SQL Server 計數器 閥值列表

List of Threshold Value for SQL Server Performance Counters

By : Kasim Wirama, MCDBA, MCITP

 

Monitoring database performance is proactive task that a DBA should do. Possible bottleneck for SQL Server is memory, processor, I/O subsystem, tempdb database, and locking. For SQL Server specifics, below list is a minimum set of counters that should be used for monitoring SQL Server performance daily. I would give preferred value but your environment might require the value stick with the preferred one because each database environment is unique. For your initial baseline, it could your reference.

1.       1. SQL Server:Access Methods with counter : Forwarded Records/sec

Preferred value : less than 10 of 100 Batch Requests/Sec

 

2.       2. SQL Server:Access Methods with counter : Full Scans/sec

Preferred value : (Index searches/sec)/(Full scans/sec) should be greater than 1000

 

3.      3. SQL Server:Access Methods with counter : Index Searches/sec

Preferred value : same with point 2

 

4.       4. SQL Server:Access Methods with counter : Page Splits/sec

Preferred value : less than 20 of 100 Batch Requests/sec

 

5.       5. SQL Server:Buffer Manager with counter : Buffer Cache Hit Ratio

Preferred value : greater than 90%

 

6.       6. SQL Server:Buffer Manager  with counter : Free List Stalls/sec

Preferred value : less than 2

 

7.       7. SQL Server:Buffer Manager with counter : Free Pages

Preferred value : greater than 640

 

8.       8. SQL Server: Buffer Manager with counter : Lazy Writes/Sec

Preferred value : less than 20

 

9.       9. SQL Server: Buffer Manager with counter : Page Life Expectancy

Preferred value : greater than 600 seconds.

 

10.   10. SQL Server: Buffer Manager with counter : Page Lookup/sec

Preferred value : (Page Lookup/sec)/(Batch Requests/sec) should less than 100

 

11.   11. SQL Server: Buffer Manager with counter : Page Reads/sec

Preferred value : less than 90

 

12.   12. SQL Server: Buffer Manager with counter : Page Writes/sec

Preferred value : same with point 11

 

13.   13. SQL Server: General Statistics with counter : Logins/sec

Preferred value : less than 2

 

14.   14. SQL Server: General Statistics with counter : Logouts/sec

Preferred value : less than 2

 

15.   15. SQL Server: Latches with counter : Latch Waits/sec

Preferred value : (Total Latch Wait Time)/(Latch Waits/Sec) less than 10

 

16.   16. SQL Server: Latches with counter : Total Latch Wait Time (ms)

Preferred value : same with point 15

 

17.   17. SQL Server: Locks with counter : Lock Wait Time(ms)

Preferred value : should not exceeds 60 seconds.

 

18.   18. SQL Server: Locks with counter : Lock Waits/sec

Preferred value : should be zero

 

19.   19. SQL Server: Locks with counter : Number of Deadlocks/sec

Preferred value : less than 1

 

20.   20. SQL Server: SQL Statistics with counter : SQL Compilations/sec

Preferred value : less than 10% of number of Batch Requests/sec

 

21.   21. SQL Server: SQL Statistics with counter : SQL Re-Compilations/sec

Preferred value : less than 10% of number of SQL Compilations/sec

 

Generally 15 seconds sample rate is sufficient and it is recommended to save it into CSV for later analysis.

    本文轉自 Fanr_Zh 部落格園部落格，原文連結：http://www.cnblogs.com/Amaranthus/archive/2011/09/21/2183078.html，如需轉載請自行聯絡原作者