SLURM Accounting Dataset for "Multivariate Analysis of Resource Interdependencies in HPC Clusters: A Validated R-Vine Copula Approach"

This dataset contains job-level resource consumption records extracted via the SLURM accounting utility sacct from the Kabré supercomputer at the National High Technology Center of Costa Rica. Records span June 18, 2024 to July 14, 2026 and cover three compute partitions with distinct hardware profiles: Dribe (memory-optimized), Kurá (general-purpose CPU), and Nukwa (GPU-accelerated). Fields included: ConsumedEnergyRaw (total energy consumed, joules), CPUTimeRAW (CPU time, seconds), ReqMem (requested memory), ReqCPUS (requested CPU count, used to classify serial vs. parallel execution mode), Partition, State (filtered to COMPLETED jobs only), and Submit (submission timestamp, used for temporal-stability analysis). No user-identifying or job-command fields are included. This dataset accompanies the paper Benavides, D. (2026), "Multivariate Analysis of Resource Interdependencies in HPC Clusters: A Validated R-Vine Copula Approach." The accompanying analysis code (R-Vine copula fitting, robust structure-selection verification, and a six-part statistical validation framework) is available at: https://github.com/DylanBC09/Multivariate-Analysis-of-Resource-Interdependencies-in-HPC-Clusters-using-R-Vine-Copulas