# Red Hat 日跟踪 · 2026-10-21
> 日期：2026-10-21（Asia/Shanghai，周三）。窗口：2026-10-20（周二）早报之后 → 2026-10-21（周三）早。上一成功早报 / 日跟踪 **2026-10-20**。抓取全库：Saved 277 / skipped 29 / errors 42；blocked_hosts：community.hpe.com、docs.redhat.com、e.huawei.com、support.huawei.com、www.globenewswire.com、www.reddit.com、www.sangfor.com。blocked 列表与 10-20 **完全相同**；B&F / NextPlatform / The Register 的 industry/ 本日仍可展开（B&F 新增 10-20 日戳 3 篇；NP/Register 无新 URL）。 原料：archive/2026-10-21/redhat/。

> 相对 2026-10-20 真正新增 URL **5**（llm-d 1.0；priority-aware serving 深测；Delta×OpenShift Dev Spaces；MLOps for edge AI；Linux Foundation 开源 AI 调研）。RHEL×OCI / MCP Gateway GA / DevSecOps demo 等为昨报已载**重抓**。

## 摘要
- **中强 NEW · llm-d 1.0 发布（Crossing the threshold to dependable open source AI serving）：** 文称「Today, Red Hat and our community partners announced the release of llm-d 1.0」；day-2 加固含 priority-aware flow control、路由器 hot standby 高可用与滚动升级；自称 agentic 基准峰值 **>6.6M 输出 token/分钟**、最多 **3,000** 并发编码 agent 零抢占；社区 **500+** 贡献者 / **25** 家组织、CNCF 治理；经 Red Hat AI Inference 交付（OpenShift AI 内或第三方 K8s）——**社区/厂商发版博，数字为厂商自称**。
- **中 NEW · Shared GPUs without shared latency（Red Hat AI 3.5 priority-aware serving 深测）：** 功能 9 月 AI 3.5 已 GA（见 2026-09-10 日跟踪），本文给出新实测：H100 上 **95%** premium 请求首 token <0.5s，batch 洪峰下拒绝数从 **48,224** 降到 **0**；新增 concurrency detector——**厂商实测博，非新 GA**。
- **中 NEW · Delta Air Lines × OpenShift Dev Spaces（Summit 2026 分享回顾）：** CDE 覆盖 **1,200+** 开发者，从 20 人试点起步；AI agent 在 OpenShift 集群内工作区运行构建/测试——**厂商案例博**（客户发言经厂商转述，非独立 VoC）。
- **轻–中 NEW · MLOps for edge AI：** 声明式/拉取式边缘舰队管理；ACM 对 OpenShift/MicroShift 做标签放置与 GitOps，Red Hat Edge Manager 管 RHEL/Device Edge（bootc delta 更新、Greenboot AI 健康检查）——**技术博**。
- **轻 NEW · Linux Foundation 开源 GenAI 调研（RH 部分赞助）：** 61% 贡献者逐行审 AI 代码、80% 重视模型来源/训练数据血缘——**赞助调研**。
- docs.redhat.com 仍 **host_blocked**。

## 原文摘录

### [llm-d 1.0](https://www.redhat.com/en/blog/crossing-threshold-dependable-open-source-ai-serving) · 中强
- 归档：`redhat/crossing-the-threshold-to-dependable-open-source-ai-serving-b92433bd.txt`
- 口径：**厂商/社区发版博**。

> Today, Red Hat and our community partners announced the release of llm-d 1.0.  
> … llm-d delivered over 6.6 million output tokens per minute at peak on agentic benchmarks. The cluster sustained up to 3,000 concurrent coding agents with zero preemptions…  
> Today, the llm-d community has more than 500 contributors across 25 organizations, all building openly under CNCF governance.

### [Priority-aware serving](https://www.redhat.com/en/blog/shared-gpus-without-shared-latency-priority-aware-serving-in-red-hat-ai) · 中
- 归档：`redhat/shared-gpus-without-shared-latency-priority-aware-serving-in-red-hat-a-62b258a3.txt`

> Built on llm-d flow control and generally available in Red Hat AI 3.5, priority-aware serving decides which request goes first when the pool is full, and applications change nothing. On a NVIDIA H100, 95% of premium requests got their first token in under half a second, and batch rejections fell from 48,224 to 0.

### [Delta × OpenShift Dev Spaces](https://www.redhat.com/en/blog/how-delta-air-lines-scaled-ai-native-development-1200-engineers-cloud-development-environments) · 中
- 归档：`redhat/how-delta-air-lines-scaled-ai-native-development-to-1-200-engineers-wi-f5779f2e.txt`

> The workspace—where code lives, builds run, and AI agents work—runs entirely in Red Hat OpenShift clusters using OpenShift Dev Spaces.

### [MLOps for edge AI](https://www.redhat.com/en/blog/mlops-edge-ai-managing-ai-models-across-distributed-fleets) · 轻–中
- 归档：`redhat/mlops-for-edge-ai-managing-ai-models-across-distributed-fleets-ac39ccda.txt`

> For OpenShift and MicroShift deployments, Red Hat Advanced Cluster Management for Kubernetes does the equivalent job through placement rules, selecting clusters by label and deploying the challenger workload only to the matching subset before going fleet-wide.

### [Building AI the open source way](https://www.redhat.com/en/blog/building-ai-open-source-way-research-insights-red-hat-and-linux-foundation) · 轻
- 归档：`redhat/building-ai-the-open-source-way-research-insights-from-red-hat-and-the-88a351ec.txt`

> Among open source contributors, 61% say they conduct detailed, line-by-line reviews and full functional testing of AI-generated code before submitting it.

## 对基础档案的影响
- **最近动态须追加**：llm-d 1.0；AI 3.5 priority-aware 实测；Delta Dev Spaces 案例。