Oracle
Content
Oracle Database Administration I Oracle Server Architecture
Overview
Set of data to store and access information Based on a relational model of rows and columns stored in tables Also Object-oriented Object(OO) structures Contains these components: Physical Memory Process
Oracle's Overall Role
Retrieve data File I/O Space Management Change Management Access Management
Database Structure
Data file(s) Control file(s) Redo logs Init.ora (spfile.ora - 9i) Trace Alert log Password
Instance Structure
Memory structures are allocated Collectively known as System Global Area. 95% defined by data block buffer cache, Shared SQL Pool, Large Pool and Java Pool. Oracle background processes are started. Single database accessible by multiple instances - RAC (Real Application Cluster) a.k.a. parallel server
Instance Defined
Defined by initialization parameter(s) that determine the size and composition Can be dynamically altered initSID.ora (ascii file) spfileSID.ora (binary file) configSID.ora (usually only used to create a database ± not widely used)
Oracle Instance
Composed of: SGA (System Global Area) PGA (Program Global Area) Oracle BACKGROUND processes
System Global Area (SGA)
Shared memory region allocated by Oracle for an Oracle database Allocated when the Oracle database is started The SGA should be in non-paged, nonswapped memory
SGA Composition
Buffer cache Shared SQL Pool Large pool (optional) Java pool Redo log buffer Dictionary cache Other miscellaneous items
PGA Composition
Non-shared memory area to which a process (server or background) can write Allocated by Oracle when a user connects to an Oracle database and a session is created
Oracle Instance
Required Background Processes: DBWn LGWR SMON PMON CKPT (9i)
Oracle Instance
Optional Background Processes: (not all inclusive)
ARCn CKPT (prior to 9i) RECO Lock (LCK0) Job Queue (SNPn) Queue Monitor (QMNn) Dispatcher (Dnnn) Server (Snnn ± for dedicated MTS servers)
Oracle Instance More Details
SGA
Buffer Cache Redo Log BUFFER Shared SQL Pool Dynamic (as of Oracle 9i)
Dynamic SGA
Defined in Granules SGA < 128mb 4mb size SGA > 128mb 16mb size Can be modified on the fly (9i) Establish an overall maximum Initially allocated ± 4 for small, 3 for large (fixed SGA, buffer cache, Shared pool)
Data Buffer Cache Structure
Can be defined in two ways: * By block db_block_buffers ** By byte db_cache_size Was divided into blocks the same size as the blocks in the data files As of 9i, multiple buffer cache introduced (depends on block size of each tablespace) Size and number defined in the initialization Parameter File Holds copies of data blocks read from disk
Data Buffer Cache Contents
Stores the most recently accessed BLOCKS OF DATA Contains PINNED buffers * Data being accessed Contains DIRTY buffers * Data that has been MODIFIED, but not written to disk FREE buffers Two lists, write and LRU
REDO Log Buffers
Stores the REDO ENTRIES (changes to the database) Default is four times the OS block size Larger reduces log I/O and uncommented redo to online redo logs Changed data is moved from the database buffer cache to the redo log buffer then to the redo log files
Shared Pool
Can be dynamically modified (9i) Must be an integer multiple of the granule size * Oracle will round up if not Managed by LRU algorithm Contains: * Library Cache * Data Dictionary Cache * Control structures * Reusable Runtime Memory
Shared Pool
Shared Pool
Library Cache
* SHARED by all users for commonly used SQL statements * Shared and Private SQL areas for executed statements * PL/SQL Procedures and Packages * Allocation and Reuse of Memory
Library Cache
Shared SQL * PARSE TREES for SQL statements * EXECUTION plans for SQL statements * Size dependent on complexity Private SQL * Each session has one * Persistent and Runtime areas * Cursors
Library Cache
PL/SQL Program Units * Procedures, functions, Packages, anonymous blocks and database triggers Allocation and Reuse of Memory * Every SQL statement submitted, Oracle automatically allocates memory * Checks for existence * Allocate private SQL area * May flush the shared pool
PGA
PGA
Varies depending if dedicated or Shared servers (MTS) are used Stack space Session information (PGA verses Shared Pool) Memory allocated to hold a session¶s variables and arrays Automatically managed in 9i with the SQL memory management
Questions
What is the name of Oracle¶s memory structure? What are its contents? How is it divided? If I have 150 mb SGA, why does Oracle change the size? Can you vary sizes of Oracle blocks? If so, how?
Oracle Database
Composed of two structures: Physical (Operating System view) Logical (Oracle¶s view)
Physical View Files
Physical Structure
Made up of: Data files Redo Log Files ± two or more with members One or more Control Files One Parameter File Trace and alert log files One Password file
Datafile
Disk space for a tablespace One or more per tablespace Associated with only one tablespace Consists of segments: * UNDO, Temporary, Data, index, undo, LOB and cache (for the most part)
Redo Log Files
Characteristics Record changes made to the data by various transactions Help RECOVER the data in case of a failure Streamline the process of how data is written to disk
Control Files
Record of the overall architecture and status of the database Easily moved and recovered Contain the following: Database NAME Datafiles NAMES and LOCATIONS Redo log file NAMES, Log sequence number and LOCATIONS
Parameter File
Attributes of an instance are defined through an Oracle Parameter File Determines the size and characteristics of the instance Stored in an ASCII file or housed with the database in a binary file
Parameter File
The only Oracle file that is ASCII All other files are binary and not directly changeable by the DBA Read-only when the instance starts Most changes require a stop and restart of the instance
Trace and Alert Logs
Trace files contain information about significant events Alert log records the commands and results of major events in the life of the database Monitor daily Rename for historical purposes
Logical View Structures
Logical View Structures
Tablespace(s) Schema objects Segments Extents Data blocks
Tablespace(s)
Logically divides a database One or more datafiles are explicity created for each one Logically divided into segments that are further divided into extents and blocks Can be defined as read-only A unit of space to store a schema¶s objects
Object(s)
Schema(s) A collection of objects owned by one Oracle user Object(s) Tables, indexes, views, sequences, stored procedures, functions, packages, UNDO segments
Extents and Blocks
A collection of the smallest object in a database An extent can be of a varying size Extents MUST be contiguous Blocks are the finest collection of space in an Oracle database
Questions
What are the physical files of a database system? What are the logical entities of a database? What is the smallest structure? Which structure MUST be contiguous? Which structure will tell you event information?
Background Processes Review
Background Processes
Every instance has BACKGROUND PROCESSES BACKGROUND PROCESSES perform I/O, handle process cleanup and monitor the ORACLE database Provide support, increase performance, assist in recovery and make the database more reliable
Background Processes
DBW0 (DBWn) LGWR PMON SMON CKPT ARCn
Database Buffer
After the changed data (redo entries) are written to the online redo log files, the changed data will be written to disk These changed data are called µdirty¶ Dirty data are moved to the µwrite¶ list Dirty data can be on the µwrite¶ list and on the LRU list
DBWn
At some time after the redo entries have been written into the online redo logs, the database writer (DBWn) writes the dirty buffer back to the database This is called µdelayed write¶ and is based on initialization parameters
DBWn (Database Writer)
Can have one or more running Only available on systems with Asynch I/O Can use I/O slaves to duplicate the above on systems without it Performs all WRITES to the database Keeps the BUFFER CACHE clean and free
When does the DBWR write?
The dirty list reaches a threshold length A DBWR timeout LGWR issues a checkpoint
LGWR (Log Writer)
Writes the REDO log entries (located in the REDO LOG BUFFER) to the REDO LOG files Only the changed data, not the entire Oracle block(s) is written to the redo logs Issues checkpoints
Online Redo Log Files
The online redo log files hold the µredo entries¶ Only the changed bytes are written to the online redo log file The entire data block is not written to these logs
Online Redo Log Files
There is a fixed number of online redo log file groups Defined at database creation time Can add or drop groups or members (discussed later)
Online Redo Log Files
Redo log file groups cycle If there are three redo log file groups:
When the first group is full, LGWR starts writing tothe second group When the second group is full, LGWR writes to the third group When the third group is full, LGWR writes to the first group
LGWR (Log Writer)
What causes the LGWR to clear the REDO LOG BUFFER and transfer the transactions to disk? When a user process COMMITS a TRANSACTION TIME-OUT (caused every three seconds) When the redo log buffer becomes 1/3 FULL When the DBWR writes the dirty buffers to disk
Overview
Set of data to store and access information Based on a relational model of rows and columns stored in tables Also Object-oriented Object(OO) structures Contains these components: Physical Memory Process
Oracle's Overall Role
Retrieve data File I/O Space Management Change Management Access Management
Database Structure
Data file(s) Control file(s) Redo logs Init.ora (spfile.ora - 9i) Trace Alert log Password
Instance Structure
Memory structures are allocated Collectively known as System Global Area. 95% defined by data block buffer cache, Shared SQL Pool, Large Pool and Java Pool. Oracle background processes are started. Single database accessible by multiple instances - RAC (Real Application Cluster) a.k.a. parallel server
Instance Defined
Defined by initialization parameter(s) that determine the size and composition Can be dynamically altered initSID.ora (ascii file) spfileSID.ora (binary file) configSID.ora (usually only used to create a database ± not widely used)
Oracle Instance
Composed of: SGA (System Global Area) PGA (Program Global Area) Oracle BACKGROUND processes
System Global Area (SGA)
Shared memory region allocated by Oracle for an Oracle database Allocated when the Oracle database is started The SGA should be in non-paged, nonswapped memory
SGA Composition
Buffer cache Shared SQL Pool Large pool (optional) Java pool Redo log buffer Dictionary cache Other miscellaneous items
PGA Composition
Non-shared memory area to which a process (server or background) can write Allocated by Oracle when a user connects to an Oracle database and a session is created
Oracle Instance
Required Background Processes: DBWn LGWR SMON PMON CKPT (9i)
Oracle Instance
Optional Background Processes: (not all inclusive)
ARCn CKPT (prior to 9i) RECO Lock (LCK0) Job Queue (SNPn) Queue Monitor (QMNn) Dispatcher (Dnnn) Server (Snnn ± for dedicated MTS servers)
Oracle Instance More Details
SGA
Buffer Cache Redo Log BUFFER Shared SQL Pool Dynamic (as of Oracle 9i)
Dynamic SGA
Defined in Granules SGA < 128mb 4mb size SGA > 128mb 16mb size Can be modified on the fly (9i) Establish an overall maximum Initially allocated ± 4 for small, 3 for large (fixed SGA, buffer cache, Shared pool)
Data Buffer Cache Structure
Can be defined in two ways: * By block db_block_buffers ** By byte db_cache_size Was divided into blocks the same size as the blocks in the data files As of 9i, multiple buffer cache introduced (depends on block size of each tablespace) Size and number defined in the initialization Parameter File Holds copies of data blocks read from disk
Data Buffer Cache Contents
Stores the most recently accessed BLOCKS OF DATA Contains PINNED buffers * Data being accessed Contains DIRTY buffers * Data that has been MODIFIED, but not written to disk FREE buffers Two lists, write and LRU
REDO Log Buffers
Stores the REDO ENTRIES (changes to the database) Default is four times the OS block size Larger reduces log I/O and uncommented redo to online redo logs Changed data is moved from the database buffer cache to the redo log buffer then to the redo log files
Shared Pool
Can be dynamically modified (9i) Must be an integer multiple of the granule size * Oracle will round up if not Managed by LRU algorithm Contains: * Library Cache * Data Dictionary Cache * Control structures * Reusable Runtime Memory
Shared Pool
Shared Pool
Library Cache
* SHARED by all users for commonly used SQL statements * Shared and Private SQL areas for executed statements * PL/SQL Procedures and Packages * Allocation and Reuse of Memory
Library Cache
Shared SQL * PARSE TREES for SQL statements * EXECUTION plans for SQL statements * Size dependent on complexity Private SQL * Each session has one * Persistent and Runtime areas * Cursors
Library Cache
PL/SQL Program Units * Procedures, functions, Packages, anonymous blocks and database triggers Allocation and Reuse of Memory * Every SQL statement submitted, Oracle automatically allocates memory * Checks for existence * Allocate private SQL area * May flush the shared pool
PGA
PGA
Varies depending if dedicated or Shared servers (MTS) are used Stack space Session information (PGA verses Shared Pool) Memory allocated to hold a session¶s variables and arrays Automatically managed in 9i with the SQL memory management
Questions
What is the name of Oracle¶s memory structure? What are its contents? How is it divided? If I have 150 mb SGA, why does Oracle change the size? Can you vary sizes of Oracle blocks? If so, how?
Oracle Database
Composed of two structures: Physical (Operating System view) Logical (Oracle¶s view)
Physical View Files
Physical Structure
Made up of: Data files Redo Log Files ± two or more with members One or more Control Files One Parameter File Trace and alert log files One Password file
Datafile
Disk space for a tablespace One or more per tablespace Associated with only one tablespace Consists of segments: * UNDO, Temporary, Data, index, undo, LOB and cache (for the most part)
Redo Log Files
Characteristics Record changes made to the data by various transactions Help RECOVER the data in case of a failure Streamline the process of how data is written to disk
Control Files
Record of the overall architecture and status of the database Easily moved and recovered Contain the following: Database NAME Datafiles NAMES and LOCATIONS Redo log file NAMES, Log sequence number and LOCATIONS
Parameter File
Attributes of an instance are defined through an Oracle Parameter File Determines the size and characteristics of the instance Stored in an ASCII file or housed with the database in a binary file
Parameter File
The only Oracle file that is ASCII All other files are binary and not directly changeable by the DBA Read-only when the instance starts Most changes require a stop and restart of the instance
Trace and Alert Logs
Trace files contain information about significant events Alert log records the commands and results of major events in the life of the database Monitor daily Rename for historical purposes
Logical View Structures
Logical View Structures
Tablespace(s) Schema objects Segments Extents Data blocks
Tablespace(s)
Logically divides a database One or more datafiles are explicity created for each one Logically divided into segments that are further divided into extents and blocks Can be defined as read-only A unit of space to store a schema¶s objects
Object(s)
Schema(s) A collection of objects owned by one Oracle user Object(s) Tables, indexes, views, sequences, stored procedures, functions, packages, UNDO segments
Extents and Blocks
A collection of the smallest object in a database An extent can be of a varying size Extents MUST be contiguous Blocks are the finest collection of space in an Oracle database
Questions
What are the physical files of a database system? What are the logical entities of a database? What is the smallest structure? Which structure MUST be contiguous? Which structure will tell you event information?
Background Processes Review
Background Processes
Every instance has BACKGROUND PROCESSES BACKGROUND PROCESSES perform I/O, handle process cleanup and monitor the ORACLE database Provide support, increase performance, assist in recovery and make the database more reliable
Background Processes
DBW0 (DBWn) LGWR PMON SMON CKPT ARCn
Database Buffer
After the changed data (redo entries) are written to the online redo log files, the changed data will be written to disk These changed data are called µdirty¶ Dirty data are moved to the µwrite¶ list Dirty data can be on the µwrite¶ list and on the LRU list
DBWn
At some time after the redo entries have been written into the online redo logs, the database writer (DBWn) writes the dirty buffer back to the database This is called µdelayed write¶ and is based on initialization parameters
DBWn (Database Writer)
Can have one or more running Only available on systems with Asynch I/O Can use I/O slaves to duplicate the above on systems without it Performs all WRITES to the database Keeps the BUFFER CACHE clean and free
When does the DBWR write?
The dirty list reaches a threshold length A DBWR timeout LGWR issues a checkpoint
LGWR (Log Writer)
Writes the REDO log entries (located in the REDO LOG BUFFER) to the REDO LOG files Only the changed data, not the entire Oracle block(s) is written to the redo logs Issues checkpoints
Online Redo Log Files
The online redo log files hold the µredo entries¶ Only the changed bytes are written to the online redo log file The entire data block is not written to these logs
Online Redo Log Files
There is a fixed number of online redo log file groups Defined at database creation time Can add or drop groups or members (discussed later)
Online Redo Log Files
Redo log file groups cycle If there are three redo log file groups:
When the first group is full, LGWR starts writing tothe second group When the second group is full, LGWR writes to the third group When the third group is full, LGWR writes to the first group
LGWR (Log Writer)
What causes the LGWR to clear the REDO LOG BUFFER and transfer the transactions to disk? When a user process COMMITS a TRANSACTION TIME-OUT (caused every three seconds) When the redo log buffer becomes 1/3 FULL When the DBWR writes the dirty buffers to disk
