Help me build my ZFS box

SilverJS said:
Either way, I agree, a 7hour downtime is not a lot, unless I'm away, which I tend to be somewhat often with the job.

It's NOT downtime, the pool/data will be available all the time, it's just how long would take for HOTSPARE to RESILVER, so the ZFS pool will still be REDUNDANT.
 
vermaden said:
Yes, with 2 * RAIDZ(3) You will get about 2 x performance of RAIDZ2(6) because these RAIDZ(3) are striped.

What makes you think that? RAIDZ2 stripes data across 6 disks (2 parity), RAIDZ stripes data across 2x3 disks (2 parity). Performance should be very similar.

The big difference is in fault-tolerance where RAIDZ2 is much better for the same number of disks. See this for a comparison of different protection schemes.
 
SilverJS said:
I see! I'm just starting to get into all of this, and a LOT more research is required - but so far, using FreeNAS 8.0.1 BETA3 and a RAIDZ2(6)+hotspare, I'm getting write and read speeds of about 11-12 Mb per second, which to me seems abysmally low...again, I'll have to research if this is due to network bottleneck or whatever (I'm very green!), but we'll see. Either way, I agree, a 7hour downtime is not a lot, unless I'm away, which I tend to be somewhat often with the job. I guess that's another consideration.

11-12Mb/s would be normal for a 100mbit network.
 
jalla said:
What makes you think that?
Its the same difference, like putting 4 drives into 4-way mirror [1] and putting drives into stripe of 2 mirrors [2], the second one would be 2 * faster. Its like comparing RAID10 to RAID11.

[1]
Code:
[1]-[2]-[3]-[4] MIRROR

[2]
Code:
[1]-[2] MIRROR \
                 STRIPE
[3]-[4] MIRROR /

RAIDZ(3) + RAIDZ(3) is like RAID50 while RAIDZ2(6) is 'only' a RAID6.
 
OP says nothing about mirroring. The question here is about two striped vdevs vs one.

Mirroring two RAIDZ vdevs halves the effective storage.
 
@jalla

Have you ever heard about examples maybe?

Start understanding what you are reading, this could help take this conversation to another level.
 
@vermaden
You're answering a question that hasn't been posed. As such, you bring nothing useful to the discussion.
 
Comparing 2x RAIDz1 vs 1x RAIDz2, the former will provide twice the write IOPS than the latter. I think read performance should be similar for both options but don't have that 100% clear myself. But I think stating twice the performance is over simplifying a little...

With regard rebuild times, that is in a perfect world. I see over 24 hour rebuilds on disks with similar average write performance, which may be due to the rather old Dell server I have but also due to other factors such as other IO load on the zpool etc etc..

Andy.
 
AndyUKG said:
Comparing 2x RAIDz1 vs 1x RAIDz2, the former will provide twice the write IOPS than the latter. I think read performance should be similar for both options but don't have that 100% clear myself. But I think stating twice the performance is over simplifying a little...
I don't agree.

The limiting factor is the combined number of IOPS the disks can sustain. If you're reading/writing 128k records that should distribute the load evenly across all 6 disks in parallell, regardless of whether they are organized in 2x3 or 1x6 vdevs.

There are different scenarios that could affect perfomance slightly (read vs write, sequencial vs random, small vs large block), but basically with the same number of disks and parity, plain raidz and raidz2 should have similar performance.
 
Back
Top