Workshops Archive

What Your Job Queue Isn’t Telling You: Workflow Monitoring with Drona


Workshop: HUST-26: 13th International Workshop on HPC User Support Tools

Authors: Ananya Adiki, Owen Schillaci, Surada Suwansathit, Andrii Kryvenko, Huy Le, Huy Hoang, Marinus Pennings, and Honggao Liu (Texas A&M University)

Abstract: Scientific workloads on high performance computing (HPC) systems often involve multi-stage workflows beyond simple job submission. Existing tools provide limited monitoring, exposing primarily scheduler-level metadata and forcing users to piece together workflow progress, performance behavior, and domain-specific results through fragmented interfaces. A new and improved version of the Drona Workflow Engine addresses this gap by providing a schema-driven framework combined with a support library for constructing integrated job monitoring artifacts throughout the workflow lifecycle. This paper presents Drona’s monitoring architecture, based on dynamic elements, retriever functions, workflow abstractions, and a support library, enabling real-time observability and interactive job management during execution within a unified interface. We demonstrate Drona monitoring dashboards for various domain-specific workflows.


Back to HUST-26: 13th International Workshop on HPC User Support Tools Archive Listing

Back to Full Workshop Archive Listing