Arvados

Things we would like to know at the end of every pipeline:

Did we utilize nodes correctly? How much total CPU time did we use? crunchstat_summary is a great method for this, but what if we automated that in one job at the end of every pipeline?

Other things could be:
What percentage of reads did we map correctly?
Whats the average size of each read mapped?
How much total output data did we make?
How much did this run cost?
How big was the input data?
How many nodes did we use at each step?
What was that graph that we used to capture real time crunchstat usage?

There's a lot of questions we can brainstorm and have in this one report. This is an example template report we could try to emulate.

https://cloud.curoverse.com/collections/qr1hi-4zz18-4z27d9yelk8shv3 (click on magnifying glass)