MRI: Acquisition of a High-Capacity Data Analysis System for Gravitational-Wave Detection with Advanced LIGO

University Of Wisconsin-Milwaukee, Milwaukee WI

Investigators

Abstract

The Laser Interferometer Gravitational Wave Observatory (LIGO) is an NSF-funded project to detect gravitational waves, the ripples in space and time emitted by violent astrophysical events such as supernovae or colliding pairs of neutron stars or black holes. The first detection of these waves in September 2015 confirmed a central prediction of one the most fundamental theories in science: Einstein's theory of general relativity. With this and subsequent detections, LIGO is now operating as an astronomical observatory allowing us to explore the Universe using a completely new sense: gravitational waves. This award provides computing resources at UWM to analyze the LIGO data and will accelerate the discoveries enabled by the LIGO instruments. The computing cluster is designed to address the specific needs of Advanced LIGO data analysis. It is an integral part of the LIGO Scientific Collaboration (LSC) effort to analyze the full Advanced LIGO data sets. LIGO analysis activities will be prioritized on 90% of the cluster; other activities on the remaining 10% will include searches for radio pulsars/transients, the detection of gravitational waves using pulsar timing arrays by NANOGrav, and numerical astrophysics. Analyses to be run on the new cluster include searches for compact binaries and related parameter-estimation. The facility will be available to all 1000+ members of the LSC, and thus will have broad international impact. Moreover, these facilities have also been used by researchers at UWM and elsewhere for interdisciplinary research in fields such as astronomy and astrophysics. This will continue, thus ensuring that this instrument will have a broad scientific reach at UWM and beyond. The deployment and use of this instrument will also provide a hands-on opportunity for education and training to undergraduates, graduate students, post-doctoral scientists and academic staff in the use of large-scale computing hardware to tackle complex technical questions. The baseline computational instrument will consist of 720 execute nodes with 2,280 modern E3-1241v3 CPU cores. The group will add several high-memory login nodes, storage servers (1,152 TB gross storage) and networking equipment. Theoretical peak performance will be about 300 TFLOPS (AVX2) with sufficient storage to host all of the calibrated data from advanced detectors (30 TB/yr) and several months of raw data (600 TB/yr). Benchmarks indicate that the cluster will be able to process over 28 million template matches in real time. The scale of this project is commensurate with a new aspect of this mission within LIGO data analysis as a secondary site for running the real-time (latency of seconds) search for compact binary mergers -- a task best handled at a dedicated site such as UWM. This instrument will facilitate and participate the discovery and the study of the exotic astrophysical events throughout the Universe that produce gravitational waves. This award will provide part of the computational resources needed to analyze the data as it is delivered by the new Advanced LIGO instruments. It will support an array of scientific activities including the detection of new gravitational-wave signals and the broader astrophysical and astronomical follow-on analyses.