raidz2 a bit big
Adam McDougall
mcdouga9 at egr.msu.edu
Sat May 23 13:26:46 UTC 2009
On Sat, May 23, 2009 at 06:04:04AM -0700, Randy Bush wrote:
>> NAME STATE READ WRITE CKSUM
>> tank ONLINE 0 0 0
>> raidz2 ONLINE 0 0 0
>> da0s3 ONLINE 0 0 0
>> da1s3 ONLINE 0 0 0
>> da2s1 ONLINE 0 0 0
>> da3s1 ONLINE 0 0 0
>> da4s1 ONLINE 0 0 0
>> da5s1 ONLINE 0 0 0
>> da6s1 ONLINE 0 0 0
>> da7s1 ONLINE 0 0 0
>> da8s1 ONLINE 0 0 0
>> da9s1 ONLINE 0 0 0
>> da10s1 ONLINE 0 0 0
>> da11s1 ONLINE 0 0 0
>
> Reads on such configurations are very slow.
how are writes?
> If one of your disk, for example, is capable of 100 IO per/sec, then:
> with 12 disks in one raidz2 vdev you get only 100 IOPS
> with 4 disks in raidz2 (total 3 raidz2 vdevs) you get 300 IOPS
> with 2 disks in mirror (total 6 mirror vdevs) you can get 1200 IOPS
ok. sounds nice. but then, don't i have six file systems and have to
start playing lay-out design games?
randy
For an example: (btw the read speed is fantastic in a mirror and the write
is notably faster than raidz, but if your I/O is all going to go through a
gig nic, then it may not matter such as if you are just using it for a
low concurrent user stash of large files)
zpool create tank mirror aacd0s1d aacd1s1d mirror aacd2s1d aacd3s1d mirror aacd4s1d aacd5s1d mirror
aacd6s1d aacd7s1d
# zpool status
pool: tank
state: ONLINE
scrub: none requested
config:
NAME STATE READ WRITE CKSUM
tank ONLINE 0 0 0
mirror ONLINE 0 0 0
aacd0s1d ONLINE 0 0 0
aacd1s1d ONLINE 0 0 0
mirror ONLINE 0 0 0
aacd2s1d ONLINE 0 0 0
aacd3s1d ONLINE 0 0 0
mirror ONLINE 0 0 0
aacd4s1d ONLINE 0 0 0
aacd5s1d ONLINE 0 0 0
mirror ONLINE 0 0 0
aacd6s1d ONLINE 0 0 0
aacd7s1d ONLINE 0 0 0
More information about the freebsd-fs
mailing list