Software and computing: thinking in new ways

tape-robot-trj-may2016
Inside the Grid Computing Center at Fermilab (Photo: R.Hahn)

This is the first of a series of articles on Software and Computing for Nus to SURF. Subsequent articles will focus on software sharing, online systems for the ProtoDUNEs, networking, storage, batch systems, simulations, and novel approaches.

The 35-ton team is busy analyzing the data from the data run and preparing publications, while the ProtoDUNE detector teams are in their final design phase, getting ready for data in 2018. At the same time, the Far Detector is in preliminary design and the Near Detector is still at a pre-technology-selection stage. The sheer variety of activities in DUNE — and their staggered life cycles — present significant software and computing (S&C) challenges. The collaborators contributing to the DUNE S&C effort have their work cut out for them, providing software support, CPU cycles, networking and storage for all these overlapping activities.

DUNE S&C collaborators are making efficient use of existing software and computing infrastructure, relying heavily on proven tools, resources, and architectures. They have also been contributing to these tools, benefitting both DUNE and the broader community. Given the novelty and scale of this experiment, many of the current solutions require redevelopment, forcing us to think in new ways.

The Far Detector simulation and reconstruction effort’s use of and contributions to LArSoft, a simulation and reconstruction software suite, which is also shared by the MicroBooNE, SBND, ArgoNeuT, and LArIAT collaborations, provides a good example. LArSoft itself relies heavily on the shared tools GEANT4, ROOT, and the art framework. Scaling LArSoft up to even one DUNE Far Detector module requires improving its speed and memory footprint while retaining the precision of the modeling.

Software and computing issues also affect DUNE’s data acquisition and trigger systems, just as for collider experiments, and these system designs will need to maximize the physics output of DUNE.

The S&C efforts will take considerable resources, time, and coordination with other DUNE detector systems — and lots of original thinking.

–Tom Junk, Software and Computing Coordinator