Re: Linux and SSD and how to decrease io delay : question, please

DanielMariella Petrini wrote:
Hi,

I have run a benchmark using the latest Linux kernel
and having
still the elevator set to noop (for each hard drive)
and vm.dirty_expire_centisecs = 6000

The system has 24GB of RAM, 2 SSD Mtron drives on a
RAID controller.
The databases are spread across two ext3 filesystems
mounted on the SSD drives (one filesystem maps to one
SSD drive and the other one to the other SSD drive.
Each drive is seen as a single volume).
The system has 2 cpus with 2 cores each.

  

Have you mounted those filesystem with the "noatime" option? If you don't do that you risk causing a *lot* of unwanted seek and write activity to the journal.

Before starting each benchmark the system has
approximately 16GB of RAM already cached (so that most
of the file reads are reduced)
  

-- 
Bill Davidsen [off-list ref]
 "Woe unto the statesman who makes war without a reason that will still
 be valid when the war is over..." Otto von Bismark