SQL Server 2014 Monitoring and Profiling
0 likes122 views
The document discusses various techniques for monitoring and profiling Microsoft SQL Server to enhance performance, including execution plans, indexing strategies, and troubleshooting locked columns. It emphasizes the importance of performance metrics like transaction rates and buffer cache hit ratios, and suggests tools such as sp_blitzcache for profiling queries. Additionally, it offers insights into optimizing indexing and diagnosing issues with stored procedures for better database management.
![Investigating index usage (index_usage_stats)
• SELECT OBJECT_NAME(ddius.[object_id], ddius.database_id) AS [object_name] ,
ddius.index_id ,
ddius.user_seeks ,
ddius.user_scans ,
ddius.user_lookups ,
ddius.user_seeks + ddius.user_scans + ddius.user_lookups
AS user_reads ,
ddius.user_updates AS user_writes ,
ddius.last_user_scan ,
ddius.last_user_update
FROM sys.dm_db_index_usage_stats ddius
WHERE ddius.database_id > 4 -- filter out system tables
AND OBJECTPROPERTY(ddius.OBJECT_ID, 'IsUserTable') = 1
AND ddius.index_id > 0 -- filter out heaps
ORDER BY ddius.user_scans DESC](https://image.slidesharecdn.com/mssql2014-administration-161115125825/85/SQL-Server-2014-Monitoring-and-Profiling-16-320.jpg)
![Identify indexes that have never been
accessed
• SELECT OBJECT_NAME(i.[object_id]) AS [Table Name] , i.name FROM
sys.indexes AS I INNER JOIN sys.objects AS o ON i.[object_id] =
o.[object_id] WHERE i.index_id NOT IN ( SELECT ddius.index_id
FROM sys.dm_db_index_usage_stats AS ddius
WHERE ddius.[object_id] = i.[object_id] AND i.index_id =
ddius.index_id AND database_id = DB_ID() ) AND o.[type] = 'U‘
ORDER BY OBJECT_NAME(i.[object_id]) ASC](https://image.slidesharecdn.com/mssql2014-administration-161115125825/85/SQL-Server-2014-Monitoring-and-Profiling-17-320.jpg)
SQL Server 2014 Monitoring and Profiling
- 1. Microsoft SQL Server Monitoring and Profiling Alireza Sadeghipour, Abouzar Noori 24th Azar 1394
- 2. Microsoft SQL Server: Monitoring and Profiling • Performance Monitoring • Estimated Execution Plan • Indexing • Database Engine Tuning Advisor • SQL Server Profiling
- 3. Before MS SQL • Be aware of Select N+1 Problem when using EF lazy loading • Log EF queries to find the problem:
- 5. Resource Monitor • First of all, check Resource Monitor and find sqlserver.exe
- 6. SQL Server Performance Counters
- 7. SQL Server Performance Counters • Access Methods: Full Scans/sec • Buffer Manager: Buffer Cache Hit Ratio • Database: Transactions/sec • Database: Write Transactions/sec • General Statistics: User Connections
- 8. SQL Server Performance Counters (cont.) • Locks: Lock Requests / Sec • Locks: Lock Waits / Sec • Locks: Lock Timeouts/sec • Locks: Number of Deadlocks/sec • SQL Statistics: Batch Requests/Sec • SQL Statistics: SQL Re-Compilations/sec • SQL Statistics: SQL Compilations/sec
- 9. Checking Execution Plan with SSMS
- 10. Actual Execution Plan in SQL Server
- 11. IO Statistics • Copy message into http://www.statisticsparser.com/ • Logical reads is the most prominent counter
- 12. Indexing
- 13. Indexing • Clustered index • Primary Key: Includes all columns • Non-clustered index • One or more columns: Include as many columns as needed • SELECT * FROM Device WHERE Serial = ‘0123456789’ • Execution plan suggests to add: • CREATE NONCLUSTERED INDEX (…) ON Device (Serial)
- 14. Indexing (cont.) • A new index is a new copy of the table, containing “key columns” and “include columns” • 4,516 logical read • SSMS hints: • Logical reads decreases to 13
- 15. Indexing (cont.) • Indexing helps • SELECT item using COLUMN = VALUE • COLUMN compare VALUE • INDEX name on table (a, b) • Sort on a values, then b values in each bucket on a • Not good for • SELECT id FROM table WHERE a > 100 and b = 200 • SELECT id FROM table WHERE b = 200 • Good for • SELECT id FROM table WHERE b > 100 and a = 200 • SELECT id FROM table WHERE a > 100
- 16. Investigating index usage (index_usage_stats) • SELECT OBJECT_NAME(ddius.[object_id], ddius.database_id) AS [object_name] , ddius.index_id , ddius.user_seeks , ddius.user_scans , ddius.user_lookups , ddius.user_seeks + ddius.user_scans + ddius.user_lookups AS user_reads , ddius.user_updates AS user_writes , ddius.last_user_scan , ddius.last_user_update FROM sys.dm_db_index_usage_stats ddius WHERE ddius.database_id > 4 -- filter out system tables AND OBJECTPROPERTY(ddius.OBJECT_ID, 'IsUserTable') = 1 AND ddius.index_id > 0 -- filter out heaps ORDER BY ddius.user_scans DESC
- 17. Identify indexes that have never been accessed • SELECT OBJECT_NAME(i.[object_id]) AS [Table Name] , i.name FROM sys.indexes AS I INNER JOIN sys.objects AS o ON i.[object_id] = o.[object_id] WHERE i.index_id NOT IN ( SELECT ddius.index_id FROM sys.dm_db_index_usage_stats AS ddius WHERE ddius.[object_id] = i.[object_id] AND i.index_id = ddius.index_id AND database_id = DB_ID() ) AND o.[type] = 'U‘ ORDER BY OBJECT_NAME(i.[object_id]) ASC
- 18. Tune Your Indexing Strategy with SQL Server DMVs • https://www.simple-talk.com/sql/performance/tune-your-indexing- strategy-with-sql-server-dmvs/
- 19. Database Engine Tuning Advisor
- 23. Some Columns in DB are Locked! • Use sp_who2 and check BlkBy to see who has blocked whom • sp_whoisactive does it in a better way
- 25. Profile Using Execution Plan Cache • Install sp_BlitzCache • Run: sp_blitzcache @sort_order = 'reads', @top = 20 • You will find “Query Text”, “Total Reads”, “# Executions”, “Execution Plan”, and some hints about how to improve performance of those queries. • It gave us a hint to add the following index: • We thought indexing this table is not important since it is solely used in admin panel. But sp_BlitzCache tells that 20% of reads are related to this function.
- 26. Other Diagnosing Stored Procedures • sp_BlitzIndex: Gives you hints on how much your indexes are used and suggestions for new indexes and their estimated effect • sp_Blitz: Checks common health measures of your DB • sp_AskBrent: For 5 seconds watches your DB and gives you a report on: • Blocking or long-running queries • If a backup, DBCC, or index maintenance job was running • Query wait statistics to figure out SQL Server’s current bottleneck • Perf. counters for CPU use, slow drive response times, or low Page Life Expectance • You can create an agent job which runs sp_AskBrent every 15 minutes and appends the result to a table in your DB (more info) • All these stored procedures are packaged in the ZIP file containing sp_BlitzCache
- 27. Evidences of Effect! • # of lock requests / sec • Avg. wait time • CPU usage
- 28. Questions?