[OpenIndiana-discuss] ISCSI SSD Performance

[Sašo Kiselkov]

On 02/18/2013 01:42 PM, Grant Albitz wrote:
> The results below were done locally and are Gigabytes ps. I will try going back to the h310.

Then you should definitely expect higher numbers. You can test to see if
it is ZFS that's slowing you down by doing raw-device writes/reads, like
so (don't do this to devices which are part of an active ZFS pool!):

# dd if=/dev/urandom of=testfile bs=1024k count=10
# dd if=testfile bs=1024k >> testfile	# build a large test file
					# interrupt with ^C when it's
					# large enough (a few GB)

# # to test writes
# dd if=testfile of=/dev/rdsk/c1t0d0s0 bs=1024k

# # to test reads
# dd if=/dev/rdsk/c1t0d0s0 of=/dev/null bs=1024k

Watch the iostat %b column, you should be able to max the device out.
For instance, my SATA-II attached SSD looks like this when I try this:

# dd if=/dev/rdsk/c2t4d0s0 of=/dev/null bs=1024k
^C4929+0 records in
4928+0 records out
5167382528 bytes (5.2 GB) copied, 19.7543 s, 262 MB/s

                    extended device statistics
    r/s    w/s   kr/s   kw/s wait actv wsvc_t asvc_t  %w  %b device
  247.6    0.0 253521.8    0.0  0.0  1.0    0.1    3.9   1  96 c2t4d0
                    extended device statistics
    r/s    w/s   kr/s   kw/s wait actv wsvc_t asvc_t  %w  %b device
  248.0    0.0 253965.6    0.0  0.0  1.0    0.1    3.9   1  96 c2t4d0
                    extended device statistics
    r/s    w/s   kr/s   kw/s wait actv wsvc_t asvc_t  %w  %b device
  250.0    0.0 255998.7    0.0  0.0  1.0    0.0    3.8   1  96 c2t4d0

The writes may not reach device saturation, as the /dev/rdsk interfaces
are running through a char-device emulation layer in the kernel.

Cheers,
--
Saso