Work

Reproducibility package for "GPU Faults Across Cloud Providers: Characterization and Operational Insights at Scale" (SC26 2026)

Year: 2026

Type: dataset

Abstract: Reproducibility materials accompanying the SC26 (2026) paper "GPU Faults Across Cloud Providers: Characterization and Operational Insights at Scale." This package provides the shareable reproducibility artifact: the SQL query patterns, statistical methodology, generalized data schema, and the aggregate result tables reported in the paper. The raw production telemetry is proprietary and is not included; all cluster and provider identifiers are anonymized, consistent with the paper. See README.md for contents and usage.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.22104096

Authors: Sebastian, Kenneth, Ayalasomayajula, Devendra, Jolly, Annu, Motwani, Rashi, Kuang, Abby, Guardia, Giorgio Di, Miryala, Aman, Thottakara, Sivakumar Arayandi, Prasanna, Ranjitha, Kooner, Satvir

Institution: Nvidia (United States)

Language: English

Topic: Cloud Computing and Resource Management

Subfield: Information Systems

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W7204239049

Cites:

Cited by:

Related to: 10