vSphere 6.5 has also introduced a new filesystem, for block-based storage—vSphere VMFS 6. It adds new capabilities as compared to previous filesystems:
|
Feature |
VMFS-3 |
VMFS-5 |
VMFS-6 |
|
Supported ESXi versions |
All |
5.x and 6.x |
6.5 |
|
VMFS datastore up to 64 TB |
Yes |
Yes |
Yes |
|
VMDK larger than 2 TB |
No |
Yes |
Yes |
|
Unified block size (1 MB) |
No |
Yes |
Yes |
|
Atomic Test and Set (ATS) enhancements |
No |
Yes |
Yes |
|
Sub-blocks for space efficiency |
64 KB |
8 KB |
64 KB (dynamic) |
|
Small file support |
No |
1 KB |
1 KB |
|
Disk partitions type |
master boot record (MBR) |
GUID partition table (GPT) by default |
GPT |
|
Physical block size |
512n |
512n |
512n or 512e |
|
Automatic space reclamation (UNMAP) |
No |
No (manually) |
Yes |
VMFS 6 introduces two new internal block size concepts for file creation—Large File Block (LFB) with a size of 512 MB and Small File Blocks (SFB) with a size of 1 MB and these are used to back files on the VMFS 6 volume. Note that the VMFS block size remains 1 MB sized. Thin disks are backed by SFBs. Eager Zeroed Thick (EZT) or Lazy Zeroed Thick (LZT) disks are backed by LFBs as much as possible; SFBs are used for the portion of the disk that does not fit into an LFB. For more information, see http://cormachogan.com/2017/08/16/vmfs-6-large-small-file-blocks/.
In VMFS 6, most of the datastore management tasks remain the same as VMFS 5, such as increasing the capacity of a datastore, resignaturing a datastore, managing the pointer block cache, and checking metadata consistency with vSphere On-disk Metadata Analyzer (VOMA) (for more information see the Storage Guide available on https://storagehub.vmware.com/).