An Agentic Framework for Triaging Incidents in Production Cloud Infrastructure
Incidents occur frequently in large-scale cloud infrastructures and can cause significant service disruptions if not triaged and mitigated quickly. Existing automated triage systems rely on complete incident reports or uniform agent quality, which are often infeasible in production due to fragmented observability and heterogeneous team maturity. We present Comfey, a decentralized agentic triage framework designed for this operational reality. Each team in Azure deploys a lightweight Comfey triage agent that analyzes domain-specific data, decides whether to accept or reject an incident, and collaborates with other agents via a shared routing table. Comfey has been running in Azure production for 22 months, where it has triaged around 19,500 incidents with 91.5% accuracy (validated by on-call engineers). It reduces average triage time by 6 times and mitigation time by 4.3 times compared to manual processes, with minimal setup requirements.
Wed 8 JulDisplayed time zone: Eastern Time (US & Canada) change
10:30 - 12:30 | CloudIndustry Papers / Research Papers / Journal-First Paper at MB 2.435 Chair(s): Mariam El Mezouar Royal Military College | ||
10:30 20mTalk | Aloha: Localizing Batch Failures in Large-scale Cloud Systems via Contrast Analysis and Human-in-the-Loop Agent Industry Papers Shenglin Zhang Nankai University, Yujia Wu Nankai University, Jinghuan Ren Nankai University, College of Software, Yongqian Sun Nankai University, Wenwei Gu Nankai University, Chaoyun Zhang Microsoft, Liqun Li Microsoft Research, Qingwei Lin Microsoft, Dongmei Zhang Microsoft, Saravanakumar Rajmohan Microsoft 365, Chetan Bansal Microsoft Research, Minghua Ma Microsoft | ||
10:50 20mTalk | Attention Enhanced Entity Recommendation for Intelligent Monitoring in Cloud Systems Industry Papers Fiza Husain Independent, Anson Bastos Microsoft, Anjaly Parayil Microsoft, Ayush Choure Independent, Chetan Bansal Microsoft Research, Rujia Wang Microsoft, Saravanakumar Rajmohan Microsoft 365 | ||
11:10 20mTalk | An Agentic Framework for Triaging Incidents in Production Cloud Infrastructure Industry Papers Yuhan Yao Microsoft, Yuxuan Jiang University of Michigan Ann-Arbor, Minghua Ma Microsoft, Madhura Vaidya Microsoft, Jieren Deng Microsoft, Yigong Hu Boston University, Chetan Bansal Microsoft Research, Ze Li Microsoft Azure, Murali Chintalapati Microsoft Azure | ||
11:30 20mTalk | TSGuard: Automated User-Centric Incident Diagnosis for AI Workloads in the Cloud Research Papers Yitao Yang The Chinese University of Hong Kong, Yangtao Deng The Chinese University of Hong Kong, Yifan Xiong Microsoft Research, Baochun Li University of Toronto, Hong Xu The Chinese University of Hong Kong, Peng Cheng Microsoft Research Asia Pre-print | ||
11:50 20mTalk | Exploring the impact of cloud computing on software architecture for sustainability: A practitioners' perspective Journal-First Paper | ||
12:10 20mTalk | AccessRefinery: Fast Mining Concise Access Control Intents on Public Cloud Research Papers Ning Kang Xi'an Jiaotong University, Peng Zhang Xi'an Jiaotong University, Jianyuan Zhang Xi'an Jiaotong University, Hao Li Xi'an Jiaotong University, Dan Wang Xi'an Jiaotong University, Zhenrong Gu Xi'an Jiaotong University, Weibo Lin Huawei Cloud, Shibiao Jiang Huawei Cloud, Zhu He Huawei Cloud, Xu Du Huawei Cloud, Longfei Chen Huawei Cloud, Jun Li Huawei, Xiaohong Guan Xi'an Jiaotong University DOI Pre-print | ||