Applies to: SQL Server (all supported versions) Azure SQL Database Azure SQL Managed Instance
This article describes the
tempdb system database, a global resource available to all users connected to an instance of SQL Server, Azure SQL Database, or Azure SQL Managed Instance.
tempdb system database is a global resource that holds:
Temporary user objects that are explicitly created. They include global or local temporary tables and indexes, temporary stored procedures, table variables, tables returned in table-valued functions, and cursors.
Internal objects that the database engine creates. They include:
- Work tables to store intermediate results for spools, cursors, sorts, and temporary large object (LOB) storage.
- Work files for hash join or hash aggregate operations.
- Intermediate sort results for operations such as creating or rebuilding indexes (if
SORT_IN_TEMPDBis specified), or certain
ORDER BY, or
Each internal object uses a minimum of nine pages: an IAM page and an eight-page extent. For more information about pages and extents, see Pages and extents.
Version stores, which are collections of data pages that hold the data rows that support features for row versioning. There are two types: a common version store and an online-index-build version store. The version stores contain:
- Row versions that are generated by data modification transactions in a database that uses
READ COMMITTEDthrough row versioning isolation or snapshot isolation transactions.
- Row versions that are generated by data modification transactions for features, such as online index operations, Multiple Active Result Sets (MARS), and
- Row versions that are generated by data modification transactions in a database that uses
tempdb are minimally logged so that transactions can be rolled back.
tempdb is re-created every time SQL Server is started so that the system always starts with a clean copy of the database. Temporary tables and stored procedures are dropped automatically on disconnect, and no connections are active when the system is shut down.
tempdb never has anything to be saved from one session of SQL Server to another. Backup and restore operations are not allowed on
Physical properties of tempdb in SQL Server
The following table lists the initial configuration values of the
tempdb data and log files in SQL Server. The values are based on the defaults for the
model database. The sizes of these files might vary slightly for different editions of SQL Server.
|File||Logical name||Physical name||Initial size||File growth|
|Primary data||tempdev||tempdb.mdf||8 megabytes||Autogrow by 64 MB until the disk is full|
|Secondary data files||temp#||tempdb_mssql_#.ndf||8 megabytes||Autogrow by 64 MB until the disk is full|
|Log||templog||templog.ldf||8 megabytes||Autogrow by 64 megabytes to a maximum of 2 terabytes|
The number of secondary data files depends on the number of (logical) processors on the machine. As a general rule, if the number of logical processors is less than or equal to eight, use the same number of data files as logical processors. If the number of logical processors is greater than eight, use eight data files. Then if contention continues, increase the number of data files by multiples of four until the contention decreases to acceptable levels, or make changes to the workload/code.
The default value for the number of data files is based on the general guidelines in KB 2154845.
To check current size and growth parameters for
tempdb, query view
Moving the tempdb data and log files in SQL Server
To move the
tempdb data and log files, see Move system databases.
Database options for tempdb in SQL Server
The following table lists the default value for each database option in the
tempdb database and whether the option can be modified. To view the current settings for these options, use the sys.databases catalog view.
|Database option||Default value||Can be modified|
|Database Availability Options||ONLINE
|PAGE_VERIFY||CHECKSUM for new installations of SQL Server
NONE for upgrades of SQL Server
|Service Broker Options||ENABLE_BROKER||Yes|
For a description of these database options, see ALTER DATABASE SET Options (Transact-SQL).
tempdb in Azure SQL
The behavior of tempdb in Azure SQL Database differs from the behavior SQL Server, Azure SQL Managed Instance, and SQL Server on Azure VMs.
tempdb in SQL Database
Single and pooled databases in Azure SQL Database support global temporary tables and global temporary stored procedures scoped to the database level, and stored in
tempdb. Global temporary tables and global temporary stored procedures are shared for all users' sessions within the same database. User sessions from other databases can't access global temporary tables. For more information, see Database scoped global temporary tables (Azure SQL Database).
For single and pooled databases in Azure SQL Database, out of all the system databases, only the master database and
tempdb database are accessible. For more information, see What is a logical server in Azure?
To learn more about tempdb sizes in Azure SQL Database, review:
- vCore purchasing model: single databases, pooled databases
- DTU purchasing model: single databases, pooled databases
tempdb in SQL Managed Instance
Azure SQL Managed Instance supports temporary objects in the same way as SQL Server, where all global temporary tables and global temporary stored procedures are accessible by all user sessions within the same managed instance. Likewise, all system databases are accessible.
To learn more about tempdb sizes in Azure SQL Managed Instance, review resource limits.
The following operations can't be performed on the
- Adding filegroups.
- Backing up or restoring the database.
- Changing collation. The default collation is the server collation.
- Changing the database owner.
tempdbis owned by sa.
- Creating a database snapshot.
- Dropping the database.
- Dropping the guest user from the database.
- Enabling Change Data Capture.
- Participating in database mirroring.
- Removing the primary filegroup, primary data file, or log file.
- Renaming the database or primary filegroup.
- Setting the database to
- Setting the database or primary filegroup to
Any user can create temporary objects in
tempdb. Users can access only their own objects, unless they receive additional permissions. It's possible to revoke the connect permission to
tempdb to prevent a user from using
tempdb. We don't recommend it because some routine operations require the use of
Optimizing tempdb performance in SQL Server
The size and physical placement of the
tempdb database can affect the performance of a system. For example, if the size that's defined for
tempdb is too small, part of the system-processing load might be taken up with autogrowing
tempdb to the size required to support the workload every time you restart the instance of SQL Server.
If possible, use instant file initialization to improve the performance of growth operations for data files.
Preallocate space for all
tempdb files by setting the file size to a value large enough to accommodate the typical workload in the environment. Preallocation prevents
tempdb from expanding too often, which affects performance. The
tempdb database should be set to autogrow to increase disk space for unplanned exceptions.
Data files should be of equal size within each filegroup, because SQL Server uses a proportional-fill algorithm that favors allocations in files with more free space. Dividing
tempdb into multiple data files of equal size provides a high degree of parallel efficiency in operations that use
Set the file growth increment to a reasonable size to prevent the
tempdb database files from growing by too small a value. If the file growth is too small compared to the amount of data that's being written to
tempdb might have to constantly expand. That will affect performance.
To check current size and growth parameters for
tempdb, use the following query:
SELECT name AS FileName, size*1.0/128 AS FileSizeInMB, CASE max_size WHEN 0 THEN 'Autogrowth is off.' WHEN -1 THEN 'Autogrowth is on.' ELSE 'Log file grows to a maximum size of 2 TB.' END, growth AS 'GrowthValue', 'GrowthIncrement' = CASE WHEN growth = 0 THEN 'Size is fixed.' WHEN growth > 0 AND is_percent_growth = 0 THEN 'Growth value is in 8-KB pages.' ELSE 'Growth value is a percentage.' END FROM tempdb.sys.database_files; GO
tempdb database on a fast I/O subsystem. Use disk striping if there are many directly attached disks. Individual or groups of
tempdb data files don't necessarily need to be on different disks or spindles unless you're also encountering I/O bottlenecks.
tempdb database on disks that differ from the disks that user databases use.
Performance improvements in tempdb for SQL Server
Starting with SQL Server 2016 (13.x),
tempdb performance is further optimized in the following ways:
- Temporary tables and table variables are cached. Caching allows operations that drop and create the temporary objects to run very quickly. Caching also reduces page allocation and metadata contention.
- The allocation page latching protocol is improved to reduce the number of
UP(update) latches that are used.
- Logging overhead for
tempdbis reduced to reduce disk I/O bandwidth consumption on the
- Setup adds multiple
tempdbdata files during a new instance installation. You can accomplish this task by using the new UI input control in the Database Engine Configuration section and the command-line parameter
/SQLTEMPDBFILECOUNT. By default, setup adds as many
tempdbdata files as the logical processor count or eight, whichever is lower.
- When there are multiple
tempdbdata files, all files autogrow at the same time and by the same amount, depending on growth settings. Trace flag 1117 is no longer required.
- All allocations in
tempdbuse uniform extents. Trace flag 1118 is no longer required.
- For the primary filegroup, the
AUTOGROW_ALL_FILESproperty is turned on and the property can't be modified.
For more information on performance improvements in
tempdb, see the blog article TEMPDB - Files and Trace Flags and Updates, Oh My!.
Memory-optimized tempdb metadata
Metadata contention in
tempdb has historically been a bottleneck to scalability for many workloads running on SQL Server. SQL Server 2019 (15.x) introduces a new feature that's part of the in-memory database feature family: memory-optimized tempdb metadata.
This feature effectively removes this bottleneck and unlocks a new level of scalability for tempdb-heavy workloads. In SQL Server 2019 (15.x), the system tables involved in managing temporary table metadata can be moved into latch-free, non-durable, memory-optimized tables.
Currently the memory-optimized tempdb metadata feature is not available in Azure SQL Database or Azure SQL Managed Instance.
Watch this seven-minute video for an overview of how and when to use memory-optimized tempdb metadata:
Configuring and using memory-optimized tempdb metadata
To opt in to this new feature, use the following script:
ALTER SERVER CONFIGURATION SET MEMORY_OPTIMIZED TEMPDB_METADATA = ON;
This configuration change requires a restart of the service to take effect.
You can verify whether or not
tempdb is memory-optimized by using the following T-SQL command:
If the server fails to start for any reason after you enable memory-optimized
tempdb metadata, you can bypass the feature by starting the SQL Server instance with minimal configuration through the -f startup option. You can then disable the feature and restart SQL Server in normal mode.
To protect the server from potential out-of-memory conditions, you can bind
tempdb to a resource pool. This is done through the
ALTER SERVER command rather than the steps you would normally follow to bind a resource pool to a database.
ALTER SERVER CONFIGURATION SET MEMORY_OPTIMIZED TEMPDB_METADATA = ON (RESOURCE_POOL = 'pool_name');
This change also requires a restart to take effect, even if memory-optimized tempdb metadata is already enabled.
Memory-optimized tempdb limitations
Toggling the feature on and off is not dynamic. Because of the intrinsic changes that need to be made to the structure of
tempdb, a restart is required to either enable or disable the feature.
A single transaction is not allowed to access memory-optimized tables in more than one database. Any transactions that involve a memory-optimized table in a user database won't be able to access
tempdbsystem views in the same transaction. If you try to access
tempdbsystem views in the same transaction as a memory-optimized table in a user database, you'll receive the following error:
A user transaction that accesses memory optimized tables or natively compiled modules cannot access more than one user database or databases model and msdb, and it cannot write to master.
BEGIN TRAN; SELECT * FROM tempdb.sys.tables; -----> Creates a user in-memory OLTP transaction in tempdb INSERT INTO <user database>.<schema>.<mem-optimized table> VALUES (1); ----> Tries to create a user in-memory OLTP transaction in the user database but will fail COMMIT TRAN;
Queries against memory-optimized tables don't support locking and isolation hints, so queries against memory-optimized
tempdbcatalog views won't honor locking and isolation hints. As with other system catalog views in SQL Server, all transactions against system views will be in
READ COMMITTED(or in this case,
READ COMMITTED SNAPSHOT) isolation.
Columnstore indexes can't be created on temporary tables when memory-optimized
tempdbmetadata is enabled.
Due to the limitation on columnstore indexes, use of the
sp_estimate_data_compression_savingssystem stored procedure with the
COLUMNSTORE_ARCHIVEdata compression parameter is not supported when memory-optimized
tempdbmetadata is enabled.
These limitations apply only when you're referencing
tempdb system views. You can create a temporary table in the same transaction as you access a memory-optimized table in a user database, if desired.
Capacity planning for tempdb in SQL Server
Determining the appropriate size for
tempdb in a SQL Server production environment depends on many factors. As described earlier, these factors include the existing workload and the SQL Server features that are used. We recommend that you analyze the existing workload by performing the following tasks in a SQL Server test environment:
- Set autogrow on for
- Run individual queries or workload trace files and monitor
- Execute index maintenance operations such as rebuilding indexes, and monitor
- Use the space-use values from the previous steps to predict your total workload usage. Adjust this value for projected concurrent activity, and then set the size of
Monitoring tempdb use
Running out of disk space in
tempdb can cause significant disruptions in the SQL Server production environment. It can also prevent applications that are running from completing operations. You can use the sys.dm_db_file_space_usage dynamic management view to monitor the disk space that's used in the
-- Determining the amount of free space in tempdb SELECT SUM(unallocated_extent_page_count) AS [free pages], (SUM(unallocated_extent_page_count)*1.0/128) AS [free space in MB] FROM tempdb.sys.dm_db_file_space_usage; -- Determining the amount of space used by the version store SELECT SUM(version_store_reserved_page_count) AS [version store pages used], (SUM(version_store_reserved_page_count)*1.0/128) AS [version store space in MB] FROM tempdb.sys.dm_db_file_space_usage; -- Determining the amount of space used by internal objects SELECT SUM(internal_object_reserved_page_count) AS [internal object pages used], (SUM(internal_object_reserved_page_count)*1.0/128) AS [internal object space in MB] FROM tempdb.sys.dm_db_file_space_usage; -- Determining the amount of space used by user objects SELECT SUM(user_object_reserved_page_count) AS [user object pages used], (SUM(user_object_reserved_page_count)*1.0/128) AS [user object space in MB] FROM tempdb.sys.dm_db_file_space_usage;
To monitor the page allocation or deallocation activity in
tempdb at the session or task level, you can use the sys.dm_db_session_space_usage and sys.dm_db_task_space_usage dynamic management views. These views can help you identify large queries, temporary tables, or table variables that are using lots of
tempdb disk space. You can also use several counters to monitor the free space that's available in
tempdb and the resources that are using
-- Obtaining the space consumed by internal objects in all currently running tasks in each session SELECT session_id, SUM(internal_objects_alloc_page_count) AS task_internal_objects_alloc_page_count, SUM(internal_objects_dealloc_page_count) AS task_internal_objects_dealloc_page_count FROM sys.dm_db_task_space_usage GROUP BY session_id; -- Obtaining the space consumed by internal objects in the current session for both running and completed tasks SELECT R2.session_id, R1.internal_objects_alloc_page_count + SUM(R2.internal_objects_alloc_page_count) AS session_internal_objects_alloc_page_count, R1.internal_objects_dealloc_page_count + SUM(R2.internal_objects_dealloc_page_count) AS session_internal_objects_dealloc_page_count FROM sys.dm_db_session_space_usage AS R1 INNER JOIN sys.dm_db_task_space_usage AS R2 ON R1.session_id = R2.session_id GROUP BY R2.session_id, R1.internal_objects_alloc_page_count, R1.internal_objects_dealloc_page_count;
Submit and view feedback for