FORM, a Fine-grained Object Reading/Writing Model for DUNE
Barnali Chowdhury, P. van Gemmeren, Michael Kirby · EPJ Web of Conferences · 2025
DUNE’s current processing framework (art) was branched from the event processing framework of CMS, a collider-physics experiment. Therefore art is built on event-based concepts as its fundamental processing unit. The “event” concept is not always helpful for neutrino experiments, such as DUNE. In DUNE, each event is represented by a trigger record, which can be much larger than a typical collider event — often several gigabytes, compared to just megabytes for collider events. To avoid allocating large chunks of memory due to the large and complex nature of DUNE’s events, the experiment is developing a framework (Phlex) that is able to break apart trigger records into smaller segments for more granular processing, and then stitch those chunks back together into an event. For an event-processing framework to function efficiently, it must be integrated with an input/output (I/O) system that supports fine-grained data handling. FORM (Fine-grained Object Reading/Writing Model) is a DUNE project focused on developing a data storage and I/O system that enables information to be written and accessed in smaller, more manageable units supporting framework that perform fine-grained event processing. To support fine-grained processing, data objects are partitioned into segments and stored separately in accessible locations. This approach allows the I/O system to read and write individual segments, avoiding the high memory usage that comes from handling large monolithic data objects. The complexity of data storage and I/O operations is encapsulated within the FORM infrastructure, making it transparent to client-side components like processing algorithms. By writing and reading multiple smaller entries as discrete events, FORM improves concurrency and scalability in the data processing pipeline.