On 29/04/13 20:21, James Harper wrote:
> dd if=/dev/zero of=test.bin bs=512 count=128 oflag=sync

All tests below on ext4 except the RAIDz one.

SATA spinning disks sw RAID0, kernel 3.5.0:
65536 bytes (66 kB) copied, 2.65235 s, 24.7 kB/s

SATA spinning disks sw RAID5, kernel 2.5.0:
65536 bytes (66 kB) copied, 4.36826 s, 15.0 kB/s

SATA spinning disks sw RAID10, kernel 3.2.0:
65536 bytes (66 kB) copied, 2.65235 s, 24.7 kB/s

(the size of this test is small enough that it's below the stripe size, 
hence similarities above)

SATA spinning disks, sw RAIDZ w/SSD log:
65536 bytes (66 kB) copied, 0.32133 s, 204 kB/s

Older SSD, kernel 3.5.0:
65536 bytes (66 kB) copied, 0.826156 s, 79.3 kB/s

A modern but budget SSD, kernel 3.2.0:
65536 bytes (66 kB) copied, 1.26909 s, 51.6 kB/s

Higher end SSD, kernel 3.2.0:
65536 bytes (66 kB) copied, 0.0313182 s, 2.1 MB/s


However after all that.. my guess is that your 10-year-old PC was one 
where xfs didn't support barriers.. Try mounting your modern PC with 
"nobarrier" and see what happens to the performance?
_______________________________________________
luv-main mailing list
[email protected]
http://lists.luv.asn.au/listinfo/luv-main

Reply via email to