Brent Miszalski

Astronomer with research interests in late stages of binary stellar evolution and especially binary central stars of Planetary Nebulae. I developed the simulated annealing algorithm that optimises the target allocation for multi-object spectroscopy instrumentation of 2dF-like instruments. Currently working on virtual observatory services and data reduction pipeline automation in the Data Central team at Australian Astronomical Optics - Macquarie University.


Affiliation – Australian Astronomical Optics - Macquarie University Position – Astronomical Specialist Research Software Engineer Homepage – datacentral.org.au

Talks

Orchestration of Dockerized Data Reduction Pipelines from a RESTful Web Service

Data reduction pipelines are traditionally run on a researcher's personal computer on a small amount of data. The pipeline may have complex software dependencies that preclude the researcher from installing it on a faster server. Even if the pipeline runs on the server, its deployment in a multiprocessing environment may be problematic. Here we present a modern solution to allow for the on demand reduction of the ever-increasing volume of data stored in telescope archives. We have developed a Python web service that accepts 2dF-AAOmega observations and determines the steps needed to reduce the data. Each step runs 2dfdr commands from a Docker container on a fast server. We utilise docker-py to remotely execute these commands from within Celery tasks, allowing for a robust, configurable data reduction workflow to be assembled and executed asynchronously by Celery across several processors. Data Central plans to offer the service to users when requesting data from the newly revamped AAT archive, allowing effortless access to freshly reduced data. The service is extensible to other pipelines and would form a solid basis for developing IVOA SODA services, while slight modifications could unlock quick turnaround reductions of transient triggered observations.