3

I'm aware that VMware ESXi supports the creation of datastores on local Advanced Format 512e disks as of v.6.5. However, all the (very scarce) information I can find seems to suggest that virtual disks created on that datastore will still be exposed to the virtual machine as 512n disks.

For some workloads, performance could be seriously degraded if the guest OS believes the disk is natively 512 byte sectored, producing lots of read-modify-write operations.

Why am I not hearing anything about this? Perhaps my information is incorrect and it exposes 512e to the guest? Or is there a setting for whether the guest will see 512n/512e/4kn?

Kevin
  • 1,580
  • 4
  • 23
  • 35

1 Answers1

3

VMDK will be still exposed as 512n in any case since it is the way how the VMware vSphere works. However, starting from VMware 6.5 you can make RDM of the 512e disk directly into VM. Moreover, your RAID controller must support those 512e and/or 4k disks and either emulate 512n or path-through all the sector size into the VMware hypervisor. (Please note that 4k disk support was added in vSphere 6.7)

Stuka
  • 5,445
  • 14
  • 13
  • 1
    Thanks for confirming that. That to me seems to be a pretty big flaw. It's too bad you can't at least specify the physical sector size to expose; better yet, the correct physical size should be reported to the guest OS automatically. I doubt it'd be _that_ hard to implement for 512e, and it could avoid large performance issues on some guests. I can see it would be slightly trickier if the datastore is backed by multiple physical disks, but even then, they could just report the largest sector size of any of the disks. Oh, well. – Kevin Feb 07 '19 at 23:08