Sample layout for a simulation with a grid of [nx,ny,nz] grid points per processor.
There are (px,py,pz) processors along the x, y, and z direction.
The number of ghost layers is set here to NG and is of course variable in reality.
This results in a global grid size of mxgrid=nx*px+2*NG in a monolithic layout.
Likewise, the global grid size without grid cells is defined as nxgrid=nx*px.
The chunk dimensions are denoted in square brackets.
The group and dataset structure and dimensions would be as follows:


The main idea in the collect-xy variant is to first combine the data along the x and y directions
via MPI. This is done on all processors in parallel and it is an independent communication.
In small setups with px=1 there would be only communication along y with py processors.

We can include the lower and upper x ghost layers in the collection along x and y,
as this is only a relatively small increase of the communicated data for boundary processors:

data/
	ax	[mxgrid,mygrid,nz],(pz)
	ay	[mxgrid,mygrid,nz],(pz)
	az	[mxgrid,mygrid,nz],(pz)
	lnTT	[mxgrid,mygrid,nz],(pz)
	lnrho	[mxgrid,mygrid,nz],(pz)
	ux	[mxgrid,mygrid,nz],(pz)
	uy	[mxgrid,mygrid,nz],(pz)
	uz	[mxgrid,mygrid,nz],(pz)
ghost/
	z	=> see "chunked" scheme, but only for the z layer


The alternative is not to include x and y layers in the collection:

data/
	ax	[nxgrid,nygrid,nz],(pz)
	ay	[nxgrid,nygrid,nz],(pz)
	az	[nxgrid,nygrid,nz],(pz)
	lnTT	[nxgrid,nygrid,nz],(pz)
	lnrho	[nxgrid,nygrid,nz],(pz)
	ux	[nxgrid,nygrid,nz],(pz)
	uy	[nxgrid,nygrid,nz],(pz)
	uz	[nxgrid,nygrid,nz],(pz)
ghost/
	x,y,z	=> see "chunked" scheme


In summary, one must note that the actual writing (and accessing of the HDF5 library)
is done here from only pz processors, therefore chunks can be much larger.
The previous collection step might be a bottleneck, because only one processor
receives the data from px×py processors, which has to be a sequential communication!
Only after the collection step has finished, the writing via HDF5 calls may begin.

I wonder if this scheme can at all perform better than the "collect-x" scheme.