This section explains the philosophy behind Bosun development and how we can fit a model to it.
Bosun tasks acts over data sources and sinks. Tasks can be divided in four categories: Prepare, Compile, Run and Archive. Data sources and sinks can be Experiments, Code, Artifacts and Storage.
In a very high level executing a model can be divided in four parts:
This step can include data manipulation (download, munging, copying) and directory creation.
Code checkout from repositories, instrumentation and compilation are good fits for this step.
How to run the model: batch system submission, consistency checks, status and automatic restarts.
Sources and sinks¶
Small tasks, composable tasks¶
Tasks should be small, and complex tasks should be broken in smaller ones.