UKAEA work with adios to transport data between fusion reactor and CSD3. This is a prototype and very much a work in progress.
Dependencies: Adios2: Wands requires Adios2 to send one dimensional arrays. ornladios/ADIOS2#3545 fixed a bug that prevented using one dimensional arrays in python bindings of Adios2. To send data via wide area network, the Adios2's Dataman engine is used. If installed via spack both dataman and python need to be specified.
Installation via Spack (in an environment):
spack env create <name>
spack env activate <name>
spack add [email protected] +dataman +python
spack install
HDF5 and Boost are needed for for the CPP server and can, for example also be installed in the spack environment.
spack add hdf5+hl+cxx
spack add boost
spack install
To install wands it is recommended to create a virtual environment and use pip to install wands:
python -m venv <pip_env_name>
source <pip_env_name>/bin/activate
python -m pip install -e .
It is possible to build and run only the cpp server on the data site. This means that the last pip environment step ins't needed. To run all pytests on the client side both, cpp server and python wands need to be built.
To test the installation, run the test suite:
pytest
It is recommended to create a build directory in the server directory.
mkdir server/build
cd server/build
cmake ../
cmake --build .
./wands_server
CSD3 specific instructions to run the code:
- start an interactive session on CSD3:
salloc -t xx:xx:xx -Nx -ppnx -A <project-account> -p <partition> bash
- inquire the IP adress of the allocated node:
nslookup cpu-q-xxx.data.cluster
- a) If you want to send data from a remote location to CSD3. At the moment this is done via ssh tunneling (started on remote machine)
ssh -R 12345:localhost:12345 -o ProxyJump=<login-node> <CSD3ID>@<IP from nslookup>
b) Send data from CSD3 to remote machine also with an SSH tunnel (started on remote machine)
ssh -L 12345:localhost:12345 -o ProxyJump=<login-node> <CSD3ID>@<IP from nslookup>
cd <server/build> //or any other build directory chosen
./wands_server
from wands import Wands
data_source = "file.h5"
signals = ["group1/dataset1","group2/dataset2"]
local_cache = "path/to/local/directory" # where signals will be saved for future requests
wo = Wands(local_cache, Port = "12345")
data_dict = wo.request(data_source, signals)
#use data
Note: Setting up an SSH tunnel with proxy jump requires two different TOTP tokens.