Aggregated thermal telemetry statistics and analysis scripts for the Summit cooling degradation study

Aggregated statistics, analysis outputs and scripts accompanying the manuscript Separating coolant-flow deficit from cold-plate thermal-resistance degradation using in-service telemetry of a liquid-cooled GPU supercomputer (submitted to Microelectronics Reliability). The data are derived from 141 daily files (2020-01, 2020-08, 2021-02, 2021-08, 2022-01) of the public OLCF Summit power and thermal telemetry (doi:10.13139/OLCF/1861393; ClusterWise release, CC BY 4.0). The raw 10-s telemetry is not redistributed. agg_*.zip (12 archives): per-day aggregates by host, socket and GPU position: high-load temperature offsets (one archive per month), idle/baseline proxies, XID failure records, job exposure, and power-binned offsets (2021-08; 2022-01 in three parts). analysis_results.zip: outputs behind every table and figure (variance components, timelines, two-segment series-heating test, power-binned decomposition, XID job-conditioned models, manifold model). code.zip: Python and R scripts. GPU_thermal_v1.1_additions.zip (new in v1.1): injection-validation and detection-limit scripts and results (Section 4.4, Tables 3–4), two-segment and power-binned scripts and outputs, the interactive floor thermal maps (Supplementary File S1, self-contained HTML) with their cabinet × slot data, and README_v1.1.md. README.md: column definitions and file guide. Support for 10.13139/OLCF/1861393 is provided by the U.S. Department of Energy, project STF218 under Contract DE-AC05-00OR22725. This research used resources of the Oak Ridge Leadership Computing Facility, which is a DOE Office of Science User Facility.