Conclave: A Unified Platform for Real-Time Slurm Cluster Data Monitoring

Effective management of Slurm-based highperformance computing (HPC) clusters requires the integration of large and diverse and distributed data sources, including command-line utilities, accounting databases, and host-level performance metrics. The heterogeneity and fragmentation of these data hinder monitoring, fault diagnosis, and performance optimization, which can degrade overall system efficiency and resource utilization, as well as decrease user satisfaction. This paper presents Conclave, a unified data monitoring application designed as a 'single pane of glass' platform for Slurm cluster operations. By consolidating job, node, and system-level information into a centralized interface, Conclave provides real-time visualization and historical trend analysis. The platform enables rapid fault isolation, validation of configuration parameters such as Quality of Service (QoS), and correlates user job profiles with node performance metrics to support informed resource allocation and capacity planning. Through this integration into a cohesive analytics environment, Conclave enhances operational decision-making, reduces troubleshooting overhead, and improves the reliability and scalability of Slurm-managed systems.