I've read as much as I can find on Cormac and Duncan's blogs, but I still have a couple of questions about VSAN reads and writes:
Is there any integrity checking for reads? I haven't seen anything about error checking or other methods when data is read to make sure there hasn't been any corruption. I know that increasing FTT results in more copies being stored, but my understanding is that a block being read will only happen from one copy, and only relies on the 2nd+ copy in the event of a disk/disk group failure.
The information I've seen about writes is that 30% of the flash device is used for write buffer, and Cormac noted in his blog that the data is de-staged to HDD at regular intervals. Is there any kind of sizing information about how many HDDs are needed? For example, if I have 1000 desktops doing 10 write ops per second, do I need to have enough HDDs to satisfy 10000 write IOPS? Or does the write buffer do some sort of random->sequential translation so that I need less write IOPS?
Do all of the HDDs get used equally? Is it possible for a HDD in a disk group to get full before the others? I think I understand it correctly that unless striping is being used a VMDK will always write to the same HDD in the disk group, even if the disk group has multiple drives? Or even without striping turned on will the writes for a VM use all of the HDDs in a disk group?
Thank you for taking the time to answer my questions.
