Ce diaporama a bien été signalé.
Nous utilisons votre profil LinkedIn et vos données d’activité pour vous proposer des publicités personnalisées et pertinentes. Vous pouvez changer vos préférences de publicités à tout moment.
10 Reasons
why you should prefer
PostgreSQL to MySQL
Anand Chitipothu
Who am I?
Anand Chitipothu
Independent Software Consultant & Trainer
Ex. Internet Archivist
MySQL or PostgreSQL?
The world's most popular
open source database
The world's most advanced
open source database
Quick Comparison
MySQL
MyISAM
MySQL
InnoDB
PostgreSQL
Transactions No Yes Yes
Foreign Key
Constraints
No Yes Yes
Locking T...
Who Uses MySQL?
● Facebook
● FriendFeed
● Quora
● Flickr
● Second Life
Who Uses PostgreSQL
● Skype
● Heroku
● Instagram
● Disqus
● Yahoo!
● NASA
MySQL ate my cake
mysql> CREATE TABLE cake (name VARCHAR(3));
Query OK, 0 rows affected (0.02 sec)
mysql> INSERT INTO cake (name) VALUES
('p...
MySQL ate my cake
mysql> SELECT * FROM cake;
+------+
| name |
+------+
| pan |
+------+
1 row in set (0.03 sec)
OMG! Wher...
PostgreSQL?
testdb=# CREATE TABLE cake (name VARCHAR(3));
CREATE TABLE
testdb=# INSERT INTO cake (name)
VALUES ('pancake')...
Ever Seen This?
CREATE TABLE users (
id integer auto_increment,
username varchar(40),
password varchar(8)
);
Data Conversion Errors - MySQL
mysql> CREATE TABLE test (x INTEGER);
Query OK, 0 rows affected (0.00 sec)
mysql> INSERT IN...
Data Conversion Errors - MySQL
mysql> SELECT * FROM foo;
+------+
| x |
+------+
| 0 |
+------+
1 row in set (0.00 sec)
Data Conversion Errors - PostgreSQL
testdb=# CREATE TABLE test (x integer);
CREATE TABLE
testdb=# INSERT INTO test (x) VAL...
Parallels - PHP vs. Python
$ php -r '$x = "bad-number";
$y = (int)$x;
echo $y."n";'
0
Parallels - PHP vs. Python
$ python -c 'print int("bad-number")'
Traceback (most recent call last):
File "<string>", line ...
File Layout
File Layout - MySQL MyISAM
/var/lib/mysql/dbname
● dbname.MYD - data of all tables
● dbname.MYI - indexes
File Layout - MySQL InnoDB
/var/lib/mysql/
● ibdata1 - data of all databases, including
tables and indexes
It is possible ...
/var/lib/postgresql/9.3/main/base
● 131384/ - directory per database
○ 2654 - one (or more) files for each table/index
○ 2...
Database Maintenance
CREATE INDEX - MySQL MyISAM
While CREATE INDEX is in progress:
● Entire table is locked for writes
● A new index file (dbn...
CREATE INDEX - InnoDB
● Entire table rebuilt to create an index.
● Seems to have improved in recent versions
CREATE INDEX - PostgreSQL
● CREATE INDEX
○ locks the table for writes
● CREATE INDEX CONCURRENTLY
○ Doesn’t hold the lock ...
Takes long time as it needs to rewrite:
● the index file (dbname.MYI) for MyISAM
● the ibdata1 file for InnoDB
DROP INDEX ...
DROP INDEX - PostgreSQL
● Almost instantaneous
● Just need to delete the files corresponding to
that index
ADDING NEW COLUMN - MySQL
Entire table data needs to be rewritten.
ADDING NEW COLUMN - PostgreSQL
Almost instantaneous if the new column has a
default value.
Connection Model
Connection Model - MySQL
A thread for each connection
PROS
● Very easy to create a new conn
CONS
● Difficult to scale on m...
Connection Model - PostgreSQL
A process for each connection
PROS
● better concurrency
● complete isolation
● plays nicely ...
$ top
$ top
kill 17618
kill -STOP 17618
kill -CONT 17618
Query Planning
EXPLAIN
SELECT name, total
FROM names
WHERE year=1995
ORDER BY total DESC
LIMIT 10;
The Query
MySQL - Query Plan
| id | select_type | table | type | possible_keys |
+----+-------------+-------+------+---------------+...
MySQL - Add Index
CREATE INDEX names_total_idx
ON names(total)
MySQL - Query Plan With Index
| id | select_type | table | type | possible_keys |
+----+-------------+-------+------+-----...
Limit (cost=246987.39..246987.42 rows=10 width=13)
-> Sort (cost=246987.39..247467.64 rows=192099 width=13)
Sort Key: tota...
CREATE INDEX names_total_idx
ON names(total)
PostgreSQL - Add Index
Limit (cost=0.43..1891.80 rows=10 width=13)
-> Index Scan Backward using
names_total_idx on names
(cost=0.43..36332875.67 ...
PostgreSQL - Query Plan
names=# EXPLAIN ANALYZE SELECT …
Limit (cost=0.43..1891.80 rows=10 width=13)
(actual time=0.037..0...
PostgreSQL - Complex Query
EXPLAIN
SELECT name, sum(total) as count
FROM names
WHERE year > 1980
GROUP BY name
ORDER BY co...
PostgreSQL - Query Plan
Sort (cost=254889.31..255063.44 rows=69653 width=13)
Sort Key: (sum(total))
-> HashAggregate(cost=...
Replication
MySQL Replication
Replication Format
● Statement based
● Row based
● Mixed
Mode
● binlog
● GTID
PostgreSQL Replication
● Synchronous / Asynchronous
● Streaming / Log shipping
Master Slave 1
Log Shipping
Streaming
Slave...
Data Recovery
PG - Point In Time Recovery
● Postgres maintains Write Ahead Log of all
changes made to the database.
● The WAL files can ...
Other Interesting
Features of PostgreSQL
Partial Indexes
SELECT * FROM comments
WHERE email LIKE '%@spam.com';
CREATE INDEX comments_spam_idx
ON comments
WHERE ema...
Functional Indexes
SELECT tag, count(*)FROM posts
GROUP BY lower(category);
CREATE INDEX post_category_idx
ON post (lower(...
JSON datatype
CREATE TABLE book (
id serial primary key,
data JSON
);
INSERT INTO book (data) VALUES (
'{"title": "Tom Saw...
JSON datatype
SELECT * FROM book
WHERE data->>'author' = 'Mark Twain'
id| data
--+----------------------------------------...
SELECT total_time/calls AS t, calls, query
FROM pg_stat_statements
ORDER BY t DESC
8606.75|3|select name, sum(total) as co...
Summary
PostgreSQL is better than MySQL in
● Data Consistency
● Query Planning
● Stability
● Database Maintenance
● Data R...
Credits
PostgreSQL Logo - By Jeff MacDonald
http://pgfoundry.org/docman/?group_id=1000089
Thanks!
Anand Chitipothu
@anandology
Prochain SlideShare
Chargement dans…5
×

Ten Reasons Why You Should Prefer PostgreSQL to MySQL

45 605 vues

Publié le

Presented at Root Conf 2015.

https://rootconf.in/2015/

https://rootconf.talkfunnel.com/2015/40-10-reasons-why-you-should-prefer-postgresql-to-mys

Publié dans : Technologie
  • Secrets To Making Up, These secrets will help you get back together with your ex. ♥♥♥ http://scamcb.com/exback123/pdf
       Répondre 
    Voulez-vous vraiment ?  Oui  Non
    Votre message apparaîtra ici
  • mysql&gt; alter table cake add column qty integer; Query OK, 0 rows affected (0.03 sec) Records: 0 Duplicates: 0 Warnings: 0 mysql&gt; insert into cake (name, qty) values ('pancake', 'many'); ERROR 1406 (22001): Data too long for column 'name' at row 1 mysql&gt; insert into cake (name, qty) values ('pan', 'many'); ERROR 1366 (HY000): Incorrect integer value: 'many' for column 'qty' at row 1
       Répondre 
    Voulez-vous vraiment ?  Oui  Non
    Votre message apparaîtra ici
  • what about mysql strict mode?
       Répondre 
    Voulez-vous vraiment ?  Oui  Non
    Votre message apparaîtra ici
  • mysql&gt; create table cake (name varchar(3)); Query OK, 0 rows affected (0.02 sec) mysql&gt; insert into cake (name) values ('pancake'); ERROR 1406 (22001): Data too long for column 'name' at row 1
       Répondre 
    Voulez-vous vraiment ?  Oui  Non
    Votre message apparaîtra ici
  • mysql 5.7
       Répondre 
    Voulez-vous vraiment ?  Oui  Non
    Votre message apparaîtra ici

Ten Reasons Why You Should Prefer PostgreSQL to MySQL

  1. 1. 10 Reasons why you should prefer PostgreSQL to MySQL Anand Chitipothu
  2. 2. Who am I? Anand Chitipothu Independent Software Consultant & Trainer Ex. Internet Archivist
  3. 3. MySQL or PostgreSQL? The world's most popular open source database The world's most advanced open source database
  4. 4. Quick Comparison MySQL MyISAM MySQL InnoDB PostgreSQL Transactions No Yes Yes Foreign Key Constraints No Yes Yes Locking Table Row/MVCC Row/MVCC
  5. 5. Who Uses MySQL? ● Facebook ● FriendFeed ● Quora ● Flickr ● Second Life
  6. 6. Who Uses PostgreSQL ● Skype ● Heroku ● Instagram ● Disqus ● Yahoo! ● NASA
  7. 7. MySQL ate my cake
  8. 8. mysql> CREATE TABLE cake (name VARCHAR(3)); Query OK, 0 rows affected (0.02 sec) mysql> INSERT INTO cake (name) VALUES ('pancake'); Query OK, 1 row affected, 1 warning (0.03 sec) MySQL ate my cake
  9. 9. MySQL ate my cake mysql> SELECT * FROM cake; +------+ | name | +------+ | pan | +------+ 1 row in set (0.03 sec) OMG! Where is my “cake”?
  10. 10. PostgreSQL? testdb=# CREATE TABLE cake (name VARCHAR(3)); CREATE TABLE testdb=# INSERT INTO cake (name) VALUES ('pancake'); ERROR: value too long for type character varying(3)
  11. 11. Ever Seen This? CREATE TABLE users ( id integer auto_increment, username varchar(40), password varchar(8) );
  12. 12. Data Conversion Errors - MySQL mysql> CREATE TABLE test (x INTEGER); Query OK, 0 rows affected (0.00 sec) mysql> INSERT INTO test (x) VALUES ('bad-number'); Query OK, 1 row affected, 1 warning (0.01 sec)
  13. 13. Data Conversion Errors - MySQL mysql> SELECT * FROM foo; +------+ | x | +------+ | 0 | +------+ 1 row in set (0.00 sec)
  14. 14. Data Conversion Errors - PostgreSQL testdb=# CREATE TABLE test (x integer); CREATE TABLE testdb=# INSERT INTO test (x) VALUES ('bad- number'); ERROR: invalid input syntax for integer: "bad-number" LINE 1: INSERT INTO foo (x) VALUES ('bad-number'); ^
  15. 15. Parallels - PHP vs. Python $ php -r '$x = "bad-number"; $y = (int)$x; echo $y."n";' 0
  16. 16. Parallels - PHP vs. Python $ python -c 'print int("bad-number")' Traceback (most recent call last): File "<string>", line 1, in <module> ValueError: invalid literal for int() with base 10: 'bad-number'
  17. 17. File Layout
  18. 18. File Layout - MySQL MyISAM /var/lib/mysql/dbname ● dbname.MYD - data of all tables ● dbname.MYI - indexes
  19. 19. File Layout - MySQL InnoDB /var/lib/mysql/ ● ibdata1 - data of all databases, including tables and indexes It is possible to tell mysql, by changing a config flag, to make it use one file for table.
  20. 20. /var/lib/postgresql/9.3/main/base ● 131384/ - directory per database ○ 2654 - one (or more) files for each table/index ○ 2703 ○ 2683 ○ 2683.1 ○ 2683.2 ○ ... File Layout - PostgreSQL
  21. 21. Database Maintenance
  22. 22. CREATE INDEX - MySQL MyISAM While CREATE INDEX is in progress: ● Entire table is locked for writes ● A new index file (dbname.MYI) need to created
  23. 23. CREATE INDEX - InnoDB ● Entire table rebuilt to create an index. ● Seems to have improved in recent versions
  24. 24. CREATE INDEX - PostgreSQL ● CREATE INDEX ○ locks the table for writes ● CREATE INDEX CONCURRENTLY ○ Doesn’t hold the lock for entire period ○ Slower than plain CREATE INDEX ● Each index is a new file
  25. 25. Takes long time as it needs to rewrite: ● the index file (dbname.MYI) for MyISAM ● the ibdata1 file for InnoDB DROP INDEX - MySQL
  26. 26. DROP INDEX - PostgreSQL ● Almost instantaneous ● Just need to delete the files corresponding to that index
  27. 27. ADDING NEW COLUMN - MySQL Entire table data needs to be rewritten.
  28. 28. ADDING NEW COLUMN - PostgreSQL Almost instantaneous if the new column has a default value.
  29. 29. Connection Model
  30. 30. Connection Model - MySQL A thread for each connection PROS ● Very easy to create a new conn CONS ● Difficult to scale on multi-core systems ● difficult monitor threads
  31. 31. Connection Model - PostgreSQL A process for each connection PROS ● better concurrency ● complete isolation ● plays nicely with UNIX tools (ps, top, kill) CONS ● lot of overhead for creating new conn
  32. 32. $ top
  33. 33. $ top kill 17618 kill -STOP 17618 kill -CONT 17618
  34. 34. Query Planning
  35. 35. EXPLAIN SELECT name, total FROM names WHERE year=1995 ORDER BY total DESC LIMIT 10; The Query
  36. 36. MySQL - Query Plan | id | select_type | table | type | possible_keys | +----+-------------+-------+------+---------------+ | 1 | SIMPLE | names | ALL | NULL | key |key_len|ref |rows |Extra | -----+-------+-----+--------+---------------+ NULL | 5 |const|12420176|Using where; | | | | |Using filesort | 1 row in set (0.00 sec)
  37. 37. MySQL - Add Index CREATE INDEX names_total_idx ON names(total)
  38. 38. MySQL - Query Plan With Index | id | select_type | table | type | possible_keys | +----+-------------+-------+------+---------------+ | 1 | SIMPLE | names | ref | NULL | key |key_len|ref |rows |Extra | ---------------+-------+-----+------+--------------+ names_total_idx| 5 |const|10 |Using where; | 1 row in set (0.00 sec)
  39. 39. Limit (cost=246987.39..246987.42 rows=10 width=13) -> Sort (cost=246987.39..247467.64 rows=192099 width=13) Sort Key: total -> Seq Scan on names (cost=0.00..242836.20 rows=192099 width=13) Filter: (year = 1995) (5 rows) PostgreSQL - QUERY PLAN
  40. 40. CREATE INDEX names_total_idx ON names(total) PostgreSQL - Add Index
  41. 41. Limit (cost=0.43..1891.80 rows=10 width=13) -> Index Scan Backward using names_total_idx on names (cost=0.43..36332875.67 rows=192099 width=13) Filter: (year = 1995) (3 rows) PostgreSQL - Query Plan With Index
  42. 42. PostgreSQL - Query Plan names=# EXPLAIN ANALYZE SELECT … Limit (cost=0.43..1891.80 rows=10 width=13) (actual time=0.037..0.462 rows=10 loops=1) -> Index Scan Backward using names_total_idx on names (cost=0.43..36332875.67 rows=192099 width=13) (actual time=0.036..0.458 rows=10 loops=1) Filter: (year = 1995) Rows Removed by Filter: 467 Total runtime: 0.517 ms (5 rows)
  43. 43. PostgreSQL - Complex Query EXPLAIN SELECT name, sum(total) as count FROM names WHERE year > 1980 GROUP BY name ORDER BY count
  44. 44. PostgreSQL - Query Plan Sort (cost=254889.31..255063.44 rows=69653 width=13) Sort Key: (sum(total)) -> HashAggregate(cost=248589.93..249286.46 rows=69653 width=13) -> Bitmap Heap Scan on names (cost=83212.02..226363.10 rows=4445366 width=13) Recheck Cond: (year > 1980) -> Bitmap Index Scan on names_year_idx (cost=0.00..82100.68 rows=4445366 width=0) Index Cond: (year > 1980) (7 rows)
  45. 45. Replication
  46. 46. MySQL Replication Replication Format ● Statement based ● Row based ● Mixed Mode ● binlog ● GTID
  47. 47. PostgreSQL Replication ● Synchronous / Asynchronous ● Streaming / Log shipping Master Slave 1 Log Shipping Streaming Slave 2 Slave 3
  48. 48. Data Recovery
  49. 49. PG - Point In Time Recovery ● Postgres maintains Write Ahead Log of all changes made to the database. ● The WAL files can be replayed up to any given timestamp. Time machine of your database!
  50. 50. Other Interesting Features of PostgreSQL
  51. 51. Partial Indexes SELECT * FROM comments WHERE email LIKE '%@spam.com'; CREATE INDEX comments_spam_idx ON comments WHERE email LIKE '%@spam.com';
  52. 52. Functional Indexes SELECT tag, count(*)FROM posts GROUP BY lower(category); CREATE INDEX post_category_idx ON post (lower(category));
  53. 53. JSON datatype CREATE TABLE book ( id serial primary key, data JSON ); INSERT INTO book (data) VALUES ( '{"title": "Tom Sawyer", "author": "Mark Twain"}')
  54. 54. JSON datatype SELECT * FROM book WHERE data->>'author' = 'Mark Twain' id| data --+----------------------------------------------- 1 |{"title": "Tom Sawyer", "author": "Mark Twain"} (1 row)
  55. 55. SELECT total_time/calls AS t, calls, query FROM pg_stat_statements ORDER BY t DESC 8606.75|3|select name, sum(total) as count | |from names group by name order by count limit ?; 4.92|8| select name, total from names | | where year=? order by total desc limit ?; pg_stat_statements
  56. 56. Summary PostgreSQL is better than MySQL in ● Data Consistency ● Query Planning ● Stability ● Database Maintenance ● Data Recovery Worth trying PostgreSQL for your next project!
  57. 57. Credits PostgreSQL Logo - By Jeff MacDonald http://pgfoundry.org/docman/?group_id=1000089
  58. 58. Thanks! Anand Chitipothu @anandology

×