/RDataFrame-Totem

Primary LanguageC++

RDataFrame-Totem

How to run the analysis on Helix Nebula

  1. Log in to SWAN Helix Nebula

  2. On the CERNBOX tab, open a new terminal (icon >_ on the top right corner)

  3. Clone this repo:

    git clone https://github.com/JavierCVilla/RDataFrame-Totem.git

  4. Open the python notebook (DistillDistibution-AllDatasets.ipynb) from the SWAN Interface:

  5. Start the Spark cluster connection, the default configuration is ready to run the analysis

  6. Once connected, execute cells 1 to 7, this should be fairly fast since no computation will be triggered yet

  7. Cell number 8 initializes the Spark job and starts the event loop:

    • It may take some minutes for the creation of ranges
    • After a couple of minutes, you will see the Spark monitoring with the job progress

  8. Once finished, the rest of cells will show some results and save them to disk

How to run distill.py

Requirements

This script reads Totem data from eos, namely from the following path:

/eos/totem/data/cmstotem/2015/90m/Totem/Ntuple/version2/4495/

Therefore, the totem project needs to be mounted and accessible for the user.

Using pure python from a terminal

  1. Clone this repository:
git clone https://github.com/JavierCVilla/RDataFrame-Totem.git
  1. Prepare the environment
  • The code requires ROOT-6.14.00 or greater and Python.
  • Simplest way to fulfil this software dependencies is using the LCG Releases available through CVMFS.
  • The following command will setup your environment with these packages ready to be used:
source /cvmfs/sft.cern.ch/lcg/views/dev3python3/latest/x86_64-slc6-gcc62-opt/setup.sh
  • Alternatively, your own ROOT and Python installation can be used, in which case you should ensure the python ROOT module is properly configured in your environment so it can be imported:
python
>>> import ROOT
>>>
  • If the previous import failed, your PYTHONPATH may not be properly set. The easiest way to configure the environment for root is using its own setup script:
source /your/path/to/root/bin/thisroot.sh
  1. Run the code:
python distill.py <diagonal> [threads number]

Valid diagonals: d45b_56t, d45t_56b, ad45b_56b, ad45t_56t

Using HelixNebula

  • Init a session in Swan HelixNebula and select the bleeding edge software stack, this is the only one that currently provides ROOT-6.14.00.
  • Just copy distill.ipynb to your Cernbox space or to your SWAN instance in HelixNebula.
  • Open the python notebook and execute the cells.
  • HelixNebula already provides the needed environment configuration as well as access to the eos files.

Comparing results

ROOT files produced by this code and the original analysis can be compared using the rootcompare.c script to ensure the same results are produced.

After setting up the environment, compile the script using:

g++ -o rootcompare rootcompare.c `root-config --cflags --glibs`

This program receives 4 arguments:

./rootcompare fileA treenameA fileB treenameB

NOTE: This script is not meant to be generic enough to compare any pair of root files, currently it's aim to compare only files produced by this analysis.