Whiteboard diagram

Outline

The plan for automating Pre-processing has been refined further, with the notes presented on the whiteboard.

In addition verbal decisions were made regarding how event level tasks will be performed with relation to realisation level tasks. Bearing in mind that Validation and Cybershake runs have different scenario naming schemes.

This is independent of the proposed Data Structure reformat

The main changes to the workflow are that event level tasks will be added to the slurm management database.

Each realisation will query its own tasks, and those of its corresponding event for dependencies. For validation events this will mean it will check its own

This will be done in such a way that validation runs will have the same name for the event level tasks and the realisation level tasks.

New fault/event level tasks

The new fault/event level tasks will be as follows:

Median srf generation

VM params generation

Velocity model generation

Fault/event level installation

New realisation level tasks

The new realisation level tasks will be as follows:

Realisation srf generation

Velocity model perturbation generation

A step to indicate no velocity model perturbations will be generated (This allows for dependencies to work as expected. The alternative is for the VM perturbation generation to have a short circuit somehow)

Realisation level installation

Exisiting task changes

Furthermore the tasks EMOD3D and HF will have dependencies on the event/fault and realisation installations.

Steps such as rrups and empiricals will also be changed to operate on the event level, rather than the realisation level.

Task flow chart

Phase 1

  • Automate VM Parameter, VM Generation, VM Perturbation and partial Fault Install.
  • Having two separate default task configurations - one for pre-processing, one for main steps.

This means that you can generate SRFs, install and then enter into the automated workflow

(Caution: If using perturbations you will have to re-install your CS workflow after completing your pre-processing steps - this regenerates the sim_params to include perturbations (your database should remain intact))

or

Enter in at the LF stage after installing (as normal)

Phase 2

  • Refactor Install into:
    • root_install
      • DB Generation + Root Params
    • fault_install
      • FD station lists + Fault Params
    • sim_install
      • Sim Params
  • Introduce Fault level tasks (tasks can either be on a realisation level or a fault level - for validation (single realisations) a fault task and a realisation task will look identical)

Phase 3

  • Data folder merge. – Folder/Data structure refactor.
  • Script to copy an existing CS run up to a specific step.
  • No labels