Provenance Tipping Point

Capture is a known, difficult problem for provenance. Obtaining from the systems and programs exactly what happened has been a continuing struggle outside of database and workflow systems. The provenance research community has created libraries to log provenance, and has also embedded instances of capture agents within operating systems, specific programs, etc. However, it is impossible to know if we are inserting capture agents at both the optimal location and frequency in a given system for a high quality provenance graph. In this work, we develop an initial agent based model to simulate Activity and Entity interactions in a complex system of software. Using this model, we can attempt to define some generalized principles about type, frequency and distribution of provenance capture agents given a new system.