December 21, 2005 at 8:45 am
Hi,
I'm about to make a new SQL server.
I've read several places that a good hardware configuration is to place database on on disk drive and logfile on another drive - because harddisk head won't have to move on the logfile drive since it is written sequentiel.
But what is the scenario if you lets say 30 databases on that server?
Then the harddrive head on the logfile drive would still have to jump/search between different logfiles...
So is the above mentioned recommendation only good for 1 single database ??
December 22, 2005 at 12:10 am
Hi,
If you have the money and space to add 30 hard drives you will definitely have the I/O performance edge. I think that would not be realistic. When making your storage decision you should consider designing a solution that will provide fault tolerancy with some performace. Use mix of RAID 5, RAID 10 or RAID 1 settings. We had limited budget when upgrading our db server. We used RAID 5 (consist of 4 physical hd) to store mdf files and RAID 10 (4 physical drives) to store log files.
December 22, 2005 at 12:50 am
Thanks...
But lets say you'll have to host 30 databases.
What would be best:
a) 1 RAID1 (2 disks) for the data files + 1 RAID1 (2 disks) for the logfiles
or
b) 1 RAID10 (4 disks) - both data and logfiles on same logical drive
What would you choose - and why?
December 22, 2005 at 6:33 am
Hi,
We have six or seven databases running and use two separate SCSI disk controllers and three logical drives.
Controller 1 has:
Drive1: Operating System and Software, and backup repository on two mirrored 80gb drives
Drive2: Log Files on two 80Gb mirrored drives
Controller 2 has:
Drive 3: Databases on 4 x 136Gb drives in a RAID5 config with a spare 'hot swap' drive.
Chose this configuration because we have a lot of PERL scripting work going on on the software drive, and I wanted to separate the software, the log files and the database. Used two separate controllers because I got the idea from somewhere that controller contention was also a potential bottleneck.
David
If it ain't broke, don't fix it...
December 27, 2005 at 10:48 am
Even with 30 databases (or more), the best advice is to put data and log files on separate physical drives (regardless of RAID configuration).
The I/O traffic is different. As you mentioned, the log file access is sequential and data file access is random. Log file changes must be written immediately to complete a transaction, data file updates can be handled by the lazy writer thread in the background. The cache built into data disks may be useful if a data page is accessed again (and has been flushed from the buffer pool), but log data is seldom re-read.
Viewing 5 posts - 1 through 4 (of 4 total)
You must be logged in to reply to this topic. Login to reply