One feature of vSphere (and the preceding VI3 and other platforms suite) that particularly interests me is the VMware vStorage VMFS file system (or just simply VMFS). VMFS is a purpose-built clustered file system for virtual machines. A little background: A VMFS datastore is available for block-based storage protocols (iSCSI and fibre channel) and is where virtual machines can reside within standalone ESXi hosts as well as ESXi clusters. Figure A is a general representation of this arrangement for a 3 host cluster.
vSphere 5 introduces VMFS 5, which is an upgrade from VMFS 3 used in vSphere 4.x and VI3. The main change with VMFS 5 is the unified block size, which is 1 MB. VMFS 3 was able to format at 1, 2, 4, or 8 MB (in a recent tip, I recommended formatting all VMFS 3 volumes at 8 MB). With VMFS 5 supporting 1 MB block sizes, the maximum sizes for Virtual Machine Disk Formats (VMDKs) are not limited like previous block sizes. This is great because too many times a volume would be inadvertently formatted at 1 MB, and a VMDK larger than 256 GB would not be supported. VMFS 3 datastores can be upgraded, and retain their block size. However, it’s advisable to reformat the volume to VMFS 5 at 1 MB (the default size) at this point to be most in line with VAAI and other upcoming features.
Another important change is related to the sub-block algorithm allocation. VMFS implements a unique sub-block algorithm that works well for the polar distribution of file types: large VMDKs and small VMX and others. The previous sub-block was 64 KB within the parent 1, 2, 4, or 8 MB large block format. In VMFS 5, this has been reduced to 8 KB.
The most important update to the VMFS 5 file system is support for sizes up to 64 TB; the previous limit was 2 TB LUN. This requires that the storage processor can provision storage at these levels.
Which new features in VMFS 5 interest you? Let us know in the discussion.
Subscribe to the Data Insider Newsletter
Learn the latest news and best practices about data science, big data analytics, artificial intelligence, data security, and more. Delivered Mondays and Thursdays