vSphere Storage Appliance

Hi,

I've seen some discussions online in relation to typical block sizes for VDI implementations, with a block size of 4KB being quoted as the most common and average block sizes being somewhat larger e.g. 10KB. It isn't surprising for 4KB to crop up since this is (I think) the block size used for all NTFS disks greater than a couple of GBs. Many / most VDI storage infrastructures are on VMFS storage, so taking the block size for VMFS5, we get a VMFS block size of 1MB, sub-block size of 8KB. The storage array will also have an associated block size, so I'm confused about what are the actual inputs to a calculation for (e.g.) average VDI block size? Are the inputs block-sizes at the guest filesystem level, the VMFS level or the storage array block-size level? And, whichever it is, surely the average VDI block size would be exactly the guest FS / VMFS / array block size i.e. no need for any averaging?

I'm probably missing out on a fundamental concept here so any clarifications / explanations welcome.

Thanks,

number1vspherefan.

Fortunately, its much easier than all that.

The VMFS blocksize is entirely irrelevant to the guest.  Guest reads and writes are passed through directly to the storage, and so VMFS block size doesn't come into it.  If your guest produces 4KB IOs, then VMware produces 4KB IOs.

When doing sizing, most people would stick with the 4KB number you mentioned.