MySQL Overview

September 2004 | MySQL® Overview | www.mysql.com 1

MySQL® Overview

Aaron ByersSr. Sales Manager - Telecom & Networking


About MySQL


MySQL• Founded in 1980’s• WW Headquarters in Uppsala,

Sweden– US Headquarters in Seattle– Executive office in Cupertino

• 200 employees, in 18 countries• US $19.5M Investment

– Led by Benchmark Capital• Over 6 Million Active

Installations– Most popular Open Source

RDBMS– Dual Licensing Model


Where does MySQL fit?

Oracle, MS, IBM

Enterprise

SleepyCatPostgreSQL

Standalone

BirdstepSolid

EmbeddedEmbedded and

Distributed Data Management


MySQL Web & Enterprise

http://us.rd.yahoo.com/more/*http://www.yahoo.com

http://graphics.travelocity.com/tr/cl/1,,TRAVELOCITY%7C987,00.html?s=hp&c=logo&URL=http://www.travelocity.com?Service=TRAVELOCITY


Telecom Customers – Q1 2005


MySQL Technology


MySQL Server• High Performance RDBMS• Standards based SQL, aiming for SQL-99• Stable• Scalable

– Embedded in hardware– Extremely high load applications– Master/Slave Replication

• Easy to use• Modular

– Storage Engines– Compile and runtime feature flexibility to conserve resources


Client Library Support

• C API: libmysql• PHP utilizing C API• Connector/J, JDBC Type IV JDBC Driver• Connector/ODBC• Perl DBD::DBI• ADO.Net, OleDB, Ruby, Erlang, Eiffel, Smalltalk, etc.

supported by third parties


Server Design

• Written in C, C++, ASM• Majority (80%) in C

– client library, network communications, I/O, storage engines• Parse tree and optimizer implemented C++• String functionality in ASM on platforms that require it for

performance.


Process Model

• Multithreaded• 1 Connection means 1 Thread• Minimum overhead per client

– Network buffers (configurable, limits max query size)– Thread Stack (e.g. 128-256K on Linux)– read_buffer_size – read_rnd_buffer_size– sort_buffer_size– tmp_table_size

• Deployments with 1000s of concurrent connections not uncommon


MySQL Storage Engines


MySQL Internal Architecture

Parse OptimizeRetrieve

Store

MySQL

MyI

SA

M

Inno

DB

MyS

QLC

lust

er

MySQL DatabaseManagement Level

Table Handler /Storage Engine

Level

Hea

p

Cus

tom


What properties are Storage Engine dependent?

• Very few!– Transactional capabilities– Locking– Backup and Recovery– Optimization (partly)– Pricing (licenses, support)

• Storage Engines are Transparent– Tables of different types within one query


MyISAM

• Fastest for read or write applications, since there are no transaction overheads

• Large-grained (table-level) locks• Non-locking inserts when inserted at end of table (logging

applications)• Less disk space since there is no transaction log or

multiversioning overhead• Uses less memory for updates


Heap

• Stored in memory• Does not survive server restart• Hash or Btree indexes• No partial-key index searches• Fixed-length record format• No BLOB/TEXT support


Redo

Log

The InnoDB Storage EngineIn memory On disk (workspace)

Log

files

Tabl

e sp

ace

Log File 1Log File 2Log File 3

Buffer pool(buffered

data pages)

Add’t’l Mem Pool

Log Buffer(buffered

log records)

ibdata2data file

ibdata1data file

Undo Log

COMMIT(+ checkpoint)

checkpoint


InnoDB• Full ACID compliance• Row level locking with Next-Key Locking• Multi-Versioning

– Consistent non-locking read• Commit and rollback segments

– Fault and crash tolerant• Foreign Key Constraints• Large Tablespaces, Raw Partitions, Hot Backups• Transparent with other MySQL storage engines• Tuned for performance and large data sets• Next-generation index and data storage

– Clustered Indexes


InnoDB Size Limits• Columns per table: 1000

– hard limit since it is encoded in 10 bits in the transaction header

• Max row size: 4 GB

• Maximum tablespace size: 64 TB

• Transaction size: Unlimited– limited by tablespace size


MySQL Cluster• Acquisition of a database unit from Ericsson, people and all IP

• Integrated as new Storage Engine for MySQL Server– MySQL Cluster

• Draws from Ericsson’s extensive experience in buildingreliable solutions for the Telecom/IP industry

• Parallel server architecture with unmatchedavailability, performance, and scalability


Cluster Advantages• Flexible and cost-efficient scaling

– Scales as one logical database• Transparent and independent scaling of database and application• Easier development• Easier to maintain in operation (as opposed to partitioned database

instances)– Demonstrated linear scalability on low-cost hardware

• Lower hardware cost– No need for expensive specialized hardware

• Performance– In-memory performance

• Carrier grade availability characteristics– Availability mechanisms transparent to application


MySQL Technical Features


Supported Operating Systems

• AIX 4.x• Amiga• BSDI 2.x (mit-pthreads)• BSDI 3.0, 3.1, 4.x (native)• Caldera (SCO) OpenServer • Caldera UnixWare 7.0.1• DEC Unix 4.x• FreeBSD 2.x, 3.x, 4.x• HP-UX 10.20 / 11.x • Linux 2.0+

• Mac OS X• NetBSD• Novell Netware 6.5+• OpenBSD• QNX• SGI Irix 6.x • Solaris (SPARC, x86)• SunOS 4.x • Tru64 Unix • Windows 95, 98, Me, NT,

2000, XP, 2003


Supported Development ToolsC API based• PHP • Perl• ODBC• C • C++• Python • Ruby• Tcl • Eiffel• Lasso• Common LISP• Cobol

Other• Java (Connector/J)• .NET

Through ODBC• VB, VBA• Word, Excel, Access• Delphi• ASP• ADO• .NET


Standards

• What standards does MySQL follow?– Entry Level SQL-92– ODBC Level 0-3.51– JDBC Type IV Driver– JDBC Spec 3.0 to fullest MySQL capabilities

• Goals– Full SQL-99 compliance– Adopt new upcoming SQL-2003 specification


Advanced Features

• Advanced features MySQL supports– Multiple storage engines– NDB Cluster– Replication– Embeddable– Internationalization– Fulltext searching– Spatial Data

• Lets discuss each of the above in a bit more detail


Replicationwww.mysql.com/replication

• Details– Very easy to setup and configure– Very light-weight on network– Implemented as master-slave– Asynchronous logical replication

• Why?– Allows for virtually unlimited scalability of read queries– Useful for hot standby/high availability– Load balancing


Embeddable

• MySQL normally runs as a standalone daemon• Alternatively can be embedded

– MySQL + Your application = Single executable– Ability to be embedded in C or C++ programs– Footprint can vary

• Minimum about 4 MB RAM/ 4MB HDD• Depends on various factors

– Features required– Operating System– Libraries used

– Very easy to do, part of base C API


Internationalizationwww.mysql.com/unicode

• Fully supported on many levels in 4.1• Prior to 4.1, declared on a server level• Able to declare both character set and collation (sorting

order)• Many levels where they are able to be set (server,

database, table, column)• Can mix them even in the same table


Fulltext Searchwww.mysql.com/fulltext

• What?– Natural language matching technology– Given a string to match and it determines relevancy– Results sorted by relevancy– Allows boolean matching

• Why?– Very easy search engines– Much faster than freeform LIKE statements– Generally better results than LIKE


Spatial Data Manipulationwww.mysql.com/gis

• What?– Added in MySQL 4.1– Enables ability to store geometry shapes such as point,

line, polygon, etc. in both WKT and WKB formats– Fully searchable using optimized R-Tree indexes

• Why?– Very useful for GIS applications– Useful for other applications as well

• Find all stores in a set radius


The eWeek Database Benchmark

First ever fully independent benchmark


eWeek Independent Benchmark• January 2002/Updated July

2003• According to eWeek, the

benchmark was the first independent full scale test of database management systems.

• Databases tested:– MySQL 4.0– Oracle 9i– IBM DB2 7.2– Microsoft SQL Server SP2– Sybase 12.5


eWeek Latency Test● Response times of the

different databases

● It better to be at the bottom

● MySQL is the yellow line under the red Oracle line

The test was a virtual web bookstore with 50 Million books. It used a few large Windows servers and about 50 clients machines. Source code to the tests have been published!

MySQL Overview

Documents

Transcript of MySQL Overview