EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Advancements in Autonomous Incident Response: Integrating Large Language Models with Database Reliability Engineering Telemetry

Evans Ighodalo, Ayinoluwa Kolawale, Chukwudera Obumneke Anunagba, Oluranti Ogundapo, Harouna Wendpanga Yann Christian Sankara, Emmanuel Chukwuma Okeke, Aisha Ife Abdulwahab, Toussida Fatah Tanguy Minoungou

Journal of Cyber-Physical Security and Robotics · 25.10.2026 · 2 (04) · 55-76

DOI: 10.64235/bgd7kswx8 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Cyber-Physical Security and Robotics, online erschienen am 25.10.2026. Verfasst von Evans Ighodalo und 7 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Database incidents remain one of the most expensive classes of production failure, because the database sits beneath almost every user-facing service and its failure modes (lock contention, replication lag, plan regressions, storage ex-haustion, connection storms) cascade quickly. Over the past decade, autonomous database research has largely concen-trated on tuning and optimization, while incident response has stayed a human-centred discipline supported by dash-boards and alerting. The arrival of capable large language models (LLMs) that can call tools, read heterogeneous telem-etry and follow written procedures has changed this picture. Using an integrative literature review of 228 sources, comprising 18 core studies from the database systems, software engineering and AIOps literature and 210 seed studies screened from a consolidated corpus of 1,056 records, combined with conceptual modelling, this paper synthesizes the trajectory from self-tuning systems to LLM-assisted diagnosis and develops a conceptual model that couples Database Reliability Engineering (DRE) telemetry with an LLM reasoning agent inside a closed control loop. The architecture is organized around four ideas: an incident packet that compresses raw signals into a form an LLM can reason over; re-trieval of operational knowledge such as runbooks and past postmortems; graded autonomy enforced by policy rather than by the model; and verification against service level objectives with automatic rollback. We present a taxonomy of database telemetry for LLM consumption, a five-level autonomy model for remediation, and an evaluation protocol with metrics that go beyond diagnostic accuracy to include time to mitigation, unsafe action rate and operator trust. The paper discusses implications for practice, research and governance, states the limitations of a conceptual review, and sets out a future research agenda covering telemetry hallucination, adversarial content in logs, evaluation realism and the shifting role of the database engineer.

Zitieren

APA

Ighodalo, E., Kolawale, A., Anunagba, C. O., Ogundapo, O., Sankara, H. W. Y. C., Okeke, E. C., Abdulwahab, A. I., & Minoungou, T. F. T. (2026). Advancements in Autonomous Incident Response: Integrating Large Language Models with Database Reliability Engineering Telemetry. Journal of Cyber-Physical Security and Robotics, 2(04), 55-76. https://doi.org/10.64235/bgd7kswx8

Vancouver

Ighodalo E, Kolawale A, Anunagba CO, Ogundapo O, Sankara HWYC, Okeke EC, et al. Advancements in Autonomous Incident Response: Integrating Large Language Models with Database Reliability Engineering Telemetry. Journal of Cyber-Physical Security and Robotics. 2026;2(04):55-76. doi:10.64235/bgd7kswx8

BibTeX
@article{Ighodalo2026,
  author = {Ighodalo, Evans and Kolawale, Ayinoluwa and Anunagba, Chukwudera Obumneke and Ogundapo, Oluranti and Sankara, Harouna Wendpanga Yann Christian and Okeke, Emmanuel Chukwuma and Abdulwahab, Aisha Ife and Minoungou, Toussida Fatah Tanguy},
  title = {Advancements in Autonomous Incident Response: Integrating Large Language Models with Database Reliability Engineering Telemetry},
  journal = {Journal of Cyber-Physical Security and Robotics},
  year = {2026},
  volume = {2},
  number = {04},
  pages = {55--76},
  doi = {10.64235/bgd7kswx8}
}
Kostenlos starten →