RUNMON-RIFT: Adaptive Configuration and Healing for Large-Scale Parameter Inference

TL;DR

RUNMON-RIFT is a flexible tool designed to improve large-scale gravitational wave parameter inference by dynamically managing unreliable distributed computing environments and adjusting analysis settings to ensure accurate results.

Contribution

It introduces RUNMON-RIFT, a novel adaptive system that enhances the robustness and reliability of the RIFT gravitational wave analysis pipeline on distributed hardware.

Findings

Successfully identified and mitigated unreliable computing environments.

Adjusted pipeline settings to ensure comprehensive parameter coverage.

Demonstrated effectiveness through controlled experiments.

Abstract

Gravitational wave parameter inference pipelines operate on data containing unknown sources on distributed hardware with unreliable performance. For one specific analysis pipeline (RIFT), we have developed a flexible tool (RUNMON-RIFT) to mitigate the most common challenges introduced by these two uncertainties. On the one hand, RUNMON provides several mechanisms to identify and redress unreliable computing environments. On the other hand, RUNMON provides mechanisms to adjust pipeline-specific run settings, including prior ranges, to ensure the analysis completes and encompasses the physical source parameters. We demonstrate both general features with two controlled examples.