This directory contains scripts that are related to the generation of workloads used for the Batsim paper accepted at JSSPP 2016.
The different subdirectories are:
- generated_workload: The workloads that have been generated for the experiments are placed in this directory.
- NPB_bin: Contains the binary files we were able to compile from the NAS Parallel Benchmarks.
- job_calibration: Contains file related to the calibration of the MSG profiles, to make their execution times match the ones on the Graphene cluster.
- tit_to_profile_validation: Contains a little MPI C++ example. This has been useful to create the script tit2batprof.py
All (or almost all) scripts have their own help page, which can be shown just by calling the script with the --help option. The different scripts in this folder are:
- json_workload_generator.py: The main script of this folder, in charge of generating a Batsim JSON workload input from a profile list. An input example of this script can be found in profiles_example.json
- Different wrapper scripts that call json_workload_generator.py. They are used to store the parameters of the call.
- generate_batsim_workload_from_oar_result.py: This script allows to create a Batsim JSON input workload file from an OAR gantt JSON file. The resulting file will contain jobs with delay profiles.
- translate_submission_times.py: This script modifies a Batsim JSON workload file. It translates the submission times of jobs such that the first submission will be done at time 0.
- create_noisy_workload.py: This script reads a Batsim workload file, adds some noise on the submission time and the execution time of jobs (according to some parameters), then writes the result in another Batsim workload file.
- tit2batprof.py: This script reads a SimGrid time-indepent trace (TIT) file which corresponds to a single job execution and transforms it into a Batsim profile, which is added into a Batsim profiles JSON file.
- prv2tit.pl: A script that transforms a PRV file to a SimGrid TIT file, courtesy of Lucas Schnorr. More information can be found on his labbook.
- mpi_bench.py: This script generate the MPI traces using the NAS benchmarks and the Extrae tracing tool.
- extract_tit_trace.sh TODO