Results

Presentation of result files

Since we deploy the ~mad-node~ interactively, every file are stored in your $HOME directory and should be available even when the reservation is finished.

First, global information can be found about the benchmark. More particularly, check the file /MadBench_<date>/results which contains the deployment time for each test. This file is filled as the benchmark runs (so it should be empty at the beginning). Also, the output of the benchmark tool is recorded in /MadBench_<date>/stdout+stderr.

You can use a script to plot figures from the global result file by typing:

(venv) user@mad-node:~/mad$ ./plot_results_perf_nova.py MadBench_*/results

Depending on your test, you might need to adapt the plotting script.

Furthermore, as the benchmark runs, some output files should appear in the current directory. These files contain detailed information used to better understand the previous results about the experiment. Particularly you should see three types of file:

• result_<test>_<tool>_<registry>_<iteration>_components.txt: time spent for each component;
• result_<test>_<tool>_<registry>_<iteration>_tasks.txt: time spent per task;
• result_<test>_<tool>_<registry>_<iteration>_gantt.html: Gantt diagram of the run.

Backup and restore

You can either move your results to a backup archive by calling:

(venv) user@mad-node:~/mad$: make tgz_backup

or copy them to an archive by calling:

(venv) user@mad-node:~/mad$: make tgz_snapshot

Both command should output the path to the archive. You can use this information to easily copy the archive on your own computer or to share it with someone:

user@laptop:~$ scp lyon.g5k:/<path_from_the_output> ./archive.tgz
user@laptop:~$ mkdir result_expe
user@laptop:~$ tar --force-local -xvzf archive.tgz -C result_expe