On 12/23/2013 6:50 AM, Joseph Kregloh wrote:
Well the original architect started out in the 80s with banking databases they just kept that model without revisiting if it works well or not, that might explain it a little bit. But also given the size of our tables we use the physical disks and filesytem advantages to improve speed and performance of the application, but not as often as I would like. We have a pretty big database.
I've found these days, you're usually better off just stripping all your mirrors into one big raid10, and letting statistics load balance your IO. I've got stripe sets of as many as 20 small-fast drives, totalling several terabytes, using XFS (Linux), or ZFS (Solaris, BSD), or JFS2 (AIX), all of which seem to handle the large file system quite efficiently.
--
john r pierce 37N 122W
somewhere on the middle of the left coast
