biggest reason that developers use to push to get MySQL used in their company.  I would imagine that is why it is a very popular Google query.

While Facebook uses MySQL, they do not use it as-is out of the box.  In fact, their team has submitted numerous high-performance enhancements to the MySQL core and Innodb plug-in.  Their main focus has been on adding performance counters to Innodb.  Other changes focused on the IO sub-system, including the following new features :

• innodb_io_capacity – sets the IO capacity of the server to determine rate limits for background IO
• innodb_read_io_threads, innodb_write_io_threads – set the number of background IO threads
• innodb_max_merged_io – sets the maximum number of adjacent IO requests that may be merged into a large IO request

Facebook uses MySQL as a key-value store in which data is randomly distributed across a large set of logical instances. These logical instances are spread out across physical nodes and load balancing is done at the physical node level.  Facebook has developed a partitioning scheme in which a global ID is assigned to all user data. They also have a custom archiving scheme that is based on how frequent and recent data is on a per-user basis. Most data is distributed randomly.  Amazingly, it has been rumored that Facebook has 1800 MySQL servers, but only 3 full-time DBAs.

Facebook primarily uses MySQL for structured data storage such as wall posts, user information, etc. This data is replicated between their various data centers. For blob storage (photos, video, etc.), Facebook makes use of a custom solution that involves a CDN externally and NFS internally.

It is also important to note that Facebook makes heavy use of Memcache,  a memory caching system that is used to speed up dynamic database-driven websites by caching data and objects in RAM to reduce reading time. Memcache is Facebook’s primary form of caching and greatly reduces the database load. Having a caching system allows Facebook to be as fast as it is at recalling your data. If it doesn’t have to go to the database it will just fetch your data from the cache based on your user ID.

So, while “What database does Facebook use?” seems like a simple question, you can see that they have added a variety of other systems to make it truly web scalable.  But, still feel free to use the argument, “MySQL is as good or better than Oracle or MS SQL Server, heck, even Facebook uses it, and they have 500 Million users!”.

Related articles

• Facebook Showcasing Its Open Source Database (allfacebook.com)
• Google Chrome’s new False Start “Feature” (z-car.com)
• Video: How Facebook Makes MySQL Scale (readwriteweb.com)

mysql> SELECT id, technology from Enigma;
+——+————-+
| Id | technology |
+——+————-+
| 1 | PHP |
| 2 | LINUX |
| 3 | MySQL |
+——+————-+

To sort the records based on the technology, do an ORDER BY on column Technology:

mysql> SELECT id, technology from Enigma ORDER BY Technology;
+——+————-+
| Id | technology |
+——+————-+
| 1 | PHP |
| 2 | LINUX |
| 3 | MySQL |
+——+————-+

It seems that the server does not understand what I have asked, or I have not mastered the entire alphabet …
Let’s add the clause CSA (just in case!)

mysql> SELECT id, technology from Enigma ORDER BY ASC Technology;
+——+————-+
| Id | technology |
+——+————-+
| 1 | PHP |
| 2 | LINUX |
| 3 | MySQL |
+——+————-+

That does not seem to be any better!  MySQL does not sort the data as I wish, am I going crazy? Should I change my RDBMS?
Don’t worry, a careful look at the table structure allows us to see things clearly.

mysql> SHOW CREATE TABLE enigma;

Table: enigma
Create Table: CREATE TABLE `Enigma` (
`Id` int (11) DEFAULT NULL,
`Technology` enum (‘PHP’, ‘Linux’, ‘MySQL’) DEFAULT NULL
) ENGINE = MyISAM DEFAULT CHARSET = latin1

The column technology is of a type ENUM.  The Enum field is stored as an integer, and that is the index that MySQL uses for sorting.  In other words, 1 is PHP, 3 is MySQL, and 2 is Linux, so sorting does works properly, just not as expected when the Enum data field is storing character data.  So how do you get the desired result?  Simply force MySQL to use the value string and not index when sorting.   This can be done by using either the Concat() or Cast() functions, as follows:

function concat ()
mysql> SELECT id, technology from Enigma ORDER BY concat (technology);
+——+————-+
| Id | technology |
+——+————-+
| 2 | LINUX |
| 3 | MySQL |
| 1 | PHP |
+——+————-+
function cast ()
mysql> SELECT id, technology from Enigma ORDER BY cast (technology as char);
+——+————-+
| id | Technology |
+—– -+————-+
| 2 | LINUX |
| 3 | MySQL |
| 1 | PHP |
+——+———– – +

Want a quick and dirty way to Truncate all the tables in a MySQL DB?  You can use the following one line script.

mysqldump -ppassword YourDBName –no-data dumpfile | mysql dumpfile

What this does is uses mysqldump to output all the tables in the database “YourDBName” into a dump file.  This dump file contains drop statements, and create table statements for each table in the database.  What this effectively does is drops every table, and recreates it WITHOUT any data inside.  Be very careful, since this is a very powerful script, it has the ability to completely wipe out your database with no hope of recovery unless you have a current backup.

For additional ways to execute specific commands against all tables in a DB, look into using mk-find in Maatkit, it is an excellent tool for advanced MySQL scripting.

Related articles

• Exiting a MySQL Stored Procedure in the middle of the code (prodromus.com)
• MySQL 5.5.8 and Percona Server on Fast Flash card (Virident tachIOn) (mysqlperformanceblog.com)

The simplest way to attack this is by using a regular expression. For example:

SELECT * FROM BadlyDesignedTable WHERE AnswerColumn regexp ‘^[0-9]+’;

or
SELECT * FROM BadlyDesignedTable WHERE AnswerColumn RLIKE ‘^[0-9]+’;

Regex and RLIKE are compatible keywords that both represent regular expression matching.

The regex is described as follows :

^    - Start anchor, used to ensure the pattern matches start of the string.
[    - Start of character class.
0-9  - Any digit
]    - End of character class

Related articles

• How can I select this in SQL? (ask.metafilter.com)
• Test Regular Expressions in Your Browser (mt-soft.com.ar)

As an example, look at the following Stored Procedure :

CREATE PROCEDURE TestProc(Value INT)

ThisSP:BEGIN

IF  Value is null or Value=0 then
Select ‘Invalid Value’;
LEAVE ThisSP;
END IF;

Select * from Table twhere t.Value=Value;

END;

There are multiple ways in which to remove all data from a MySQL Table.

The first is to use the DROP TABLE command, which will remove the entire table with data.  This basically removes the entire table, you will not be able to recover the structure or data.

The second is DELETE * FROM Table which will remove the data,  leaving the auto-increment values alone, as well as the rest of the table structure.  This method is very slow on large tables using Innodb.  This is because each record has to be written to the log.  In general, you should never use this method to remove all data from a table.

The third method is to issue a TRUNCATE TABLE command which will quickly remove all the data from the table, reset the auto-increment values (but leave them as auto-increment columns, so they will just start at 1 and go up from there again).

Lastly, you can use the nuclear option, which is as follows :

mysqldump -ppassword YourDBName YourTblName –no-data dumpfile | mysql dumpfile

This will drop the table, and then recreate it all from scratch.

Related articles

• MySQL – A simple script to truncate all MySQL tables in a Database (prodromus.com)
• Sharing an auto_increment value across multiple MySQL tables (revisited) (mysqlperformanceblog.com)

Don’t worry, this is an easy to recover from situation. Just follow these basic steps to reset the root user password.

Stop the MySQL server process.

Start the MySQL server process with the –skip-grant-tables option. This option causes the server to start without using the privilege system at all, which gives anyone with access to the server unrestricted access to all databases. BEWARE! Anyone can access your server, so make sure off the Internet, and that you perform the following steps as quickly as possible.

Start the MySQL console client with the -u root option. Mysql -u root

SELECT * FROM mysql.user;

UPDATE mysql.user SET Password=PASSWORD(‘[password]‘) WHERE User=’[username]‘; Replace [username] with root to change root. You can also change any other user as well.

Stop the MySQL process

Start the MySQL Process normally (i.e. without the –skip-grant-tables option).

Some folks have reported that issuing a Flush Privileges command will prevent your having to perform the final Stop/Start of the MySQL server. However, this has not always worked for me, and I think to be safe, it is wise to recycle the server.

Luckily, there is a fairly easy way to speed up shutdown while still allowing the server to stay up and handle requests.  The secret is to lower the maximum percentage of dirty pages allowed.  This is controlled by the global variable innodb_max_dirty_pages_pct. On my servers I usually have this set at 75%. This variable can be changed at run-time, which allows us to modify this variable to speed up system shutdown.

First, enter the command :

set global innodb_max_dirty_pages_pct=0;

You can then enter the following to verify the new value:

show global variables where variable_name rlike ‘dirty’;

To monitor the number of dirty pages that have not been written to disk, enter the following:

show global status where variable_name rlike ‘dirty’;

You will see this number begin to drop, although it is unlikely to ever reach 0, once it drops to a low level and reaches a plateau, you can then shutdown the server. You will find that the server will shutdown very quickly, allowing you to minimize downtime.

select * from mytable where first_name collate latin1_general_cs=’damon’

This will only return rows will the first_name matches damon exactly. Damon will not match.

You can also use latin1_bin which performs an exact binary match.

If you are unsure of a field or strings collation, you can use the COLLATION(str) function to return the collation of the string argument.

Related articles

• MySQL – Disabling Binary Logging for INSERT and UPDATE (prodromus.com)