What is MariaDB Server 10.3?

MariaDB 10.3
Colin Charles, Chief Evangelist, Percona Inc.

colin.charles@percona.com / byte@bytebot.net

http://bytebot.net/blog/ | @bytebot on Twitter

Percona Webinar

26 June 2018

whoami
• Chief Evangelist, Percona Inc

• Focusing on the MySQL ecosystem (MySQL, Percona Server, MariaDB Server),
as well as the MongoDB ecosystem (Percona Server for MongoDB) + 100%
open source tools from Percona like Percona Monitoring & Management,
Percona xtrabackup, Percona Toolkit, etc. Now supporting PostgreSQL too!
“To champion unbiased open source database solutions”

• Founding team of MariaDB Server (2009-2016) [Monty Program Ab, merged with
SkySQL Ab, now MariaDB Corporation]

• Formerly MySQL AB (exit: Sun Microsystems)

• Past lives include The Fedora Project (FESCO), OpenOffice.org

• MySQL Community Contributor of the Year Award winner 2014

License
• Creative Commons BY-NC-SA 4.0

• https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode

What is MariaDB Server?
• https://mariadb.org/about/

What is MariaDB TX 3.0?
• A combination of all the offerings MariaDB Corporation & MariaDB
Foundation work on

• MariaDB Foundation: MariaDB Server (GPLv2), MariaDB Connectors for C/
Java/ODBC (LGPL)

• MariaDB Corporation: services (remote DBA, migration, consulting,
technical support) + MariaDB MaxScale proxy (Business Source License) +
MariaDB Backup (fork of Percona XtraBackup), MariaDB Admin (SQLYog) +
Monitor (Monyog) + Notifications (security alerts through a portal)

• TX Cluster includes support for Galera Cluster; AX for MariaDB
ColumnStore

Key focus points for MariaDB 10.3
• Oracle compatibility

• More storage engines

• Temporal data (system versioned tables)

• Plus some of the features from 10.2+10.1+10.0+5.5+5.3+5.2+5.1
that may not be in stock MySQL

Release dates (and a basis for comparison)
MySQL MariaDB
10.1: 17 October 2015
5.7: 21 October 2015
10.2: 23 May 2017
8.0: 19 April 2018
10.3: 25 May 2018

Let’s talk storage engines
• Storage: how the data is stored on disk

• Or in NDB (memory+disk), CassandraSE
(access a Cassandra Cluster), SphinxSE
(access the Sphinx daemon)

• Indexes: improves search operations

• Memory usage: improves data access for
speed

• Transactions: protects the integrity of
your data (Atomic-Consistent-Isolated-
Durable - ACID)

• Locking level: MyISAM (table locks),
InnoDB (row locks), old BDB (page locks)

• Data types: Data types may be
converted, MEMORY doesn’t support
TEXT, etc.

• Caching: InnoDB caches data & indexes,
MyISAM caches indexes only (relying on
OS disk cache for data)

• Full-text search capability: MyISAM has
this, InnoDB 5.6 got this

• GIS: MyISAM & Aria work (R-tree indexes
exist), InnoDB 5.7 has this

• Backups

• Foreign Keys

MariaDB storage engine offerings
• MyRocks: for write-intensive
workloads

• SPIDER: for scalability and
sharding

• InnoDB: default for read/write
operations (no longer Percona
XtraDB since MariaDB 10.2)

• ColumnStore: analytical purposes
(not included in MariaDB Server
10.3 — still a separate download)

• OQGRAPH: leaves algorithm

• note: requires libJudy

• PARTITION: updates to make
SPIDER work better

• Cassandra: still around, requires
libthrift

• CONNECT: for ETL operations

• TokuDB: requires jemalloc and
transparent hugepages to be
never (not always)

MariaDB storage engine aims
• To be a general purpose database, with purpose-built storage
engines

• Great focus on use cases

• Maximise strengths, minimise weaknesses

• InnoDB: good general purpose, balance B+-tree based engine

• MyRocks: LSM-based, fast writes, great compression

• SPIDER: sharding purposes

MyRocks
• RocksDB (Facebook) is a fork of LevelDB (Google). MyRocks is the interface to it
from MySQL/MariaDB

• Write optimised

• Focus on the endurance of flash devices to gain better lifetime (10x less write
amplification)

• Better compression than InnoDB (at least 2x)

• Ability to load data fast, avoiding compaction overheads

• Read-free replication (no random reads for updating secondary keys, only for
unique indexes; RFR does away with it all, with row-based binlog)

• Recommended read: https://mariadb.com/kb/en/library/differences-between-
myrocks-variants/

SPIDER
• Transparent sharding and re-sharding via SQL

• Partition by range/key/hash/list

• vertical partitioning engine, allows partition by columns

• Condition pushdown to the storage engine layer

• JOIN, GROUP BY done internally (on the data nodes/shards)

• direct updates/deletes (pushdown to data nodes)

• direct aggregates (sums, min, max, avg through partition engine)

• Partition improvements: full-text support, multi-range read (MRR)

• Read the docs, please! https://mariadb.com/kb/en/library/spider-storage-
engine-overview/

So the benefit of storage engines…
• mixing and matching

• sure, there are limitations…

• e.g. Galera Cluster only works with the InnoDB backend

• Feel free to mix InnoDB and MyRocks. Or get data from CONNECT
and load it into InnoDB. And so on…

Compression
• Row compression (ROW_FORMAT=COMPRESSED), to page
compression (PAGE_COMPRESSED=1), now to column compression

• Bonus? Storage engine independent

CREATE TABLE ùsers` (
ìd` int(11) NOT NULL,
`name` varchar(100) DEFAULT NULL,
`blurb` text /*!100301 COMPRESSED*/ DEFAULT NULL,
PRIMARY KEY (ìd`)
) ENGINE=ROCKSDB DEFAULT CHARSET=latin1

Compression
show status like 'column_%';
+-----------------------+-------+
| Variable_name | Value |
+-----------------------+-------+
| Column_compressions | 1 |
| Column_decompressions | 0 |
+-----------------------+-------+

Invisible columns
CREATE TABLE ùser` (
ìd` int(11) NOT NULL,
`name` varchar(100) DEFAULT NULL,
`secret` varchar(10) INVISIBLE DEFAULT
NULL,
PRIMARY KEY (ìd`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1
insert into user (id,name,secret) values
("1","colin","yes");
select * from user;
+----+-------+
| id | name |
+----+-------+
| 1 | colin |
+——+-------+
select id,name,secret from user;
+----+-------+--------+
| id | name | secret |
+----+-------+--------+
| 1 | colin | yes |
+----+-------+--------+

INSTANT ADD COLUMN
• contributed by Tencent to both MariaDB and MySQL

• https://mysqlserverteam.com/mysql-8-0-innodb-now-supports-
instant-add-column/ + https://mariadb.com/kb/en/instant-add-
column-for-innodb/

• inserts a hidden row in the table, updates the data dictionary, only
works with the last column

System versioned tables
• SQL 2011 standard. Stores history of all changes.

• Can alter a table to enable/disable/remove system versioned data

• Queries?

• AS OF to select data as of a point in time

• BETWEEN .. AND to select data between two points in time

• Partition data BY SYSTEM_TIME

• Just ALTER .. ADD SYSTEM VERSIONING or create a table WITH
SYSTEM VERSIONING

System versioned tables
create table employees (name varchar(10), salary int,
department varchar(10)) with system versioning;
insert into employees values ("colin", 1000, “mktg");
update employees set salary=10000 where name=“colin";
update employees set department="eng" where
name=“colin"
select * from employees where name="colin";
+-------+--------+------+
| name | salary | dept |
+-------+--------+------+
| colin | 10000 | eng |
+———+--------+------+
select *, ROW_START, ROW_END from employees for
SYSTEM_TIME ALL;
+-------+--------+------------
+----------------------------
+----------------------------+
| name | salary | department | ROW_START
| ROW_END |
+-------+--------+------------
+----------------------------
+----------------------------+
| colin | 10000 | eng | 2018-06-26
13:00:53.772241 | 2038-01-19 03:14:07.999999 |
| colin | 1000 | mktg | 2018-06-26
13:00:03.656662 | 2018-06-26 13:00:24.251594 |
| colin | 10000 | mktg | 2018-06-26
13:00:24.251594 | 2018-06-26 13:00:53.772241 |
+-------+--------+------------
+----------------------------
+----------------------------+

Oracle compatibility - Sequences
• Sequences to create a sequence of numeric values

• Not to be confused with replacing AUTO_INCREMENT, is an
alternative to creating unique identifiers

• With a sequence, you can compute the last number created by all
existing sequences, whereas AUTO_INCREMENT can only compute
its own last number created

Sequences
create sequence seq start
with 10 increment by 10;

select nextval(seq);

+--------------+

| nextval(seq) |

+--------------+

| 10 |

+--------------+


+--------------+

| nextval(seq) |

+--------------+

| 20 |

+--------------+


+--------------+

| nextval(seq) |

+--------------+

| 30 |

+--------------+

select lastval(seq);

+--------------+

| lastval(seq) |

+--------------+

| 30 |

+--------------+

select previous value for
seq;

+------------------------+

| previous value for seq |

+------------------------+

| 30 |

+------------------------+

Oracle PL/SQL
• PL/SQL compatibility parser added for easier migration from Oracle to MariaDB

• sql_mode=‘oracle’

• Data types (have synonyms in MariaDB): VARCHAR2 (VARCHAR), NUMBER (DECIMAL),
DATE (DATETIME), RAW (VARBINARY), BLOB (LONGBLOB), CLOB (LONGTEXT)

• CURRVAL, NEXTVAL

• EXECUTE IMMEDIATE

• Existing stored procedures, triggers

• ROW datatype for stored routines

• Cursors with parameters

• Packages

• https://mariadb.com/kb/en/library/sql_modeoracle-from-mariadb-103/

Proxy support
• Client can connect to MariaDB 10.3 via a proxy without the need to
define user privileges based on the host of the proxy

• Proxy protocol allows the proxy to provide the client IP to the server

• With audit plugin, logged client IP is now real client IP not proxy IP

Let’s not forget the goodness of MariaDB
10.2, 10.1
• Window Functions

• Recursive Common Table
Expressions

• JSON, GeoJSON functions

• Time delayed replication

• Restrict speed of reading binlog
from master

• Compressed binlog

• CHECK CONSTRAINT

• DECIMAL increased to 38 digits

• EXECUTE IMMEDIATE

• New user management functions
(SHOW CREATE USER)

• Information schema supports
user defined variables

• Binary log based Flashback for
DML statements

• Indexes for virtual columns

Other bits
• INTERSECT and EXCEPT to UNION

• Stored aggregate functions - functions that are computed over a
sequence of rows and return one result for the sequence of rows

• Idle transaction timeouts

• idle_transaction_timeout, idle_readonly_transaction_timeout,
idle_write_transaction_timeout

Some gotchas if you’re coming from MySQL
8
• JSON is not stored as a binary data type

• GTIDs are different in MariaDB (e.g. no
GTID in OK packet)

• No X Protocol, mysqlsh support

• No group replication

• PERFORMANCE_SCHEMA from MySQL
5.6

• No caching_sha256_password (ed25519)

• mysql.user.password now is
mysql.user.authentication_string

• No password expiry, last changed, etc.
however there is cracklib_password_check

• No optimiser hints, optimiser trace

• Threadpool in MariaDB!

• PAM/GSSAPI/SSPI authentication

• AWS Key Management Plugin

• Table elimination!

• User statistics

• Dynamic columns

• No SET PERSIST

• InnoDB comes from MySQL 5.7

Thank you! Please rate in the app!
Colin Charles
colin.charles@percona.com / byte@bytebot.net
http://bytebot.net/blog | @bytebot on twitter
slides: slideshare.net/bytebot

What is MariaDB Server 10.3?

Recommended

More Related Content

What's hot (20)

Similar to What is MariaDB Server 10.3? (20)

More from Colin Charles (15)

Recently uploaded (20)

What is MariaDB Server 10.3?