I realized that I'll loose too much disk space with the "double" raid
configuration suggested below. Agree?
I've done some performance testing with raidz/raidz1 vs raidz2:
bash-3.00# zpool status -v raid5
pool: raid5
state: ONLINE
scrub: none requested
config:
NAME STATE READ
WRITE CKSUM
raid5 ONLINE 0
0 0
raidz1 ONLINE 0
0 0
c7t6000402001FC442C609DC5A300000000d0 ONLINE 0
0 0
c7t6000402001FC442C609DCA4A00000000d0 ONLINE 0
0 0
c7t6000402001FC442C609DCA2200000000d0 ONLINE 0
0 0
c7t6000402001FC442C609DCABF00000000d0 ONLINE 0
0 0
c7t6000402001FC442C609DCADB00000000d0 ONLINE 0
0 0
c7t6000402001FC442C609DCAF800000000d0 ONLINE 0
0 0
c7t6000402001FC442C609F029100000000d0 ONLINE 0
0 0
errors: No known data errors
bash-3.00# zpool list
NAME SIZE USED AVAIL CAP HEALTH ALTROOT
raid5 12.6T 141K 12.6T 0% ONLINE -
bash-3.00# df -h /raid5
Filesystem size used avail capacity Mounted on
raid5 11T 41K 11T 1% /raid5
bash-3.00# echo zfs_nocacheflush/D | mdb -k
zfs_nocacheflush:
zfs_nocacheflush: 0
bash-3.00# ./filesync-1 /raid5 10000
Time in seconds to create and unlink 10000 files with O_DSYNC: 9.871197
bash-3.00# echo zfs_nocacheflush/W1 | mdb -kw
zfs_nocacheflush: 0 = 0x1
bash-3.00# ./filesync-1 /raid5 10000
Time in seconds to create and unlink 10000 files with O_DSYNC: 7.363303
Then I destroyed the raid5 pool and created a raid6 pool:
bash-3.00# zpool status -v raid6 pool: raid6 state: ONLINE scrub:
none requested
config:
NAME STATE READ
WRITE CKSUM
raid6 ONLINE 0
0 0
raidz2 ONLINE 0
0 0
c7t6000402001FC442C609DC5A300000000d0 ONLINE 0
0 0
c7t6000402001FC442C609DCA4A00000000d0 ONLINE 0
0 0
c7t6000402001FC442C609DCA2200000000d0 ONLINE 0
0 0
c7t6000402001FC442C609DCABF00000000d0 ONLINE 0
0 0
c7t6000402001FC442C609DCADB00000000d0 ONLINE 0
0 0
c7t6000402001FC442C609DCAF800000000d0 ONLINE 0
0 0
c7t6000402001FC442C609F029100000000d0 ONLINE 0
0 0
errors: No known data errors
bash-3.00# zpool list
NAME SIZE USED AVAIL CAP HEALTH ALTROOT
raid6 12.6T 195K 12.6T 0% ONLINE -
bash-3.00# df -h /raid6
Filesystem size used avail capacity Mounted on
raid6 8.8T 52K 8.8T 1% /raid6
bash-3.00# echo zfs_nocacheflush/D | mdb -k
zfs_nocacheflush:
zfs_nocacheflush: 0
bash-3.00# ./filesync-1 /raid6 10000
Time in seconds to create and unlink 10000 files with O_DSYNC: 9.879219
bash-3.00# echo zfs_nocacheflush/W1 | mdb -kw
zfs_nocacheflush: 0 = 0x1
bash-3.00# ./filesync-1 /raid6 10000
Time in seconds to create and unlink 10000 files with O_DSYNC: 7.560435
My conclusion on raidz1 vs raidz2 would be no difference in
performance and big difference in disk space available.
On 10. mars. 2009, at 09.13, Lars-Gunnar Persson wrote:
How about this configuration?
On the Nexsan SATABeast, add all disk to one RAID 5 or 6 group. Then
on the Nexsan define several smaller volumes and then add those
volumes to a raidz2/raidz zpool?
Could that be an useful configuration? Maybe I'll loose too much
space with "double" raid 5 or 6 configuration? What about performance?
Regards,
Lars-Gunnar Persson
On 10. mars. 2009, at 00.26, Kees Nuyt wrote:
On Mon, 9 Mar 2009 12:06:40 +0100, Lars-Gunnar Persson
<lars-gunnar.pers...@nersc.no> wrote:
1. On the external disk array, I not able to configure JBOD or
RAID 0
or 1 with just one disk.
In some arrays it seems to be possible to configure separate
disks by offering the array just one disk in one slot at a
time, and, very important, leaving all other slots empty(!).
Repeat for as many disks as you have, seating each disk in
its own slot, and all other slots empty.
(ok, it's just hear-say, but it might be worth a try with
the first 4 disks or so).
--
( Kees Nuyt
)
c[_]
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
.--------------------------------------------------------------------------.
|Lars-Gunnar
Persson |
|IT-
sjef |
|
|
|Nansen senteret for miljø og
fjernmåling |
|Adresse : Thormøhlensgate 47, 5006
Bergen |
|Direkte : 55 20 58 31, sentralbord: 55 20 58 00, fax: 55 20 58
01 |
|Internett: http://www.nersc.no, e-post: lars-
gunnar.pers...@nersc.no |
'--------------------------------------------------------------------------'
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
