We just had the ATLAS workflow management meeting with a round table going through our latest projects.
Already mentioned by Tadashi in presentation:
* Thoughts on metrics: how do we describe if a change in PanDA made things better or worse. Still in early stages of thinking how this could be and the machinery to collect and store metrics.
* R&Ds: Task requirement predictions through ML and tracking system error codes and messages. System errors should not count towards the user job retries.
* Submitted abstract to WLCG Environmental Sustainability Workshop about job CO2 estimations: https://indico.cern.ch/event/1450885/overview
* Planning to rewrite and modernise the PanDA server API
* Cleaner code and organization
* Json
* Documentation
* Versioning
* Data Carousel being implemented in PanDA: Instead of each task submitting independent staging requests, tasks with input on tape will go through a central machinery to make more controlled access. This should optimise and relieve tape systems from unorganised tape staging requests.