Hello Simon,
Firstly, welcome to Communities and using VMware products, pop in when you can!
Configuration:
As per your question regarding licenses, as one is based on cores and the other sockets you will have to do the math on that based on the requirements and cost of each with comparison of the possible configurations.
Maybe check out the prices of 3x800GB HDD vs 2x1.2TB HDD configuration as this will likely have better performance and less issues with multiple components of larger vmdk objects being forced to use the same disks.
Also, try to avoid SATA HDDs, they cannot keep up compared to SAS HDD drives.
Downtime:
Assuming you have a VMFS datastore on the existing standalone ESXi host, you should be able to Enhanced vMotion (storage and compute) from that local datastore to vsandatastore and the vSAN hosts(assuming Essentials Plus or higher licensing AFAIK) once they are managed by the same vCenter.
Alternatively, if you have some shared storage (SAN or NAS) you could migrate the VMs and/or back-up/restore/convert the physical boxes to here as a staging for migrating to vSAN (probably not necessary but depends on the situation).
vSAN implementation:
If 2 nodes with the specified compute resources will comfortably run the workload (low-ball this as in a node/disk-group failure scenario all will be running on the remaining host until the other is recovered/remediated) then a 2-node Direct-connect set-up would work.
Note that the Witness Appliance needs another ESXi cluster to run in, or alternatively I have seen some customers running them in the cloud when latency was within what is required (it is basically just a small VM that is added as a host to the vCenter in use by the vSAN).
Redundancy and failover:
In the simplest of terms, FTT=1 on an object (e.g. a vmdk - a hard disk from a VM) means that two copies (data components) of it are created, one is stored on each node (and a tiny witness component stored on the Witness), in the event of a failure of one copy (single-disk failure, disk-group failure or node failure) the VM can keep running from the remaining copy of the data component.
However in the case of a node failure (Hardware failure, Power issue, PSOD etc.) the VMs on this how will go down but can be restarted/re-registered right away on the remaining host manually (or using HA if using Essentials Plus or higher and set-up to do this).
I am aware that there is a substantial difference in cost between Essentials and Essentials Plus, there is also a lot of key features not in the latter (for this reason assumedly) and also the added benefit of support at a reasonable cost should you need/require to have it.
On Dell 730s the controllers work in pass-through mode, no R1 or R0 at the disk-level. This 'Raiding' is handled down the line by vSAN using Storage Policies as discussed above - e.g. FTT=1 applied to a VM or its disks is basically a R1 - striping for performance is also applicable (so yes can be essentially R10, but this is applied at the object level not at the controller/physical disk level).
I hope this helps clarify things and if nothing else to give you some insight into vSAN!
Bob
-o- If you found this comment useful please click the 'Helpful' button and/or select as 'Answer' if you consider it so, please ask follow-up questions if you have any -o-
