How to get max throughput with my components

[quei]

Hi All

I am stranded with my idea to use a ZFS pool for max throughput. I have learned that I cannot easily extend the pool with one disk increments.

 

I would appreciate your guidance now on how I should set up my NAS.

The goal should be to get the max read and write speeds with these components:

1 x AsRock X570M Pro4 with 48 GB RAM

2 x kingston 500GB nvme

2 x WD Blue 500 GB SSD

4 x Seagate Exos 7E10 8TB HDD

 

The max throughput is because we will move into a house with 10gbe infrastructure. This means the internal network is capable of handling 10gbe. Also, the ISP can handle this speed. So I have the perfect base to set up a NAS with good throughput. So my goal should be to get the max throughput with these components.

 

The initial plan was:
1 pool with 2 nvme disks with one spare (used for VM placement)

1 pool with 2 ssd disks with one spare (used for application data, mainly docker things)

1 pool with 4 hdd disks with one spare (as the main data pool)

 

So I hoped to reach around 240 - 300 MB with the hdd pool because of parallel reads and writes to the disks.

 

Something like this may be possible if I think about a new configuration.

1 pool with 2 nvme disks with one spare (used for VM placement and application data)

1 pool with 2 ssd disks with no spare (used as cache)

1 pool with 4 hdd disks with one spare (as the main data pool)

So all write will go first to the SSD, which will be fast because it's configured as a cache for the data pool, and later on, it will save to the HDD. But I think the reads will be a "problem" because it keeps the data to individual disks and not spread over all disks. Or am I wrong here?

 

Which configuration would you recommend in my case?

[JorgeB]

3 hours ago, quei said:

So all write will go first to the SSD, which will be fast because it's configured as a cache for the data pool, and later on, it will save to the HDD. But I think the reads will be a "problem" because it keeps the data to individual disks and not spread over all disks. Or am I wrong here?

I assume you mean adding to the array for the hdd "pool"? If yes read/writes to the disks will be limited to a single disk speed, you could have those 4 disks as a seperate zfs pool, but will have the expansion issue and currently the mover does not support moving data from pool to pool, so it would need to be done manually or with a script.

[quei]

What other options do I have?

[JorgeB]

I already mentioned them above, use the array if single disk speed is OK, create a pool if not, but note that a zfs pool will be much less flexible for future expansion.