PROVENANCE ANALYSIS WITHIN A.N.T.A.R.E.S.

Abstract

ANTARES, a software system for categorizing astronomic phenomena, employs a non-static pipeline. For scientific legitimacy, metadata needs to be recorded for each categorization. However, the pipeline processes up to fifty thousand alerts every thirtynine seconds. As a result, there are limited resources for recording metadata in realtime. By allowing some overhead at startup, environmental metadata can be recorded, including what version of the pipeline was used. This recording can be achieved before processing has begun. Post-hoc analysis can be used to infer the rest of the needed metadata. Provenance, the computational history of the categorization of an alert, is entirely dependent on this metadata. DAFCAA, the Dynamic Alert Flow Computation Analyzer on ANTARES, is an interactive application used to assist in analyzing provenance post hoc. This application allows a user to view what configuration was used during a specified time period by retrieving source code from GitHub. By reading values queried from the database of alerts, the user can view the results of specific computations and therefore, alert classifications.