MPD Data Lab: Towards the Modern Data Analysis Framework for the MPD Experiment

Abstract

MPDRoot is an off-line software framework for simulation, reconstruction, and physical analysis of the simulated or experimental data for MPD experiment at NICA collider. The experiment is projected to run for a few decades and to obtain ~10⁸ events of heavy ion collisions, collecting the data for physics analysis at the 100 PB scale. For overall experiment success it is imperative to have state of the art data analysis software, which integrates best of available latest technologies, while adhering to time-proven, most effective development methodologies. In this paper, we introduce the MPD Data Lab—the technological integration of Acceptance Test Driven Development and Rapid Development concepts into the MPDRoot framework. At the beginning, we standardized the existing codebase by designing and writing API. This was a necessary step to be able to plug-in the external diagnostic software entities and to make the in-depth comparison of different realizations of the reconstruction modules possible. The logic of the diagnostics is encapsulated into the separate controller—the QA Engine, while its visualization is provided by JupyterLab framework. We show how full integration of MPDRoot’s libraries into JupyterLab enables to use the power of rapid development provided by JupyterLab technology to enhance productivity by fast prototyping of MPDRoot’s algorithms. The combination of these technologies together with the existing development environment form a software complex, providing means to accomplish the long term strategic objectives—competent software development with reliable quality control and algorithm innovation.