[WIP] import experiment-base

[lgarithm]

• migrate k8s cluster tools from base experiments-base
• cleanup README docs to only refer to one virtual environment terminal
• fix broken figure links
• try more experiments to make sure they are still working
  • kernels_mpi (Fig.9b)
  • kernels_omp (Fig.10)
  • lammps (Fig.9a) (stuck on 7 MPI processes with workload network, not finish in 2 hours, tried 2 times)
  • elastic (Fig.12)
  • lulesh
  • makespan

Traceback (most recent call last):
  File "/usr/lib/python3.10/multiprocessing/process.py", line 314, in _bootstrap
    self.run()
  File "/usr/lib/python3.10/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/home/lg/code/repos/github.com/lgarithm/granny-experiments/tasks/makespan/scheduler.py", line 353, in thread_pool_thread
    run_kubectl_cmd("makespan", exec_cmd)
UnboundLocalError: local variable 'exec_cmd' referenced before assignment

• migration (Fig.11)
• motivation
• openmpi
• polybench

[csegarragonz]

hey @lgarithm could you also please update the readme so that the links to the plots also contain a reference to the figure number in the paper?

many thanks!!

[csegarragonz]

hey @lgarithm many thanks for pushing on this, it's definately getting there!!

i would say we don't need to re-run all the plots, just make sure that each experiment runs for some cluster size.
you can discard the changes in the data files so we keep the old results.

[lgarithm]

What do you mean by the old results? Currently this repo completely ignores the results and the plots folder.

i would say we don't need to re-run all the plots, just make sure that each experiment runs for some cluster size. you can discard the changes in the data files so we keep the old results.

[csegarragonz]

What do you mean by the old results? Currently this repo completely ignores the results and the plots folder.

That is true, maybe let's keep it that way. I may try to commit the results and the plots at some point later.

[lgarithm]

The network workload of LAMMPS always got stuck at 7 MPI processes:

Running LAMMPS on Granny with 7 MPI processes (workload: network, run: 1/1)

happened for multiple times.