### How to import a mesh into Microsoft Azure notebook

313

views

0

I'm working with the demo at https://notebooks.azure.com/garth-wells/libraries/FEniCS. I would like to import a mesh that is on the hard drive into the notebook. Is it possible to do this?

The mesh is in h5 format. So currently, you can run a script and import the mesh using:

f = HDF5File(mpi_comm_world(), 'mesh.h5', 'r')

mesh = Mesh()

f.read(mesh, 'mesh', False)

I'd like to do this in an Azure notebook. Any help doing this would be greatly appreciated.

Thank you,

The mesh is in h5 format. So currently, you can run a script and import the mesh using:

f = HDF5File(mpi_comm_world(), 'mesh.h5', 'r')

mesh = Mesh()

f.read(mesh, 'mesh', False)

I'd like to do this in an Azure notebook. Any help doing this would be greatly appreciated.

Thank you,

Community: FEniCS Project

### 1 Answer

2

It should be possible - although of course, Azure Notebooks are not very powerful, so don't expect to solve huge problems.

There is a "data" dropdown menu when you are in a notebook - you should be able to use that to upload a "h5" file (and any other file)

There is a "data" dropdown menu when you are in a notebook - you should be able to use that to upload a "h5" file (and any other file)

What's the equation you're solving and which solver are you applying to the linear system? That seems very efficient being able to solve a system that large and that quickly on a laptop.

written
11 months ago by
Nate

I'm solving the advection diffusion equation, with velocities computed from the Navier Stokes equation. The solver for the linear system is GMRES, with the ILU pre-conditioner. I need to run about 30,000 iterations, each time step size is 0.001 second

written
11 months ago by
Sophia Wright

Please login to add an answer/comment or follow this question.

Are there any cloud computing options that run FEniCS that are more powerful than Azure? I would like to import a 3D mesh with 2 million cells, that forms a 9.4 million dof linear system, and my laptop takes 4 min per iteration, which is too long