July 10, 2007 at 3:53 am
Hi Gurus,
I have a Problem with the Transaction Log BackUp's.
It Runs every 15 Minutes and since its a Production Server we have very High Transaction Load.
There are around 20 Databases on server and every 15 Minutes there is a Peak showing in SAN Drives.
I want to minimise this Peak.
Is there any other Solution to this.
Thanks for your Help.....!!!!!!!!!!!!!!!!!
July 10, 2007 at 4:01 am
Do you have the backup going to different drives from the log and/or data? That should help. Also, do you need all databases backed up every 15 minutes? You might be able to set different schedules based on different requirements.
July 10, 2007 at 4:21 am
July 10, 2007 at 4:35 am
The only way to minimize the peaks are to run log backups more often (so less data in each backup) or use hardware to increase the speed. Using a R1 or R10 drive setup might help.
July 10, 2007 at 5:00 am
Go for better hardware. Also use RAID levels that does not impact much on write performance.
Cheers,
Sugeshkumar Rajendran
SQL Server MVP
http://sugeshkr.blogspot.com
July 10, 2007 at 6:31 am
July 10, 2007 at 7:28 am
Are you sure that the other databases tlog backup will complete within 30 secs else you will have the problem accompanying right.
Cheers,
Sugeshkumar Rajendran
SQL Server MVP
http://sugeshkr.blogspot.com
July 11, 2007 at 2:56 am
One more Interesting Fact to share is that the SAN Drive is shared by the SQL Servers of other applications as well.
1 Most Important thing is all the SQL Servers run Trans Log BackUp after every 15 minutes.
A little bit more the SAN Drive configured for the FULL & Trans Log Backup shows a Sharp rise in READS and not Writes.
I am confused now how SQL Server performs READS operation on SAN Drive configured for the FULL & Trans Log Backups and why for what reason?
Thanks.
July 11, 2007 at 3:15 am
Does the database being used for Transactional purpose and is there large number of updates/deletes/insert happening every second. If so then there is a possibility as SQL server reads backup file and the datafiles so that the backup is consistent with the state of database till the time the backup is run.
Cheers,
Sugeshkumar Rajendran
SQL Server MVP
http://sugeshkr.blogspot.com
July 11, 2007 at 6:20 am
y0u will see an i/o spike on the drive containg the logs and the drive you are backing up to when the backups take place, this is unavoidable.
your idea of spreading out the backups so they do not all occur at the same time is a good idea and definitely worth doing.
ensure if you can log and backup drives are seperate drives through different channels
Other than that best you can do is get the fastest disks you can and use raid 10.
Are the io spikes actually causing a slow down to the apps? Is average disk q length unnacceptable?
---------------------------------------------------------------------
July 11, 2007 at 6:44 am
RAID 1+0 is the best other option that i forgot to say anyways thanks for george for having rminded for that.
Cheers,
Sugeshkumar Rajendran
SQL Server MVP
http://sugeshkr.blogspot.com
Viewing 11 posts - 1 through 10 (of 10 total)
You must be logged in to reply to this topic. Login to reply