Hi
We have used XFS on productions systems with great success. We will soon
enter a process of hardware upgrades. One of the servers was experiencing
a strange problem. By reading this list we concluded to be related to the
"default" small XFS log size on "larger" filesystems. I know I may be
wrong and the tests I will do before the upgrades will prove this to me.
This is what I thought to do , to optimize I/O for the future systems. I
will test with bonnie++ and other I/O benchmark tools different
configurations:
- different XFS log sizes
- different XFS log types (internal and external)
- different stripe size (beeing a hardware RAID5 device)
The configuration of the system for tests will be : 3 x 73 GB SCSI RAID5
array (hardware), 1 x 18 GB SCSI - onboard controller. XFS will be on the
single partition of the RAID array.
Other than playing with log size, stripe size, log position (internal into
the RAID5 array, or external on the sigle 18 gb disk) is there any other
things I can do to optimize/increase I/O ? I am interested in "fast seeks"
a lot more than actual I/O bandwidth.
Thank you
----------------------------
Mihai RUSU
Disclaimer: Any views or opinions presented within this e-mail are solely
those of the author and do not necessarily represent those of any company,
unless otherwise specifically stated.
|