Possible bad advice in [ID 1213714.1] Sun ZFS Storage Appliance: Performance clues and consideration

The referenced support document [ID 1213714.1] has the following statement I believe is untrue:

It is very important to match the client blocksize with the filesystem recordsize or volume blocksize. For example, the minimum blocksize for vmware is 1MB, hence the volblocksize for iSCSI LUNs should be 128KByte (the max) instead of 8KBytes (the default setting).

On VMFS there is a 1MB blocksize. (Up to 8MB prior to VMFS5). But guest IO is not mapped to VMFS blocks; it is essentially passed through directly. Thus a 4kB read on a guest VM will appear to the storage as a 4kB read. Thus zvols that will contain VMFS should be tuned to the workload of the guest. For most filesystems that is normally going to mean 4kB blocks. To realize that this is true simply imagine the amount of throughput that would be required to support a random workload on top of VMFS3 with 8MB blocks if