کنفرانس بین المللی محاسبات نرم

صفحه اصلی / ششمین کنفرانس بین المللی محاسبات نرم

A Systematic Literature Review on AI-Driven DevOps: Enhancing CI/CD, Monitoring, and Error Prediction

نویسندگان :

Ali Shahnazi¹ Nima Nikrouz² Amir Seyed Danesh³

1- دانشکده فنی مهندسی شرق گیلان/گیلان 2- دانشکده فنی مهندسی شرق گیلان/گیلان 3- دانشکده فنی مهندسی شرق گیلان/گیلان

کلمات کلیدی :

DevOps،AIOps،CI/CD،anomaly detection،failure prediction،generative AI

چکیده :

DevOps has become a go-to method for speeding up software delivery, relying on Continuous Integration (CI) and Continuous Deployment (CD) pipelines paired with automated monitoring and feedback loops. As systems grow larger and more complex, traditional DevOps struggles with frequent build failures, limited visibility, and difficulty predicting issues early. To overcome these, experts have introduced AIOps, using machine learning (ML), deep learning (DL), and large language models (LLMs) to boost reliability, automation, and scalability in DevOps workflows. This paper presents a systematic literature review (SLR) of academic and industrial studies from 2020 to 2025, focusing on three key areas: (i) CI/CD automation, (ii) monitoring and anomaly detection, and (iii) fault prediction and remediation. We aim to explore how AI enhances these processes, addressing questions like: What AI algorithms are used in DevOps? How does AI improve CI/CD? What methods exist for error prediction? Our review highlights AI’s strengths—such as better failure prediction, reduced mean time to resolution (MTTR), and self-healing capabilities —while noting challenges like reliance on project-specific data, explainability gaps, and integration overheads. Unlike previous surveys, this study offers an analytical comparison of approaches, backed by a case study on GitLab CI/CD that shows how AI cuts pipeline failures and optimizes resources. The paper concludes by pointing to future opportunities in generative AI, agent-based systems, and security-aware DevOps, shaping a clearer path for smarter, more adaptive software delivery.