By default our storage array uses 32K block size for LUNs provisioned for virtualization and they've been formatted for VMFS6.
Are there any caveats when it comes to the allocation unit size specified for an MBR or GPT NTFS volume (on VMDK)? I know Windows default is 4K, but didn’t know if there’s any advantage to using a higher allocation unit size. In this instance, the guest OS is Windows Server 2019 and VM function is to support CIFS/SMB file services. Thank you.
The guest OS is unaware of the underlying storage so the principles for choosing a block size remain the same as for physical machines. I always use the default for regular VMs and 64k for SQL server VMs.
The guest OS is unaware of the underlying storage so the principles for choosing a block size remain the same as for physical machines. I always use the default for regular VMs and 64k for SQL server VMs.
So, you just configure as if they were physical and use the default (4K), unless it's for SQL, usually recommended 64K for Backup, Data, Logs, TempDB?
Yep! But on the other hand, I'm not a storage expert
I read that 4K is recommended to accommodate small files and I do have those on the volumes supporting those CIFS/SMB shares, so that reinforces your response. As mentioned, I don't think we've ever deviated from default, except to support Oracle and SQL databases on VMs, but I wanted to ensure we followed best practice for these new VMs that would be supporting quite a bit of unstructured data.
From my experience with recovery of VMDKs I learned that nobody ever regrets going with the 4k default.
Many of those guys that tried larger sizes would not use it again.
Just my 2 cents
Vielen Dank!