VM host RAID

I would take 2 HDDs and create a hardware RAID 1 set (mirror) for the OS install.  Then manage the balance of the HDDs with Windows Storage Spaces.

Link:  https://technet.microsoft.com/en-us/library/hh831739(v=ws.11).aspx

Microsoft's take on designing for Hyper-V using Storage Spaces.

Link: https://technet.microsoft.com/en-us/library/dn554251(v=ws.11).aspx

An article that discusses a similar setup and a walk thru the configuration.

Link:  http://muegge.com/blog/hyper-v-cluster-with-storage-spaces/

Dan

Well, It mostly depends on the size of the drive you want to give to the virtual machines.
Since the 730P supports Raid 6, I'd create 2 Raids.
2 Disks as Raid 1 which will give you 300GB for the OS
and the rest as a Raid 6 which will give you 3,600 GB of Storage (about 3.5 Tera) with 2 disks failure tolerance which is preferred in a high availability infrastructure ESPECIALLY if it's going to be a production environment.

IMO, RAID 6 is a waste of hardware.  The probability of losing 2 HDDs in a single RAID container is quiet low.

In 24 years of IT, I've never had 2 HDDs in the same RAID container, fail, at the same time... either way, the container would still be in operation but without parity support.  You would need 3 HDDs to fail, at the same time, in the same R6 container to lose the volume.  Probability.... even lower.

A single server cannot be considered an HA setup.  The server is a single-point-of-failure.

But lets review:

 16x 300GB HDDs

1.  RAID10 container for everything
1a. RAID10 container with 16x 300GB HDDs, results in 2400GB usable space
--- you lose 50% of your space (2400GB)
--- that's 8 mirrors configured as a stripe.

2. RAID1 for OS, RAID10 for data storage
2a. RAID1 (300GB) for OS
2b. RAID10 with 14x 300GB HDDs, results in 2100GB usable space
--- this would be the fastest setup for the data storage volume.

3. RAID1 for OS , RAID6 for data
3a. RAID1 (300GB) for OS
3b.  RAID 6 with 14x 300GB HDDs, results in 3600GB usable space
--- you lose N-2 in space for R6 (600GB)

4.  RAID1 for OS , Windows Storage Spaces for data
4a. RAID1 (300GB) for OS
4b. Manage 14x 300GB HDDs as JBOD via Storage Spaces, results in 3600 to 3900GB usable space
--- create a parity pool (similar to R5/6)

Storage Spaces info:
1. https://technet.microsoft.com/en-us/library/hh831739(v=ws.11).aspx
2.  https://technet.microsoft.com/en-us/library/jj822938(v=ws.11).aspx

The advantage to the Storage Space setup is, mixed disk types, your can create tiered storage volume, mixing SSDs and HDDs for better performance.  Expansion of existing volumes is easier and quicker, HDDs failure recovery and rebuilds is faster than HW RAID due to parity being stored on all the disk, not just a single parity drive.

Dan

I have an EE article: Some Hyper-V Hardware and Software Best Practices.

We would set up either one RAID 6 array across all disks and partition 75GB for the OS and the balance ahead of time (via DiskPart in CMD after booting OS installer flash drive --> Repair) or two logical disks of the same setup.

I'm leaning towards performance and therefore I will put raid 10. What would be the cons for having one big disk to install OS and have a folder for VHDX's? This way I'd get the write performance of 8 disks and read performance of 16 disks.

Thanks. This is what I shall do.

With 8 or more spindles in RAID 6 with 1GB flash-backed cache the performance difference to RAID 10 is very little.

We do a lot of disk subsystem work for clusters. Our expectation is that 8x 10K SAS spindles in the above setup will yield at least 250 IOPS per disk at 128KB/256KB slice/block sizes and 450 IOPS per disk for 64KB slice/block size. Throughput for the former averages around 800MB/Second sustained write speeds.

Testing conditions:
RAID 6 array across all spindles with RAID on Chip, 1GB cache, and flash/non-volatile memory backing. Write-back is enabled.

+ Logical Disk 0: 75GB for host OS
+ Logical Disk 1: Balance GB/TB for Data and Testing