Hi ... on the off chance that anyone is following this thread or comes across it on a google search I wanted to post a quick followup.
Did a proof of concept consisting of:
- "Performance tier" of 2x Intel S3500 SSDs, directly connected over SATA
- "Capacity tier" of 2x HP HDD, configured as a single RAID1 logical volume behind the LSI raid controller on a Supermicro X9DRD board
The MediaType of the LSI Raid logical volume was set to HDD using the Set-PhysicalDisk command.
On the Performance tier, the ResiliencySettingName was set to Mirror
On the Capacity tier, the ResiliencySettingName was set to Simple (which it had to be, since Windows only sees the one logical drive)
A windows volume was created via New-Volume, using tiers, with ReFS as the filesystem.
Performance on CrystalDiskMark on this new volume was consistent with the performance of 2x S3500 SSDs, suggesting that things were more or less working as expected. I have not yet done any more detailed testing / benchmarking but was sufficiently encouraged that I am moving ahead with this approach at a somewhat larger scale with NVMe drives as the Performance tier and a Raid 6 array of HDDs as the Capacity tier. Happy to post updates if anyone is interested.
Did a proof of concept consisting of:
- "Performance tier" of 2x Intel S3500 SSDs, directly connected over SATA
- "Capacity tier" of 2x HP HDD, configured as a single RAID1 logical volume behind the LSI raid controller on a Supermicro X9DRD board
The MediaType of the LSI Raid logical volume was set to HDD using the Set-PhysicalDisk command.
On the Performance tier, the ResiliencySettingName was set to Mirror
On the Capacity tier, the ResiliencySettingName was set to Simple (which it had to be, since Windows only sees the one logical drive)
A windows volume was created via New-Volume, using tiers, with ReFS as the filesystem.
Performance on CrystalDiskMark on this new volume was consistent with the performance of 2x S3500 SSDs, suggesting that things were more or less working as expected. I have not yet done any more detailed testing / benchmarking but was sufficiently encouraged that I am moving ahead with this approach at a somewhat larger scale with NVMe drives as the Performance tier and a Raid 6 array of HDDs as the Capacity tier. Happy to post updates if anyone is interested.