[dba-SQLServer] SQL Server File locations and speed

Francisco Tapia fhtapia at gmail.com
Wed Sep 20 22:40:54 CDT 2006


John,
   It has to do with physical speed.  If you have one raid5 w/ data
and log files, then the OS has to go through one raid controller to
get to either the log files or the datafiles, the database is
constantly writing to the log file while transactions are occuring,
and it is also constantly reading data from the data files while in
serving up data.  If you break down the raid5 into different drives
and controller channels you increase your throughput, as each drive
has a dedicated channel... a better method than just a raid5 would be
something like a raid5 for data, and a mirror raid for your
transaction logs.  even better if you can afford it would be raid10
for data and likewise for your log files, but that's really more
likely for the most demanding applications, and last I remember you
are juggling several hundred gig, wich may just be suitable that
coupled w/ bulk logging to minimize the size of the log files would
help the speed tremendously.

On 9/20/06, JWColby <jwcolby at colbyconsulting.com> wrote:
> Does anyone have any insight on the pros and cons of using a (fast) Raid5 as
> one big partition which holds the database data files as well as the log
> files, vs breaking a single raid5 array down into partitions and placing the
> log files on one partition and the data files on another, vs building two
> raid arrays and placing the data files on one raid array and the log files
> on another.
>

-- 
-Francisco
http://sqlthis.blogspot.com | Tsql and More...



More information about the dba-SQLServer mailing list