Which is safer - disk pooling or disk spanning?

With either striping or spanning if one "disk" fails the volume is offline until that "disk" is reconnected so I would say they were equally reliable.

Performance may be a consideration, it certainly was 15 years ago where the same 2TB limit was imposed by Windows rather than a hypervisor.

For performance you have to consider the underlying hardware, if there are two separate disk arrays then striping gives a performance gain since to retrieve a big file half the blocks come from one array and half from the other whereas if two separate slices are used from a single array then half the blocks come from one area of disk and the other half from a different area of the same disk causing lots of long seeks. Thus striping can either make it faster or slower than before.

With spanning the whole file comes from either one or the other, so no performance gain but no possible slowing down either.

Bearing in mind the two scenarios above then if you can ensure separate underlying disk groups I would use striping but in every other case I would use spanning. Spanning can also concatenate different sizes whereas striping wants them to be the same size although that may be moot since you can use thin provisioning to fool it.

Maybe better consider a NAS/SAN storage array  with either raid 10   (stripe of mirrorsets) or raid 5/6  (1 or 2 redundant disk)  in a stripeset.

I had assumed they had several of them already.

Spanning as Andy pointed out provides no tolerance for a disk fault.

Pooling depends on what and how it ismanaged deals with how a"pooling disk" is defined.

Not sure I understand the 2TB Limit you mention.

Meaning disks inthe cluster storage are limited to the 2TB...

Striping isn't an issue. My question is about spanning vs disk pooling -
both using native tools to Win2k12. In either case, disks would come from
same LUN. As far as the 2tb limit, I am referring to ESX 5.1.

Disk pooling would either be mirrored, parity or *striped*. Since you want to compare it to spanning which gives you all the space with the redundancy built into the hardware I thought the only valid option of the three was striped. Maybe there's a fourth method, have they added ZFS/WAFL as a way of arranging the data?

If they are on the same LUN then spanning is the one to use for the same reason as my first post.

You are asking a question without the requisite context to what your issue seems to be or how you are trying to remedy it:
https://kb.vmware.com/s/article/2003813

A complete scenario of what it is you are trying to do and this way we would be in a position to address the issues you are dealing with in the scope of the process you are working towards.

I thought I covered everything pretty well in the question - even stating this
was 5.1 and not 5.5.

You didn't say what "disk pooling" product you were using but it was pretty obvious that it was Windows Storage Spaces.

Striping without knowing what you are doing is always unsafe.