ATLAS Trigger Simulation with Legacy Code using Virtualization Techniques

For high-quality data analysis a realistic data simulation is necessary. Thereby simulating the trigger response accurately presents the single largest challenge: while for event generation and simulated data reconstruction the latest software release is used, the trigger simulation needs to be done with the software and conditions that were used during data taking, often many years back. Therefore, having a strategy for running old trigger software becomes important, in particular when data from past years start to present a sizable fraction. We examined the requirements and possibilities for such a simulation scheme within and beyond the existing ATLAS software framework and successfully implemented a proof-of-concept simulation chain. In order to gain independence from the underlying operating system when running the archived code base unmodified, we tried several virtualization technologies. This use of virtualization also presents a first step into the study of data preservation. In this talk we will present the problems we discovered during this work, like data format compatibility, virtualization choices, multi-release simulation job setup, and the implemented solutions.