Is it possible to fully use Mirror instead of RAIDZ?

[molay]

How many drives (and of what kind) can a/this specific DELL R730XD fit?

My DELL R730XD only has 12x 3.5" and 2x 2.5" SATA/SAS bays, and NVMe conversion cards (two NVMe SSDs) can only be installed on the RISER1 card (3x half height PCIe 3.0x8).
The reason for choosing HDD is that it is large and cheap.
I also know that SSD is fast, but SATA/SAS SSD is very expensive, and the cost performance is much lower than that of NVMe SSD.
The PCIe slots are limited, and I can install up to 3x2 NVMe SSDs in the RISER1 card, but it will generate a huge amount of heat which may cause some issues.
So currently I still tend to use HDD devices.

I read the threads of the moderators and found that Mirror does bring many advantages, such as relatively less impact from disk fragmentation, much shorter re-silvering time than RAIDZn, higher IOPS, and more suitable for block storage.
My key question is: Is it possible to completely use Mirror to replace RAIDZn in daily use?
If yes, for SOHO, on the premise of ensuring the security of the data pool, is it more recommended to use 2-way Mirror or 3-way Mirror?

For projects with a large number of small files, If the access speed has reached the limit under a reasonable configuration, I can accept it.
In the future If I can find an inexpensive hardware solution that can accommodate multiple M.2/U.2 SSDs at the same time, with good cooling, I may create a separate TrueNAS server for these projects.
Then I could really fly.

 

[NugentS]

Yes HDD is cheap - but it doesn't really match all your requirements well.
However as an example I use 10 disks in mirrored vdevs (2 way) to create a 5 vdev pool of HDD's. I also include a special vdev for small files (which might be an option for you). I only use 2 way mirrors.

I have a good 3-2-1 backup regime that is tested on a fairly regular basis - so I don't worry about the only 1 spare disk per vdev. At the moment I even have a hot spare. I resilvered a drive a few days ago - it took a few hours.

I use mirrors because:
1. I use some of the space as an iSCSI volume for ESXi
2. I like the quick resilver
3. Its easier to expand with 2 new disks - and they don't have to be the same size as the other vdevs

 

[jgreco]

I'm sorry that my question may be meaningless debate, or offensive to the forum.

It's not. It's some of the following posts. The responses you gathered from @Davvo , @NugentS , @Ericloewe , were the generally useful and topical TrueNAS/ZFS targeted stuff normally found in these forums. We were having some problems with a different argumentative poster and I originally closed this thread but we discussed it a bit more and I opted for a more surgical solution. However, since I was dealing with several threads and dozens of posts, I missed noticing the wording of my closing post in this thread and failed to update it to reflect the status. Sorry for any confusion, your posts were completely fine and absolutely nothing wrong with them.

 

[Davvo]

My key question is: Is it possible to completely use Mirror to replace RAIDZn in daily use?

Yes if you are willing to sacrifice at least half of your total space and some straming writing speed (see below).

If yes, for SOHO, on the premise of ensuring the security of the data pool, is it more recommended to use 2-way Mirror or 3-way Mirror?

For the purpose of small files, streaming reads and writes are almost irrilevant; let's assume that a single drive is able to do 250 IOPS and a continuous throughput of 150 MB/s:

3 vdevs composed each of a 2-way mirror

• read IOPS: 1500
• write IOPS: 750
• streaming reads: 900 MB/s
• streaming writes: 450 MB/s
• space efficency: 50%
• parity level: 1 disk per vdev (3 total)
• total disks: 6

2 vdevs composed each of a 3-way mirror

• read IOPS: 1500
• write IOPS: 500
• streaming reads: 900 MB/s
• streaming writes: 300 MB/s
• space efficency: 33%
• parity level: 2 disks per vdev (4 total)
• total disks: 6

You excange lesser space and performance for great resiliency. I would probably use 2-way mirrors in your situation, but only if I would need to service small data. For bulk data (ie backups or media files) I would use RAIDZ2. Perhaps it's possible to service both kinds of file using special vdevs as suggested by @NugentS.

 

[molay]

Sorry for any confusion, your posts were completely fine and absolutely nothing wrong with them.

No need to apologize at all, it was just a misunderstanding.

Thanks for all the help! @Davvo , @NugentS , @Ericloewe

After comprehensive consideration, I decided to use 6x 2-way Mirrors to store all SOHO data, and pay attention to cold backup and cloud backup of important data.

Now I have another question, is it necessary for me to put macOS TimeMachine backups, static media files in separate pool of 3x 2-way Mirror vdevs? Then put the database, block storage, projects with a lot of small files in another pool of 3x 2-way Mirror vdevs?
If yes, can my Optane 900P serve as SLOG devices for both pools at the same time?

 

[Davvo]

If yes, can my Optane 900P serve as SLOG devices for both pools at the same time?

No, one SLOG per pool.

Show : But...

...if you partition it it's possibile:

SOLVED - 1 SSD as SLOG for two separate pools?

Hi, I have been doing some benchmarking on my Dell R720. I can clearly see the massive difference a ZIL drive makes when using NFS shares vs iSCSI. I have 2 pools that I would ideally include a zil in. But I have no space to add two separate SSDs in the server. So I have 1 SSD that I could use...

www.truenas.com

 

[NugentS]

No - just use different datasets. I would probably tweak the record size for different datasets. 1M for backups, 128K or smaller for small files
Yes it can BUT (and here be dragons) it will require manual setup as this is not possible through the GUI. This is not reccomended.

The SLOG will only effect sync writes (its not a write cache). VMWAre, NFS and MAC writes are I believe likley to be sync by default.

 

[molay]

Thank you for all the answers!
I have re-established the pool that only using 2-way Mirror vdevs.
It works very well with Optane 900P as ZLOG device, and iSCSI, SMB, TimeMachine speed is very fast in daily use.
For projects with a large number of small files, the current speed is also acceptable.
I'm glad the memory usage is finally exceeded 100GB.