csaxs_bec/docs/user/data_analysis/computing.md

(user.data_analysis.computing)=

Online data analysis

We can access the online computing resources from the beamline consoles when logged in as the e-account.

Allocation of nodes

Check available nodes and allocate one:

sinfo
salloc -A csaxs -p csaxs -w [ra-c-110]
ssh -Y [ra-c-110]

Or for a GPU node:

salloc -A csaxs -p gpu-csaxs --gpus=4 [-w ra-gpu-003] --mem=100G -n 8
ssh -Y [ra-gpu-003]

(note that we need to ask them to explicitly use one of our GPU nodes there, and this needs a couple of days to set up)

Specifications of the computing nodes

A few examples of the available nodes:

Node	CPU	Cores	Memory	GPUs	Partition
ra-c-110	Intel Xeon 6152	44	377 GB	—	csaxs
ra-gpu-003	AMD EPYC 7452	64	1007 GB	4x NVIDIA A100 (40 GB)	gpu-csaxs
ra-gpu-007	Intel Xeon 6530	64	1007 GB	4x NVIDIA L40S (46 GB)	gpu-csaxs

The home directory is the p group. To go to the raw directory we need to navigate there:

cd /sls/x12sa/data/[p22598]/raw

Python tools

Python package installation

Anaconda is no longer supported. Packages should now be installed directly using pip.

For example, pyFAI can be installed as follows:

module load Python/3.11.11
pip install "pyFAI[gui]"

To launch the pyFAI applications:

# Detector calibration
python -m pyFAI.app.calib2

# Batch integration
python -m pyFAI.app.integrate

Recommended: using a Python virtual environment

For long-term maintainability and reproducibility, it is recommended to use a dedicated Python virtual environment.

Create a virtual environment

cd /sls/x12sa/data/[p-group]/raw

module load Python/3.11.11
mkdir python-env
cd python-env

# Create a virtual environment named ".venv"
python3 -m venv .venv

# Activate the environment
source .venv/bin/activate

# Install required packages
pip install "pyFAI[gui]"

Use an existing virtual environment

cd /sls/x12sa/data/[p-group]/raw/python-env
source .venv/bin/activate

Troubleshooting

If you encounter the following error when launching GUI applications (e.g. pyFAI-calib2):

X11 connection rejected

reconnect to the analysis node with X11 forwarding enabled:

ssh -X [ra-c-110]

and then reactivate the virtual environment before launching the application.

Use Jupyter Notebook / Lab

Create a virtual environment following the instructions above, then:

pip install jupyter

To run Jupyter Lab:

jupyter lab --port 6006 --ip $(hostname) --no-browser

cSAXS matlab scripts

To use the cSAXS matlab scripts we can go to the cxs_software package in the raw folder.

In base we can do:

source setup-environment.sh 

And then we can use Matlab interactively as usual.

matlab &