Analysis AWS

From MEG Core
Revision as of 09:57, 1 November 2019 by Jstout (talk | contribs)
Jump to navigation Jump to search

For Workshop members without access to computing resources - Preconfigured Amazon Web Services servers can be used

Accessing compute resources on AWS:

Speak with Jeff Stout to get the current IP address and PEM key
${PATH} is the path to the PEM key provided to you

 ssh -X -i ${PATH}/AWS.pem ubuntu@IP_ADDRESS_PROVIDED

Jupyter Notebook Over a Remote Connection

1) Run one terminal to start the notebook:

 #If necessary: conda activate $CONDA_ENV
 jupyter notebook --no-browser --port=8887

2) Run another terminal to forward the port to your local machine
(Replace User and Remote_Host with the appropriate username and ip-address)

 ssh -i ${PATH}/AWS.pem -L 8887:localhost:8887 ubuntu@IP_ADDRESS_PROVIDED
 #ssh -N -L localhost:8888:localhost:8887 $USER@$REMOTE_HOST

3) In Firefox/Chrome/Etc - Log into localhost:8888
This will prompt you for a token

4) In the password/token column, copy the token created in the first terminal
(It should look like the below - only copy the portion after the token= )


Installed software

Human Neocortical Neurosolver

 #hnn is installed in the system level python >> may need to conda deactivate first

MNE python, Eelbrain, PyCTF

 conda activate workshop  (This environment has MNE, Eelbrain, and PyCtf installed)

CTF software

 singularity shell /opt/ctf/ctf.img
 singularity exec /opt/ctf/ctf.img $ctf_command

SAM version 5

 sam_cov ...
 sam_3d ...


 afni -dset $MRI_file

  1. Server Config
    1. google > fedora aws > select the AMI number for N.Virginia
    2. In AWS > launch instance > search for AMI number > (Under community) > Launch
  2. Determine an image (search for fedora in community)
  3. Create a keypair and download the key
    1. The download will be in the form of a .pem file
    2. We will need to distribute this key to the users
  4. Launch the instance
  5. Log into the instance using ssh

  1. Freesurfer
  2. MNE python
  3. pyctf

Done # SAM v5

  1. Afni ###INSTALL ME neurodebian
  2. singularity ? >> CTF (need to use workstation version of fedora) << may need to compile for ubuntu or use 2.6
  3. Eelbrain (conda forge)
  4. Jupyter

Done# Forward sim HNN software

  1. FSL ###Install outside of neurodebian
  2. Brainstorm (compiled)

Matlab Based

  1. Brainstorm
  2. Fieldtrip