MySQL Architecture

AGENDA

• MySQL Server process

• Storage Engine
• Log files
• High Availability
• Q&A

2
Server Process

4
Server Process
 The mysqld (server program) process can be sliced into the following three layers:
 Connection layer: Handles connections. This layer exists on all software (Web/mail/LDAP
server)
 SQL layer: Processes SQL queries that are sent by connected applications
 Storage layer: Handles data storage. Data can be stored in different formats and
structures on different physical media.

5
Connection Layer

Confidentia 66
l
Connection Layer
The connection layer accepts connections from applications over
several communication protocols
 TCP/IP
 UNIX sockets
 Shared memory
 Named pipes

TCP/IP works across the network. The other protocols listed above
support only local connections when the client and server are
running on the same machine. This layer maintains one thread
per connection. This thread handles query execution. Before a
connection can begin sending SQL queries, the connection is
authenticated by verification of username+password+client host

Confidentia 77
l
 Communication Protocols

• Protocols are implemented in the client libraries and drivers.

• The speed of a connection protocol varies with the local settings.

8
SQL Layer

9
SQL Layer

 After a connection is established, the following processes are handled

by the MySQL Server:
 Authorization and parser: The parser validates the correct syntax,
and then authorization verifies
 Optimizer: Creates the execution plan for each query, which is a step-
by-step instruction set on how to execute the query in the most
optimal way. Determining which indexes are to be used, and in which
order to process the tables, is the most important part of this step.
 Query execution: Fulfills the execution plan for each query.
 Query cache: Optionally configurable query cache that can be used to
memorize (and immediately return) executed queries and results
 Query logging: Can be enabled to track executed queries

10
SQL Statement Processing

11
Storage Layer

12
Storage Engine: InnoDB

 ACID Compliant
 Atomicity/Consistency/Isolation/Durability
 Full transactional support and multi-versioning
 Read Uncommitted, Read Committed, Repeatable Read, Serializable
 Foreign keys constraints
 Locking and logging
 Row-level and next-key locking
 Consistent non-locking reads
 Commit and rollback segments
 Fault tolerance and table spaces
 Large datasets, raw partitions
 Online backups
Storage Engine: MyISAM
 File based storage
 .MYD – table data
 .MYI – index data
 .FRM – table definition (schema)
 Easily maintained
 Architecture-independent data
 Files can be copied across platforms
 Low overhead
 No transactions
 Large grained table level locking
 Excels at mostly-read applications
 One third the memory/disk footprint of transactional engines
Storage Layer
 Data can be stored on disk, memory, and network. Each table in
a database can use any of the available storage engines. Disk
storage is cheap and persistent, where as memory is much faster.
 InnoDB is the default storage engine. It provides transactions,
full-text indexing, and foreign key constraints, and is thus useful
for a wide mix of queries. It is multipurpose and supports read-
intensive, read/write, and transactional workloads.
 Other storage engines include:
 MyISAM: Useful for mostly read and little update of data
 MEMORY: Stores all data in memory
 NDB: Used by MySQL Cluster to provide redundant scalable topology for highly
available data

15
Storage Engine: Overview

16
 Features Dependent on Storage Engine

• Storage medium
• Transactional capabilities
• Locking
• Backup and recovery
• Optimization
• Special features
– Full-text search
– Referential integrity
– Spatial data handling

17
How MySQL Uses Disk Space

18
How MySQL Uses Disk Space
 Program files are stored under server installation directories,
along with the data directory. The program executable and log
files are created by the execution of the various client,
administrative, and utility programs. The primary use of disk
space is the data directory.
 Server log files and status file contain information about
statements that the server processes. Logs are used for
troubleshooting, monitoring, replication, and recovery.
 InnoDB log files for all databases reside at the data directory
level
 InnoDB System Tablespace contains the data dictionary, undo
log, and buffers.

19
How MySQL Uses Disk Space
 Each database has a single directory under the data directory,
regardless of what types of tables are created in the database.
The database directories store the following:
 Data files: Storage engine – specific data files. These files can also include
metadata or index information, depending on the storage engine used.
 Format files (.frm) : Contain a description of each table and/or view structure,
located in the corresponding database directory.
 Triggers: Named database objects that are associated with a table and are
activated when a particular event occurs for the table.
 The location of the data directory depends on the configuration, operating
system, installation package, and distribution. A typical location is /var/lib/mysql.
 MySQL stores the system database (mysql) on disk. Mysql contains information
such as users, privileges, plugins, help lists, events, time-zone implementations,
and stored routines

20
 How MySQL Uses Memory

 Global
 Allocated once
 Shared by the server process and its threads

 Session
 Allocated for each thread
 Dynamically allocated and deallocated
 Used for handling query results
 Buffer sizes per session

21
How MySQL Uses Memory
 Memory allocation can be divided into the following two
categories:
 Global (per-instance memory): Allocated once when the
server starts and frees when the server shuts down. This
memory is shared across all sessions.
 When all the physical memory has been used up, the
operating system starts swapping. This has an adverse effect
on MySQL server performance and can cause the server to
crash.
 Session (per-session memory): Dynamically allocated per
session (sometimes referred to as thread). This memory can
be freed when the session ends or is no longer needed.

22
 Memory Structures
Server allocates memory in three different
categories:

23
 Summary

• Describe the MySQL client/server model

• Understand communication protocols
• Understand how the server supports storage
engines
• Explain the basics of how MySQL uses
memory and disk space

24
Thank You

25