So far the best presentation of the conference *sorry Brad
Highlights:
Clear out the run history of all your jobs. This can really impact your jobs if you let it get cluttered
Spedometer - Use PerfMon
MIIS Objects
For Imports Objects Read/Sec
Objects Synchronized/Sec
For Exports Objects Exported/Sec
But wait, there's a script: http://www.ilmbestpractices.com/Articles/Speed_of_Past_Runs
as tables grow - Page splits, fragmentation, more scans than seeks, growth of log and data files.
monitor with Page Splits/Sec in perf mon how do we avoid it
scan v seek you get more scans as get more fragmented indices
read entire tables versus using an index
monitor with PerfMon
Full Scans/Sec
Auto Create/Update stats are enabled
File Growth
if auto growth is enabled
clear run history script on website
http://www.ilmbestpractices.com
DBCC SHRINKDB('MicrosoftIdentityIntegrationServer', NO TRUNCATE)
maybe preset size if you can predict
Log file should be 1/4 of total data file
Transaction Log is key to performance as well
Recovery Models
Full - LOG EVERYTHING
Simple - only as good as last full or diff
Bulk Logged hardly ever used with MIIS - only good for bulk logged operations
Red Gate SQL Backup to compress backup files (compression eats CPU time but far less Disk IO
When you clear run history it will host your transaction log. Do this during a maintenance window and
a. Truncate log
b. Change to SImple Recovery Mode
c. Clear run history in small chunks
d. Then go to Full Recovery mode and Perform a Full Backup
you can also add some indices to better performance
indexdefrag on DB every month, index rebuild quarterly till SQL 2005 EE. With SQL2005EE we get to rebuild index with DB online as well as rebuild index in TempDB
http://www.sqlbestpractices.com
Subscribe to:
Post Comments (Atom)
0 comments:
Post a Comment