Tachyon
Intelligent Multi-Scale Modeling of Distributed Resilient Infrastructure and Workflows for Data Intensive HEP Analyses
Grant
- Tachyon: Intelligent Multi-Scale Modeling of Distributed Resilient Infrastructure and Workflows for Data Intensive HEP Analyses
- UIC PI: Zhiling Lan
- Sponsor: U.S. Department of Energy, High Energy Physics program
- Award Number: DE-SC0024271
- Partner Institutions: Rensselaer Polytechnic Institute, Argonne National Laboratory, Fermi National Accelerator Laboratory, University of Illinois Chicago, and University of California, Davis
- Start Date: October 1, 2023
- End Date: September 30, 2028
- Award Amount: $7.5M total, $1.5M to UIC
Overview
The Deep Underground Neutrino Experiment (DUNE) will generate as much as 6 GB of data every five milliseconds, nearly 300 times the volume captured by current technologies. That data must travel 800 miles from the detector in South Dakota to Fermilab’s storage system and on to Argonne National Laboratory for processing on the Aurora supercomputer, in real time or near real time, so that results can feed back into how the detectors and beam facilities are operated.
Tachyon builds a digital twin of this distributed system: an intelligent multi-scale parallel simulation framework for end-to-end modeling of a highly complex scientific workflow with a real-time component. The models account for storage and network failures and for failing computer components, so that remedies and fault tolerance can be planned before problems hit the data campaign. They also inform a scheduling policy that lets DUNE’s high energy physics analyses run on Aurora without displacing the other science applications that share the machine.
Original news story: Lan hopes to help researchers unlock the mysteries of neutrinos (UIC Computer Science, October 12, 2023)