Cached pages in TempDB (MB)

November 21, 2012 by Jonathan Allen | 0 comments | 961 views
Categories: Performance, Storage

Rating: 0.0/5 (0 votes cast)

TempDB is a key component in the general performance of a SQL Server. Knowing more about what is stored in it, and by which process, will greatly help diagnose performance issues. This metric helps a DBA analyze how well TempDB is working on their server. Seeing big changes in TempDB commitment between one database and another could be a sign of a wrongly sized TempDB, or indicate that one database is using TempDB in a sub-optimal way.

Metric definition

Metric Name:

Cached pages in TempDB (MB)

Description:

The amount of space (in MB) consumed by user and internal processes, in TempDB per database. Big changes in allocations (increases or decreases) of space for particular databases can indicate possible problems, such as poorly written T-SQL. Keep a watch on big allocations that are not subsequently de-allocated appropriately. This trace shows the net effect, so increases not followed by decreases show a de-allocation is not taking place, possibly because of a long running transaction using a temporary table. Consider this trace along with the Log size and Log used traces for TempDB and use this information to ensure your TempDB files are sized appropriately.

Enter the T-SQL query that will collect data:

WITH    TempUsage
          AS ( SELECT [ddtsu].[session_id] ,
                    [ddtsu].[user_objects_alloc_page_count] ,
                    [ddtsu].[user_objects_dealloc_page_count] ,
                    [ddtsu].[internal_objects_alloc_page_count] ,
                    [ddtsu].[internal_objects_dealloc_page_count]
                FROM [sys].[dm_db_task_space_usage] AS ddtsu
               UNION ALL
               SELECT [ddssu].[session_id] ,
                    [ddssu].[user_objects_alloc_page_count] ,
                    [ddssu].[user_objects_dealloc_page_count] ,
                    [ddssu].[internal_objects_alloc_page_count] ,
                    [ddssu].[internal_objects_dealloc_page_count]
                FROM [sys].[dm_db_session_space_usage] AS ddssu
             )
             
    SELECT SUM(( ( tu.[user_objects_alloc_page_count]			-	tu.[user_objects_dealloc_page_count] )
                 + ( tu.[internal_objects_alloc_page_count]		-	tu.[internal_objects_dealloc_page_count] ) ))
            / 128.0 AS [TempDB_Objects_MB]
        FROM [TempUsage] AS tu 
            INNER JOIN [sys].[sysprocesses] AS s
                ON tu.[session_id] = [s].[spid]
        WHERE [s].[spid] > 50
            AND [s].[dbid] = DB_ID();

Select instances to collect from:

Select all

Choose databases to collect from:

Specify databases and type

TempDB

Set collection frequency:

Collect data every 10 minutes

Note: If you suspect that there may be an issue with TempDB, increase the collection frequency to every minute so you can analyze more data.

Use collected or calculated values:

Leave the Use a calculated rate of change between collections check box turned off

Metric collection is:

Enabled

Metric definition

Alert definition