Download Fudcon talk

MySQL and Fedora:
A Developer's Overview
Brian Aker
Director of Architecture
Fudcon
April 2006
MySQL AB
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
Who am I?
• Brian Aker
– Director of Architecture, MySQL AB
– Author of mod_layout, the apache
streaming services mod_mp3, Slash
(Slashdot’s CMS System), and lot of
other things on Freshmeat....
– http://mysql.com/
– http://krow.net/
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
MySQL
• Open Source Database Company
• Swedish company started by Monty
Widenius and David Axmark
• Employs Developers World Wide
– 200^H^H^H300 employees
– 26+ countries
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
Agenda
• What's in MySQL 5.0
• Fitting in with the Fedora Community
• Differences in packaging
• Contributing to the OSS movement
• MySQL and India
• Resources
• What is coming in the Future?
• Q&A?
The World’s Most Popular Open Source Database
What is in 5.0?
• Views
• Stored Procedures
• Triggers
• GUI Tools
• Precision Math
• Cursors
• Instance Manager
• Archive
• XA
• More…..
Copyright 2005 MySQL AB
• Cluster
• Information Schema
The World’s Most Popular Open Source Database
CREATE VIEW Syntax
CREATE [OR REPLACE]
[ALGORITHM =
{MERGE | TEMPTABLE | UNDEFINED}]
VIEW view_name
[(column_list)]
AS select_statement
[WITH [CASCADED | LOCAL] CHECK OPTION]
CREATE OR REPLACE VIEW view1 ...
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
Stored Procedure Syntax
• CREATE PROCEDURE procedure_name
([parameter [...]])
[characteristic ...]
routine_body
• CREATE FUNCTION function_name
([parameter [,...]])
RETURNS data_type
[characteristic ...]
routine_body
• ALTER and DROP also fully supported
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
CREATE TRIGGER Syntax
• CREATE TRIGGER trigger_name
{BEFORE | AFTER}
{INSERT | UPDATE | DELETE} ON table_name
FOR EACH ROW trigger_action
• CREATE TRIGGER tg1
AFTER INSERT ON t1
FOR EACH ROW <compound SQL statement>;
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
MySQL 5.0: Precision Math
• Exact calculations with well defined rounding
• At least 56 digits precision
• Very fast with static memory allocation
Sample:
create table d2 (n decimal(64,3));
insert into d2 values (233221213212312312321321321321),
(34543543324325435435435),
(32432432432454374435234543456);
Query OK, 3 rows affected (0.00 sec)
select sum(n) from d2;
+------------------------------------+
| sum(n)
|
+------------------------------------+
| 531307360376620022163982600424.000 |
+------------------------------------+
1 row in set (0.01 sec)
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
XA Support
• MySQL supports distributed transaction processing:
– Two-phase commit with XA protocol available in InnoDB as of 5.0
– Commit grouping to improve performance
– XA JDBC driver
• In MySQL:
– the server acts as a transaction manager or resource manager
– storage engines can be implemented as a resource manager
Application
MySQL
 MySQL acts as RM
 Queries on ACID TX
tables included
5.0
Copyright 2005 MySQL AB
TX Scope
Other XA RM
 Included in app TX
MySQL
OR
TX Scope
XA Engine1
 MySQL acts as TM
 Storage engines act as
RM
XA Engine 2
 Different XA engine can
be in same TX
InnoDB is the first XA-Compliant Storage Engine
5.1
The World’s Most Popular Open Source Database
MySQL Cluster
• Query Cache support
• Multi Read Ranges
– batched reads for IN clause
conditions)
Copyright 2005 MySQL AB
•
Objects in Clusters can now
reach 20,000 without a recompile
(before it was 1600)
The World’s Most Popular Open Source Database
Information Schema
• INFORMATION_SCHEMA is a new, virtual database.
• INFORMATION_SCHEMA can be queried via a SELECT
statement, so there is no need to learn a new set of
commands to be able to access the metadata you need.
• MySQL creates and populates the tables automatically.
mysql> SELECT table_name, table_type, engine
-> FROM INFORMATION_SCHEMA.tables
-> WHERE table_schema = 'tp'
-> ORDER BY table_type, table_name DESC;
+------------+------------+--------+
| table_name | table_type | engine |
+------------+------------+--------+
| t2
| BASE TABLE | MyISAM |
| t1
| BASE TABLE | InnoDB |
| v1
| VIEW
| NULL
|
+------------+------------+--------+
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
Archive
• New Storage Engine
• Substantially reduces storage requirements >50%
• Information is stored packed via gzip, unzipped on the
fly
• Insert and Select only
• Great for historical data you need to keep online but
not for blazing response times
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
Where we fit in with the Fedora Community
• Just a package out of 1,500
– (okay, really 13/1795)
• Want to forge better relationships
– If you're a LAMP user, on Fedora, it should all “just work”
• Ship more Extras packages that rely on MySQL in FC
– These could be our tools (Query Browser, Adminstrator)
– Could be other apps (mytop, Gforge, SequioaERP)
– Working out “Alternatives”, like the NDB cluster, is the hard part
• Fix bugs with the package maintainer
The World’s Most Popular Open Source Database
How Red Hat and MySQL package differently
• RPMs are both provided at dev.mysql.com or within Fedora
• For the longest time, Fedora only shipped MySQL 3.23
– There were licensing misconceptions with MySQL's FLOSS
exception
– A lot of applications built against 3.23 client libraries
• #include <asm/atomic.h>
– Depends on kernel headers to build
• sparc/sparc64/ppc64 don't enable atomic_* functions as inlines
• i386: kernel not compiled for SMP -> compiled code will not be
SMP-safe (omits lock prefix before instruction)
The World’s Most Popular Open Source Database
How Red Hat and MySQL package differently
II
• No access to the extra storage engines:
– NDB Cluster
– ARCHIVE (large data store, no indexes, small footprint)
– CSV
– BLACKHOLE
– FEDERATED (stores data in remote databases; MySQL specific)
• Get the -max package to satisfy these requirements
• mysql-shared-standard required for upgrades
The World’s Most Popular Open Source Database
Some Core/MySQL statistics
• MySQL versions in various Cores:
– Core 1&2: 3.23.58 (rebuilt in Core 3)
– Core 4: 4.1.11
– Core 5: 5.0.18
• In FC-4, packages that depend on the 3.23 client libraries:
– mysqlclient10-devel, MyODBC, libdbi-dbd-mysql
• FC-5 ships the 3.23 client libraries, but nothing else
depends on it (besides its -devel package)
– mysqlclient14 (4.1.14) is provided
The World’s Most Popular Open Source Database
How you can contribute to the OSS
movement
• With Fedora, you have Extras
– package maintenance, documentation, translation
– ...or contribute to a package within Fedora, with upstream
• MySQL has many entry points
• Translation of documentation/applications
– Get access to translation repository
– In locale directory, find how_to_translate.txt
– Sign Translation Agreement, granting MySQL AB all exclusive rights
to translation
– Requires: Love for your language/dialect, and proficient English
The World’s Most Popular Open Source Database
How you can contribute to the OSS
movement II
• Conversion of applications
– Plenty of applications out there use another database
– Make them run on MySQL
– Requires: knowledge of SQL, working knowledge of application's
language DB query methods
• MySQL code-base
– BitKeeper is still used for code
• Sane build scripts exist
• http://dev.mysql.com/doc/refman/5.1/en/installing-source-tree.html
– Requires: knowledge of various bits used in MySQL
• Work on the GUI tools
• Work on the embedded version of MySQL
The World’s Most Popular Open Source Database
Resources
• We have mailing lists
– http://lists.mysql.com/
– mysql-list is very high traffic, but is a great place to learn/contribute
– Internals list if you’re into the codebase
• We have forums
– http://forums.mysql.com/
– These come with RSS feeds!
• http://dev.mysql.com/ is a great resource
• The manual is excellent!
The World’s Most Popular Open Source Database
Business Intelligence/Data Warehousing
• Table/Index Partitioning
• Full Text Search Enhancements
• Better XML Handling – XPath
• Archive engine enhancements
High Availability
• Disk-based Cluster
• Row-based Replication
• Cluster replication
Easier Manageability
• Task Scheduler
• Transaction support for Federated Engine
Higher Performance
• Faster alter table
• Faster add/drop index for MySQL Cluster
• Faster data import operations
• Better problem user and SQL identification
• New Performance/Load Testing Utility
• New MyISAM memory option
Copyright 2004 MySQL AB
The World’s Most Popular Open Source Database
Business Intelligence/Data Warehousing
Table/Index Partitioning
• Perfect for data warehouses or other VLDB situations
• Increases performance – only needed partitions are accessed
• Eases space/data management burden on DBA
• Supports range, hash, key, list, and composite methods
• Supported by all storage engines in MySQL
• More options than Microsoft, DB2, or Sybase partitioning
• Partition key must be integer (or convertible to integer)
• Local indexes only – no global primary or unique keys
• Parallelism not supported in this version
Copyright 2004 MySQL AB
The World’s Most Popular Open Source Database
Business Intelligence/Data Warehousing
XML Xpath Support
• Address parts in XML document
• Allows manipulation of
• Strings
• Booleans
• Numbers
• Models XML document as tree of nodes
Copyright 2005 MySQL AB
XML
The World’s Most Popular Open Source Database
Business Intelligence/Data Warehousing
Archive Engine Enhancements
• Faster I/O operations
• Lower Memory requirements
• Autoincrement column support
• Unique key support
• Non-unique key support
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
High Availability
MySQL Cluster Disk-Based Data
• Tables now can be designated to be disk-based
• Tablespaces used to store table disk data
• Tables can be either disk or main memory
• Indexes still main memory only
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
High Availability
Row-Based Replication
• New replication option – statement-based replication retained
• Handles all replication scenarios (deterministic, etc.)
• Safest form of replication
• Common to most other RDBMS’s
• Statement-based approach still available
• Mixed mode available that does statement and row-based replication
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
High Availability
MySQL Cluster Replication
• Replicate data from one MySQL Cluster to another
• Limitation – no cross-server logging of add/drop/alter tables
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
Higher Performance
Better Problem User/SQL Identification
• SHOW PROCESSLIST command now system table
• General and Slow Query Log now system tables
• Can easily query tables to find all or just inefficient SQL
• General/Slow Query Log use CSV engine and can be read by MS Excel
SQL
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
Higher Performance
New Performance/Load Testing Utility
• Named “mysqlslap”
• Simulates many concurrent connections to MySQL server
• Repetitively runs designated SQL load
• Can test multiple engines
• Performs creation of test schema and population of data
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
MySQL 5.1 Rollout Schedule
• Beta begins March 1, 2006
• Release Candidate expected in the Fall
• GA Fall/End of 2006
Copyright 2005 MySQL AB
The World’s Most Popular Open Source Database
Thanks!
Colin Charles
Community Engineer
colin@{mysql.com,fedoraproject.org}
The World’s Most Popular Open Source Database