# 专题跟踪 · AI 运维 · 2026-09-03
> 日期：2026-09-03（Asia/Shanghai）。对应卷宗 `topics/ai-ops.md`。AIME 原文强调确定性状态机，禁止写成 LLM Agent。

## 摘要
- VCF Operations 9.1 数据表把实时监控、Workbench、K8s/AI 负载监控、MCP/AG API 写进同一 Operate 面；告警动作默认可配但策略里自动化默认关闭。
- NCM Intelligent Operations：异常带 + Playbooks；NCM 2.x 跨多 Prism Central 聚合 AIOps 信号。许可墙在 NCM，不在 NCI Starter 赠品叙事之外的“免费大脑”。
- Scale AIME：层级 FSM，内嵌 HyperCore；厂商对比表把传统 AIOps 写成“叠在第三方上的脚本”。1–50,000 站点为厂商自称。
- 深信服 SkyOps 英文页 2026-09-03 抓取被 Cloudflare 拦截；能力以厂商营销句（150+ 场景、7–30 天预测等）存在，须标自称。中文巡检走 portal.sangforcloud.com AK/SK。
- Morpheus Agentic/OpsRamp 在 9.0 Enterprise；Proxmox 补的是 24/7 人，不是 AIOps SKU。SmartX 6.3 增量在 HA 而非大脑。

## 原文摘录

### [VMware Cloud Foundation Operations datasheet](https://www.vmware.com/docs/vmw-vcf-operations-datasheet)
- 来源：VMware 数据表
> VMware Cloud Foundation Operations provides a single console for private cloud with integrated compute, storage and network operations. … Faster issue detection and resolution with consolidated diagnostics, logs, alerts, network topology, and insights. … Real-time, predictive capacity and cost analytics and continuous performance optimization. Monitoring for Kubernetes containers, VMs, and AI workloads. Enhanced troubleshooting experience with consolidated diagnostics, metrics, logs, events and network insights. … AI-Ready Platform with APIs to integrate with AG and MCP.

### [Actions in VCF Operations](https://techdocs.broadcom.com/us/en/vmware-cis/vcf/vcf-9-0-and-later/9-1/infrastructure-operations/configuring-alerts-and-actions/actions-configuring-actions.html)
- 来源：TechDocs / Last Updated June 29, 2026
> You activate actionable alerts in your policies. By default, automation is deactivated in policies. … When an alert is triggered that is associated to the recommendation, it triggers the action without any user intervention. … For example, if you do not have access to the Power Off VM action, but you can create and modify alerts and recommendations, you can see the Power Off VM action and assign it to an alert recommendation. Then, if you automate the action in your policy, … When CPU stress on your virtual machines exceeds a critical, immediate, or warning level, the alert triggers the recommended action without user intervention.

### [NCM datasheet · Intelligent Operations](https://www.nutanix.com/library/datasheets/ncm)
- 来源：Nutanix
> Easily and quickly gain back the control and visibility of AHV and ESXi virtual environments across the organization by implementing Infrastructure AIOps. Intelligent Ops Reporting Capacity Forecast and Planning Resource Inefficiency Detection and Right-sizing Support for ESXi on non-NCI Environments Low-code/No-code Operations Automation

### [NCM 2.x 多站点 AIOps](https://www.nutanix.com/blog/next-gen-nutanix-cloud-manager)
- 来源：Nutanix 博客
> NCM 2.0 evolves into a multi‑site manager that spans multiple Prism Centrals and VMware vCenters, centralizing VM and cluster inventory, alerts, events, audits, and AIOps signals so you can define alert policies and playbooks once and apply them across domains.

### [AIME product page](https://www.scalecomputing.com/aime-autonomous-infrastructure-management-engine)
- 来源：Scale Computing / 抓取 2026-09-03
> Using a hierarchical state machine, AIME continuously monitors the full stack — storage, compute, and networking — and executes predefined corrective actions the moment conditions warrant them. … It identifies the root cause and minimizes the impact of those issues when it cannot repair them automatically, notifying users with specific problem determination and action, versus just sending a stream of data that must be interpreted. … Behavior is deterministic: the same state transition always produces the same response, making remediation predictable and auditable. … Designed for 1 to 50,000 edge sites … Learning & Adaptation | Dependent on external data lakes or ML engines | Based on live telemetry from global fleets

### [How AIME Brings Autonomous AIOps to the Edge](https://www.scalecomputing.com/resources/aiops-edge-autonomous-infrastructure-aime)
- 来源：Scale 资源页
> AIME isn’t an add-on, it’s the automation layer SC//HyperCore that delivers hands-off infrastructure operations. That distinction is critical when autonomy and speed are required at the edge.

### [Sangfor SkyOps](https://www.sangfor.com/cloud-and-infrastructure/innovations/skyops-aiops)
- 来源：检索摘要；本机 WebFetch 2026-09-03 返回 Cloudflare block（Ray ID a351b9b50e5dc95b）。下列句子来自公开检索片段，**厂商自称**。
> Sangfor SkyOps is an advanced AIOps solution … cloud-based platform connects to your on-premises Sangfor Private Cloud (HCI/VDI) … predict potential issues, such as disk failures or spikes in resource usage … reducing Mean Time To Respond (MTTR) by 50%. … It can recognize over 150 typical fault scenarios, predict hardware failures 7-30 days in advance, and pinpoint root causes in just one second, all with more than 90%

### [SKYOPS 巡检必读](https://support.sangfor.com.cn/productDocument/read?category_id=332127&product_id=107&version_id=688)
- 来源：深信服支持 / 更新时间：2025-09-05
> 支持AK/SK 和 邮箱密码 两种方式巡检，优先推荐使用 AK/SK 方式，因为邮箱巡检成功率较低 … 登录https://portal.sangforcloud.com/ … 选择“运维服务”-“云端智能大脑”，点击“云端代理管理” … 选择与线下连接的云端代理，点击“查看密钥”，即可获取AK/SK

## 对卷宗的影响
- 四列拆法（异常/根因/自愈/出境）继续有效。AIME 必须写确定性 FSM。
- SkyOps 国际页可达性差，国内材料以中文知识中心+云端门户为准。
- MCP 进入 VCF Ops 数据表，运维写操作暴露给 Agent 的风险写入规划含义。