/* here be dragons */

experiment/scripts contains the code used to run the experiments.

For running the timed benchmark, use

n_jobs_runner.py <number of jobs> <path to dir to monitor> <path to job>

this dispatches n jobs containing synced_workflow.py to the cluster (for
example monitoring/test.job). As of now, this should be run from the same
directory as <path to directory to monitor>, since the script in the job
checks that enough receipts from abel have shown up (ie the jobs have
started), so no workflow starts before all jobs (that contain the workflow)
have started on abel.

You can do a 

watch "python check_finished.py <path to dir to monitor>"

to see when the jobs are done, although the heuristic to do that is a little
crude, ie it checks that the slurm.out file is of the expected length (57,
hardcoded, sorry).

get_minutes_from_abel_receipt.py <path to dir to monitor>

gets you the time of the fastest finishing job, the average runtime of all
jobs and the slowest finishing job. The receipts from my experiments are in
experiment/speed_out.

running the disk-space benchmark:

check_growth.py <raw text file to run the workflow on>

which gets you the (incremental) growth of the db after each tool in the test
workflow. NB this uses the lap library and assumes access to pymongo, run it
with the python in the lap-tree!

Some trivial newplotting can be found in /experiments/plot

fin