June 11, 2013 at 1:30 pm
This is most likely my biggest weakness at the moment, but I do not know how to go about justifying better hardware with some real facts that the non techs would understand. So what is the best approach to this.
Not sure if this would help, but the basis system info:
I am running a virtual SQL 2012 server 96GB RAM 8 cores.
It is attached to a Equilogics SAN Array (15k disks)
I am getting PATEIOLATCH_EX (you can add almost all the codes after the _)
Surprisingly, partitioning actual speed the system up tremendously.
Some stats:
to delete 107,000 records it takes about 15 seconds batching 5000 at time (this seems long)
to Select 107,000 and insert them into another table is take 12 seconds using the SSIS data flow component
SQL_Noob
June 11, 2013 at 2:53 pm
1) virtual environment could be misconfigured. I have NEVER been to a client that already had such an environment set up with SQL Server running on it where they did everything optimal, and most of the time things are HORRIBLY bad.
2) First thing to do is TUNE your database application/structures/code. Aggregate profiler analysis is a huge help with this.
3) Then do fileIO stall and waitstats analysis
4) the "slow" DELETEs could have many causes, such as having to scan a big table per 5K batch, too many indexes on the table, heap with lots of forwarding pointers, blocking from other users, etc, etc.
Best,
Kevin G. Boles
SQL Server Consultant
SQL MVP 2007-2012
TheSQLGuru on googles mail service
June 12, 2013 at 8:38 pm
Starting doing tuning today. Got a good chunk of data during heavy load.
How would I go about tracking the metrics for File IO? My idea was to save Database I/O metrics over time (maybe a week or so) and chart it.
I feel noobish since it's been about 13 years since I did any server type administrative work.
Thanks for the help.
June 13, 2013 at 7:19 am
pjbaratelli (6/12/2013)
Starting doing tuning today. Got a good chunk of data during heavy load.How would I go about tracking the metrics for File IO? My idea was to save Database I/O metrics over time (maybe a week or so) and chart it.
I feel noobish since it's been about 13 years since I did any server type administrative work.
Thanks for the help.
A week is FAR too long to look at IO stalls. When I am tuning I do 2-5 minute differential analysis while the server is under load. Download my Tune Like a Guru session from here: http://sqlsaturday.com/81/schedule.aspx
I hope you grabbed a long-run profiler trace to disk so you can then load it into a SQL Table or run it through Qure to do aggregate execution analysis. Many people fail to recognize that the thing(s) that are most important to tune are not necessarily the sprocs/calls that run the longest. If you can knock 25% of the duration of something that runs in 0.4 seconds but gets called 200 times per minute that could be the biggest win for the server and your app.
Best,
Kevin G. Boles
SQL Server Consultant
SQL MVP 2007-2012
TheSQLGuru on googles mail service
Viewing 4 posts - 1 through 3 (of 3 total)
You must be logged in to reply to this topic. Login to reply