1) Idd, partitioning only applies to ESX.
2) In my opinion this is certainly a ‘must’ although I have seen virtual infrastructures with misaligned disks which are not affected due to the low load on the environment. In stressed environments this will have more affect and you should align the partitions to get the best IO performance.

Keep in mind that there are two locations/disk/volumes to align.

First of all you should align the datastore/VMFS. Just use the vSphere client to create your datastore, this automatically does it for you.

Second you should align the virtual machine partitions in the vmdk disk.

Thanks for the article.
Two questions though:

1.) The part of “partitioning” only applies to the use of ESX and not ESXi?
2.) During my search on partitioning I stumbled on a (few) posts on aligning
partitions, because it has an impact on performance.
http://communities.vmware.com/message/1326824#1326824
Is this a ‘must do’ ?

Thanx Jaap

Some best practices seem to be copied from other articles, might nice to add a link to the source.

1) You have a good head These were ESX3 numbers. Updated with vSphere now.
2) Added that. I prefer su because I’m not such a linux/SC wizkid.
3) I only use it in very small environments (2-3 ESX hosts) where it’s easy too keep it consistent. In larger environment there’s a physical DNS (and AD) most of the time.
4)I’m curious what you use as block size and on which arguments you base your decision. Block size is always a bit of a grey area. VMware states that block sizes do not have a performance penalty and VMFS uses sub channels so small file sizes are no issue either.
5)This is idd no ‘regular’ diagram. I used it to indicate that such setups are complex but it’s easy to see which connection is used for which function. In this case I did not use all nics but the customer wanted to use all to improve on redundancy and capacity. Because of that I added the spare nics to the management and VM network.

2) I would recommend to use SUDO instead of SU, as SUDO leaves the perfect audit trail

3) I never recommend a host file as it leads to inconsistency

4) Lun Size and Blocksizes, I wouldn’t recommend default based on those arguments.

5) The Networking Diagram doesn’t seem like a “regular” diagram with 3 nics for the management layer… why is that?

@J.R. Kalf: Thanks, I added a disk alignment section.

@Marius Redelinghuys: Thanks again. I added a hardware assistance and storage to the ESX(i) section.

With your additions this Best Pratices article is getting better by the day. Great community effort!