rajibhossen / workload-driven-autoscaling-ensemble-cluster24

Scripts and Datasets for Cluster 2024 Paper - "Enabling Workload-Driven Elasticity in MPI-based Ensembles"
0 stars 1 forks source link

Repository Organizational Suggestions #1

Open vsoch opened 3 months ago

vsoch commented 3 months ago

OK I think I'm going to stop - here is what we can do. Start with the organization described above in the main README. Then one at a time, walk through the instructions you are providing to the used in the subdirectory README. Every file should be there. Instead of having README-.md (that is hard to read) have a structure for an experiment directory that looks like the following:

experiment-name
   strategy
       applicationA
           README.md
       applicationB/
           README.md

Every file should be scoped to exactly where it was used, and anything mentioned in the README should be there. If you need to copy files (and have redundancy) that's fine. For example, when you said you manually changed values in an eksctl config? Actually create each of those files for the reader.

We will likely need a few rounds of changes.

rajibhossen commented 3 months ago

Further reorganization suggestions by @vsoch

setup/
   cluster-autoscaler/
   custom-metrics/
experimentA/
   ....
experimentA/
    strategy/
      applicationA/
        README.md
      applicationB/
        README.md

With scaling by kubescaler.

setup/
  cluster-autoscaler/
  custom-metrics/
experiments/
  scaling/
    cluster-creation/
    cluster-deletion/
  auto-scaling/
    experiment-name/
       strategy/
         appA/
         appB/

In the readme.

# Name of Experiment

Summary of experiment

1. Cluster Creation

eksctl command stuff

2. Setup Autoscaler

You'll need to setup the autoscaler, which is this thing (bla bla). See [the instructions here](../../../setup) for how to do that.