跳转至

弃权协议：Clos 网络架构的根本原因分析 (RCA)

文章背景与核心概要

本文介绍了由 Madhava Gaikwad 和 Deepak Pandey 在第 20 届 USENIX 操作系统设计与实现研讨会 (OSDI '26) 上发表的研究成果。针对超大规模数据中心网络中遥测数据普遍存在的噪声大、不完整及异步性问题，传统的基于评分的归因方法往往难以奏效，导致故障定位不稳定或错误。

为了解决这一挑战，作者提出了 CoreSec，这是一个生产级的根本原因分析 (RCA) 系统。该系统摒弃了传统的加权融合方法，转而采用一种 PAM 风格的“弃权代数”。通过结合拓扑感知配置，CoreSec 能够做出确定性的决策，在证据模糊时明确选择“弃权”，并随着新数据的积累实现单调收敛。该研究为自动化云网络诊断提供了一个稳健且可解释的底层框架。

弃权协议：Clos 网络架构的根本原因分析 (RCA)

执行摘要

The Abstention Protocol: RCA for Clos Fabrics is a research paper authored by Madhava Gaikwad and Deepak Pandey, presented at the 20th USENIX Symposium on Operating Systems Design and Implementation (OSDI '26).

The paper addresses the longstanding challenges of Root Cause Analysis (RCA) in large-scale hyperscale datacenter networks, where telemetry is frequently noisy, partial, and asynchronous. Traditional score-based attribution approaches struggle under these conditions, often resulting in unstable or incorrect fault localization. To solve this, the authors introduce CoreSec, a production-grade RCA system that replaces weighted fusion with a PAM-style abstention algebra. Combined with topology-aware configurations, CoreSec yields deterministic decisions, explicitly abstains when evidence is ambiguous, and converges monotonically as new data accumulates—proving to be a resilient, explainable foundation for automated cloud network diagnosis.

元数据与出版详情

  • arXiv 标识符： arXiv:2608.21412 [cs.AI]
  • 主要学科： 人工智能 (cs.AI)
  • 次要学科： 分布式、并行与集群计算 (cs.DC)；网络与互联网架构 (cs.NI)
  • 作者：
  • Madhava Gaikwad
  • Deepak Pandey
  • 提交日期： 2026 年 8 月 11 日
  • 会议/期刊参考： 第 20 届 USENIX 操作系统设计与实现研讨会 (OSDI '26)，西雅图，华盛顿州，2026 年 7 月，第 405–421 页
  • DOI： 10.48550/arXiv.2608.21412
  • 许可协议： 知识共享署名 4.0 (查看许可图标来源：license icon)
  • arXiv Identifier: arXiv:2608.21412 [cs.AI]
  • Primary Subject: Artificial Intelligence (cs.AI)
  • Secondary Subjects: Distributed, Parallel, and Cluster Computing (cs.DC); Networking and Internet Architecture (cs.NI)
  • Authors:
  • Madhava Gaikwad
  • Deepak Pandey
  • Submission Date: August 11, 2026
  • Conference/Journal Reference: 20th USENIX Symposium on Operating Systems Design and Implementation (OSDI '26), Seattle, WA, July 2026, pp. 405–421
  • DOI: 10.48550/arXiv.2608.21412
  • License: Creative Commons Attribution 4.0 (View license icon source: license icon)

摘要

大型数据中心网络中的根本原因分析 (RCA) 极具挑战性，因为遥测数据通常存在噪声、不完整且异步。在这种条件下，基于评分的方法性能会下降，往往导致不稳定或错误的归因。

作者提出了 CoreSec，这是一个生产级的 RCA 系统，它用 PAM 风格的弃权代数取代了加权融合。遥测代理通过控制标志进行组合，从而在证据模糊时产生确定性的决策和明确的弃权。CoreSec 将这种代数与拓扑感知配置相结合，能够捕获 Clos 架构中的故障面，并随着证据的积累实现单调收敛。

CoreSec 已在超大规模环境中部署，在无需重新调整的情况下，在各种环境中提供了稳定且可解释的 RCA 行为。作者的运营经验表明，这种带有弃权机制的结构化组合为现实世界云网络中的自动化 RCA 奠定了实用的基础。

Root cause analysis (RCA) in large datacenter networks is challenging because telemetry is noisy, partial, and asynchronous. Score-based approaches degrade under these conditions, often yielding unstable or incorrect attributions.

The authors present CoreSec, a production RCA system that replaces weighted fusion with a PAM-style abstention algebra. Telemetry agents are composed using control flags that yield deterministic decisions and explicit abstention when evidence is ambiguous. CoreSec combines this algebra with topology-aware configurations that capture failure surfaces across Clos fabrics and converge monotonically as evidence accumulates.

Deployed at hyperscale, CoreSec provides stable and explainable RCA behavior across diverse environments without retuning. The authors' operational experience shows that structured composition with abstention forms a practical foundation for automated RCA in real-world cloud networks.

访问与资源

外部引用工具与指标

External Citation Tools & Metrics