zfs discuss - Jul 2010 - raidz2 + spare or raidz3 and no spare for nine 1.5 TB SATA disks?

The only other zfs pool in my system is a mirrored rpool (2 500 gb disks). This
is for my own personal use, so it''s not like the data is mission
critical in some sort of production environment.

The advantage I can see with going with raidz2 + spare over raidz3 and no spare
is I would spend much less time running in a degraded state when a drive fails
(I''d have to RMA the drive and wait most likely a week or more for a
replacement).

The disadvantage of raidz2 + spare is the event of a triple disk failure. This
is most likely not going to occur with 9 disks, but certainly is possible. If 3
disks fail before one can be rebuilt with the spare, the data will be lost.

So, I guess the main question I have is, how much a performance hit is noticed
when a raidz3 array is running in a degraded state?

Thanks

- Jack
-- 
This message posted from opensolaris.org

On Jul 27, 2010, at 10:37 PM, Jack Kielsmeier wrote:
> The only other zfs pool in my system is a mirrored rpool (2 500 gb disks).
This is for my own personal use, so it''s not like the data is mission
critical in some sort of production environment.
> 
> The advantage I can see with going with raidz2 + spare over raidz3 and no
spare is I would spend much less time running in a degraded state when a drive
fails (I''d have to RMA the drive and wait most likely a week or more
for a replacement).
raidz3 with no spare will be better than raidz2+spare in all single-set cases.
> The disadvantage of raidz2 + spare is the event of a triple disk failure.
This is most likely not going to occur with 9 disks, but certainly is possible.
If 3 disks fail before one can be rebuilt with the spare, the data will be lost.
> 
> So, I guess the main question I have is, how much a performance hit is
noticed when a raidz3 array is running in a degraded state?
The performance will be similar, but in the non-degraded case, the raidz3 
will perform better for small, random reads.
 -- richard

-- 
Richard Elling
richard at nexenta.com   +1-760-896-4422
Enterprise class storage for everyone
www.nexenta.com

> The performance will be similar, but in the non-degraded case, the
> raidz3
> will perform better for small, random reads.
Why is this? The two will have the same amount of data drives....

Vennlige hilsener / Best regards

roy
--
Roy Sigurd Karlsbakk
(+47) 97542685
roy at karlsbakk.net
http://blogg.karlsbakk.net/
--
I all pedagogikk er det essensielt at pensum presenteres intelligibelt. Det er
et element?rt imperativ for alle pedagoger ? unng? eksessiv anvendelse av
idiomer med fremmed opprinnelse. I de fleste tilfeller eksisterer adekvate og
relevante synonymer p? norsk.

On Jul 28, 2010, at 8:34 AM, Roy Sigurd Karlsbakk wrote:
>> The performance will be similar, but in the non-degraded case, the
>> raidz3
>> will perform better for small, random reads.
> 
> Why is this? The two will have the same amount of data drives....
The simple small, random read model for homogeneous drives:
	I = small, random IOPS of one drive
	D = number of data disks
	P = number of parity disks
	total IOPS = I * (D+P)/D

raidz2: P=2
	total IOPS = I * (D+2)/D

raidz3: P=3
	total IOPS = I * (D+3)/D

 -- richard

-- 
Richard Elling
richard at nexenta.com   +1-760-896-4422
Enterprise class storage for everyone
www.nexenta.com

Thanks,

Looks like I''ll be using raidz3.
-- 
This message posted from opensolaris.org