When Automation Passes but Configuration Fails: Auditing AI-Generated Cloud Runbooks: A Cross-Site Validation Agenda
DOI:
https://doi.org/10.66687/yywhgb31Keywords:
Artificial Intelligence, Cloud Computing, Information Storage and Retrieval, Algorithms, Computer SecurityAbstract
Cloud runbooks can report successful execution even when their final storage, identity or rollback state violates operating policy. The difficulty is particularly visible when a system is transferred to a new setting. The objective is to develop checks that distinguish a successful command sequence from a correct and recoverable infrastructure state. Relevant methodological and data-quality literature informs the design of the sampling and validation strategy [3–7]. The proposed method is a software-assurance experiment using mutated runbooks, controlled API faults and machine-readable postconditions. Comparators will include human-authored scripts, assistant-generated scripts and scripts with policy verification gates. Related enterprise-storage studies inform the assessment of operational reliability [1] and the deployment assumptions adopted for system-level comparison [2]. The cross-site extension holds out an entire institution or deployment location rather than relying only on random test splits. Statistical and computational sources help define the reproducibility checks and comparison procedures [8–12]. The primary outcomes are configuration violations, undetected partial failures, idempotence and time spent diagnosing recovery. Potential biases include privilege boundaries, rollback safety, configuration drift and reviewer-blinded assessment. Between-site heterogeneity would be examined through prespecified subgroup estimates and transparent domain-shift diagnostics. Additional research informs the prespecified stress tests and limits on interpretation [13–17]. The intended contribution is an auditable research protocol rather than an unsupported claim of effectiveness; related work frames its transferability limits [18–23].
References
[1] Mohammed Nazir. (2023). FlexGroup and FlexCache Optimization for AWS-Based CAE/HPC File Workloads: A Latency and Throughput Evaluation. International Journal of Business & Computational Sciences, 3(1), 25–31. https://ijbcs.org/index.php/IJBCS/article/view/2023-01-04
[2] Mohammed Nazir. (2021). RTO and RPO Optimization Using Asynchronous Replication in Multi-Petabyte Enterprise Storage Environments. Journal of Computing, Intelligence and Information Sciences, 2(1). https://jciis.com/index.php/jciis/article/view/2020-01-04
[3] Cidon, A. et al. (2013). Copysets: Reducing the Frequency of Data Loss in Cloud Storage. USENIX ATC '13. https://www.usenix.org/conference/atc13/technical-sessions/presentation/cidon
[4] Vrable, M., Savage, S. & Voelker, G.M. (2012). BlueSky: A Cloud-Backed File System for the Enterprise. USENIX FAST '12. https://www.usenix.org/conference/fast12/bluesky-cloud-backed-file-system-enterprise
[5] Patel, T. et al. (2020). Uncovering Access, Reuse, and Sharing Characteristics of I/O-Intensive Files on Large-Scale Production HPC Systems. USENIX FAST '20. https://www.usenix.org/conference/fast20/presentation/patel-hpc-systems
[6] Maneas, S. et al. (2022). Operational Characteristics of SSDs in Enterprise Storage Systems: A Large-Scale Field Study. USENIX FAST '22. https://www.usenix.org/conference/fast22/presentation/maneas
[7] Mahmud, T. et al. (2023). ConfD: Analyzing Configuration Dependencies of File Systems for Fun and Profit. USENIX FAST '23. https://www.usenix.org/conference/fast23/presentation/mahmud
[8] Cao, Z., Kuenning, G. & Zadok, E. (2020). Carver: Finding Important Parameters for Storage System Tuning. USENIX FAST '20. https://www.usenix.org/conference/fast20/presentation/cao-zhen
[9] Mahdisoltani, F., Stefanovici, I. & Schroeder, B. (2017). Proactive Error Prediction to Improve Storage System Reliability. USENIX ATC '17. https://www.usenix.org/conference/atc17/technical-sessions/presentation/mahdisoltani
[10] Rodriguez, L.V. et al. (2021). Learning Cache Replacement with CACHEUS. USENIX FAST '21. https://www.usenix.org/conference/fast21/presentation/rodriguez
[11] Wong, D.L.-K. et al. (2024). Baleen: ML Admission & Prefetching for Flash Caches. USENIX FAST '24. https://www.usenix.org/conference/fast24/presentation/wong
[12] Zhou, W. et al. (2025). 3L-Cache: Low Overhead and Precise Learning-based Eviction Policy for Caches. USENIX FAST '25. https://www.usenix.org/conference/fast25/presentation/zhou-wenbin
[13] Li, T. et al. (2024). MinFlow: High-performance and Cost-efficient Data Passing for I/O-intensive Stateful Serverless Analytics. USENIX FAST '24. https://www.usenix.org/conference/fast24/presentation/li
[14] Jiao, B., Goel, A. & Wang, A.-I.A. (2025). Silhouette: Leveraging Consistency Mechanisms to Detect Bugs in Persistent Memory-Based File Systems. USENIX FAST '25. https://www.usenix.org/conference/fast25/presentation/jiao
[15] Chen, W. et al. (2025). IMPRESS: An Importance-Informed Multi-Tier Prefix KV Storage System for Large Language Model Inference. USENIX FAST '25. https://www.usenix.org/conference/fast25/presentation/chen-weijian-impress
[16] Guo, C. et al. (2017). On Calibration of Modern Neural Networks. Proceedings of Machine Learning Research, 70, 1321–1330. https://proceedings.mlr.press/v70/guo17a.html
[17] Mihailescu, M., Soundararajan, G. & Amza, C. (2013). MixApart: Decoupled Analytics for Shared Storage Systems. USENIX FAST '13. https://www.usenix.org/conference/fast13/technical-sessions/presentation/mihailescu
[18] Jiao, Z. et al. (2024). The Design and Implementation of a Capacity-Variant Storage System. USENIX FAST '24. https://www.usenix.org/conference/fast24/presentation/jiao
[19] Qin, R. et al. (2025). Mooncake: Trading More Storage for Less Computation — A KVCache-centric Architecture for Serving LLM Chatbot. USENIX FAST '25. https://www.usenix.org/conference/fast25/presentation/qin
[20] Burke, Q. et al. (2025). On Scalable Integrity Checking for Secure Cloud Disks. USENIX FAST '25. https://www.usenix.org/conference/fast25/presentation/burke
[21] Nachman, A., Yadgar, G. & Sheinvald, S. (2020). GoSeed: Generating an Optimal Seeding Plan for Deduplicated Storage. USENIX FAST '20. https://www.usenix.org/conference/fast20/presentation/nachman
[22] Alagappan, R. et al. (2019). Protocol-Aware Recovery for Consensus-Based Storage. USENIX ATC '19. https://www.usenix.org/conference/atc19/presentation/alagappan
[23] Yang, Z. et al. (2023). λ-IO: A Unified IO Stack for Computational Storage. USENIX FAST '23. https://www.usenix.org/conference/fast23/presentation/yang-zhe
Published
Issue
Section
License
Copyright (c) 2026 Sávio Sampaio
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors retain copyright in their published work. Unless an individual Version of Record identifies different terms, JMRIS articles are distributed under the Creative Commons Attribution 4.0 International License (CC BY 4.0): https://creativecommons.org/licenses/by/4.0/. This permits sharing and adaptation, including commercial use, with appropriate attribution, a link to the license, and indication of changes. Third-party material may be subject to separate terms. Existing articles retain the license originally granted to their Version of Record; a policy update does not retrospectively replace that license.
