[K12OSN] Impact of filesystems on performance

[Chris Kacoroski]

Hi,

I have been doing some testing of a 3ware raid card and found a problem 
with ext3 filesystems on the 3ware card (not sure if this applies to 
other kinds of raid devices).  Tests to replicate are:

Write: time dd if=/dev/zero of=/mountpoint/bigfile bs=1024k count=1024
Read: time dd if=/mountpoint/bigfile of=/dev/null bs=1024k count=1024

For concurrent tests, I just used different file names and started up 
mulitple xterms.  I took the average of three test runs for the numbers 
below.

Using Ext3, when doing a single read or write I was getting 41MB/s, 2 
concurrent writes was 36MB/s, and when I changed to a concurrent read 
and write it dropped to 4.5MB/s.   On the advice of 3ware I changed to 
the jfs filesystem and saw a big improvement.  Single read or write is 
60MB/s, 2 concurrent writes are 79MB/s, a concurrent read and write was 
99MB/s, and 1 read with 2 writes concurrently was 100MB/s.

Summary, if you use SATA disks, I highly recommend using the jfs 
filesystem over the ext3 filesystem.

cheers,

ski

--
"When we try to pick out anything by itself, we find it
   connected to the entire universe"		John Muir

Chris "Ski" Kacoroski, ckacoroski at nsd.org, 425-489-6263