Scribd
Upload
202 views

Postgre SQL

PostgreSQL uses a multi-process architecture with a postmaster process that forks child processes to handle client connections. It uses shared memory to share data between processes. The write-ahead log stored in pg_xlog is used for crash recovery. Various background processes perform vacuuming, writing data to disk, replication and statistics collection. The optimizer considers different scan and join methods and join order when planning queries using statistics collected on the data.

Uploaded by

SureshVitus
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
202 views

Postgre SQL

PostgreSQL uses a multi-process architecture with a postmaster process that forks child processes to handle client connections. It uses shared memory to share data between processes. The write-ahead log stored in pg_xlog is used for crash recovery. Various background processes perform vacuuming, writing data to disk, replication and statistics collection. The optimizer considers different scan and join methods and join order when planning queries using statistics collected on the data.

Uploaded by

SureshVitus
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 35

PostgreSQL Architecture

gnes Kovcs
Budapest, 2015-01-20

Agenda

Introduction
Overview of architecture
Process structure
Shared memory
Concurrency handling
The Optimizer

Introduction

What is PostgreSQL?
open source
object-relational database system
runs on Linux, UNIX (AIX, BSD, HP-UX, SGI IRIX, Mac OS X, Solaris, Tru64), and
Windows
ACID compliant
supported data types: INTEGER, NUMERIC, BOOLEAN, CHAR, VARCHAR, DATE,
INTERVAL, TIMESTAMP and binary large objects
native programming interfaces for C/C++, Java, .Net, Perl, Python, Ruby, Tcl, ODBC

PostgreSQL in numbers
Limit

Value

Maximum Database Size

Unlimited

Maximum Table Size

32 TB

Maximum Row Size

1.6 TB

Maximum Field Size

1 GB

Maximum Rows per Table

Unlimited

Maximum Columns per Table

250 - 1600 depending on column types

Maximum Indexes per Table

Unlimited

What PostgreSQL can do?

Multi-Version Concurrency Control (MVCC)


Point in time recovery
Tablespaces
Asynchronous replication
Nested transactions (savepoints)
Online/hot backups
Query planner/optimizer
Write Ahead Logging
International character sets, multibyte character encodings, Unicode, and it is localeaware for sorting, case-sensitivity, and formatting

Overview of architecture

Database file layout


PGDATA base directory for the Database Server: traditionally it contains
configuration and data files + data directory
example location: /var/lib/pgsql/data
Multiple clusters, managed by different server instances, can exist on the same
machine
configuration files and pid file location can be configured any where, it can reside
under PGDATA also

Subdirectories within PGDATA


Item
PG_VERSION
base
global
pg_clog
pg_dynshmem

pg_logical
pg_multixact
pg_notify
pg_replslot
pg_serial
pg_snapshots
pg_stat
pg_stat_tmp
pg_subtrans
pg_tblspc
pg_twophase
pg_xlog

Description
A file containing the major version number of PostgreSQL
Subdirectory containing per-database subdirectories
Subdirectory containing cluster-wide tables, such as pg_database
Subdirectory containing transaction commit status data
Subdirectory containing files used by the dynamic shared memory
subsystem
Subdirectory containing status data for logical decoding
Subdirectory containing multitransaction status data (used for shared
row locks)
Subdirectory containing LISTEN/NOTIFY status data
Subdirectory containing replication slot data
Subdirectory containing information about committed serializable
transactions
Subdirectory containing exported snapshots
Subdirectory containing permanent files for the statistics subsystem
Subdirectory containing temporary files for the statistics subsystem
Subdirectory containing subtransaction status data
Subdirectory containing symbolic links to tablespaces
Subdirectory containing state files for prepared transactions
Subdirectory containing WAL (Write Ahead Log) files

The /base subdirectory


contains the user database files
subdirectory names are the database OIDs

Data Pages
pages are located under the
database subdirectories
page default size: 8k
additional sizes:4k and 16k but
needs compilation of postgresql

for general purpose 8k is best


practice

Important user accessible files

PGVERSION: major version number of installation


postgresql.conf: main configuration file for PostgreSQL installation
pg_hba.conf: configures the client authentication method
pg_ident.conf: configures OS and PostgreSQL authentication name mapping
postmaster.opts: default command line options for the postmaster
postmaster.pid: PID of the postmaster and identification the main directory

Write Ahead Logs I.

located under /pg_xlog


REDO logs of PostgreSQL database Server
per default it is used during crash recovery
related main parameters are
wal_level
archive_mode, archive_command

Write Ahead Logs II.


wal_level: defines what to log into the WAL files
minimal (default): transaction logging is skiped for bulk operations like CREATE TABLE
AS, CREATE INDEX, CLUSTER, COPY etc.
enough to recover after a crash or immediate shutdown

archive: needed for archiving WAL files


hot_standby: enable to open standby read only
logical: supports logical decoding

archive_mode: default is false, if true WAL files are archived with the command given
at archive_command

Process structure

Postmaster
main PostgreSQL program
postgres@agideb:~$ pg_ctl status
pg_ctl: server is running (PID: 2971)
/usr/lib/postgresql/9.1/bin/postgres "-D"
"/var/lib/postgresql/9.1/main" "-c"
"config_file=/etc/postgresql/9.1/main/postgresql.conf"

postmaster is listening and if user


connection comes in it forks postgres server
processes
postgres server process is the copy of
postmaster

Additional important background


processes I.
Statistics Collector
Background Writer
writes dirty pages to disk
runs repeatedly (time is defined by multiple parameters and actual statistics)

WAL Writer

Additional important background


processes
II.
Auto Vacuum daemon
optional but highly recommended
automates VACUUM and ANALYZE commands
multiple processes:
a.
b.

auto vacuum launcher: persistent


auto vacuum workers: started by launcher for all databases

What is VACUUM?

recover or reuse disk space occupied by deleted or updated


rows
update data statistics
update visibility maps speeds up index only scans
protects against transactional ID Wraparound

Shared Memory

Why do we need shared memory in


PostgreSQL?
to mainly enable communication
between postmaster and postgres
server processes
child processes cannot propagate
information
shared memory is available to share
the same information to all processes

to cache pages

Structure of a block tuple

Structure of a single file system tuple

Handling concurrency

Multi-Version Concurrency Control


(MVCC)
Readers do not block writers, writers do not block readers
PostgreSQL guarentees this even with the strictest isolation level

Standard SQL Transaction Isolation Levels


available in PostgreSQL
Isolation Level

Dirty Read

Nonrepeatable Read

Phantom Read

Read uncommitted

Possible

Possible

Possible

Read committed

Not possible

Possible

Possible

Repeatable read

Not possible

Not possible

Possible

Serializable

Not possible

Not possible

Not possible

The Optimizer

Decisions taken by the optimizer


Scan Method
Sequential Scan
Bitmap Index Scan
Index Scan

Join Method
Nested Loop
Hash Join
Merge Join

Join Order

Optimizer statistics
distribution of data:

100 most common values


histograms with 100 buckets
granularity can be changed to have more data to calculate distribution
ALTER TABLE

statistic collection cannot be turned off


statistics cannot be backed up individually for an object

Sources
http://momjian.us/main/presentations/overview.html
http://www.postgresql.org/
http://raghavt.blogspot.hu/2011/04/postgresql-90-architecture.html

You might also like