Geoscience Australia (GA) stores large quantities of satellite data at the National Computer Infrastructure (NCI) facilities located at ANU in Canberra. From these data, GA produces a number of data products that are made available to customers via various publishing mechanisms.
The main computer at NCI is Raijin, a batch-processing super computer. The GA product production line employs Raijin and the Portable Batch System (also know as PBS) to submit batch jobs for processing on Raijin.
The purpose of this document is to give new starters at GA a quick introduction in how to write PBS jobs to run on Raijin. I've provided some sample code and some tips on writing code and preparing jobs scripts.
Here is a glossary of terms to help get started with PBS. You should pay careful attention to the anti-terms otherwise there is potential to waste time looking for PBS features that are not implemented avaialble on Raijin.
NCI (and others) provide some excellent online documentation. Here is a list of preliminary reading for new starters.
Here is a collection of design tips to help develop code to run on Raijin under PBS.
Read about different workload distribution stratgies used by GA.
Example code has been prepare to demonstrate some of the batch processing techniques used by GA
TODO