dsx-ai-factory/run-ai
Run:ai is the AI workload orchestration platform for Kubernetes. This monorepo contains the control plane services, in-cluster Kubernetes operator/controllers, shared libraries and clients, the installer, and the local dev environment.
Go · ★ 759 · Fork 126 · 在 GitHub 查看
在手机上浏览源码、订阅每日 AI 趋势日报与项目架构深度解读:获取 CopoHub 客户端(iOS & HarmonyOS)→
GitHub Trending 上榜记录
dsx-ai-factory/run-ai 已在 GitHub Trending 每日榜单上榜 2 天,最高排名第 54,首次上榜 2026-11-12,最近一次上榜 2026-11-13。
| 日期 | 排名 | Star | 当日增长 |
|---|---|---|---|
| 2026-11-13 | #54 | 759 | +8 |
| 2026-11-12 | #54 | 756 | +8 |
同期上榜的项目
- alibaba/open-code-review – Secure, fast, efficient, battle-tested at Alibaba's scale. Hybrid architecture code review tool: deterministic pipelines + LLM Agent, precise line-level comments, built-in multi-language ruleset (NPE, thread-safety, XSS, SQL injection), OpenAI & Anthropic compatible.
- GoogleChrome/modern-web-guidance-src – Keep your coding agent up to date with the latest web best practices
- openai/codex-security – OpenAI's Codex Security CLI and TypeScript SDK for finding, validating, and fixing security vulnerabilities. npm: https://www.npmjs.com/package/@openai/codex-security
- ZhuLinsen/daily_stock_analysis – LLM 驱动的多市场股票智能分析系统:多源行情、实时新闻、决策看板与自动推送,支持零成本定时运行。 LLM-powered multi-market stock analysis system with multi-source market data, real-time news, decision dashboard, automated notifications, and cost-free scheduled runs.
- we-promise/sure – The personal finance app for everyone (by everyone)
- backnotprop/plannotator – The workspace to stay in the loop and on top of agents - shape, decide, plan, annotate, review code, share with team, and send all feedback to agents with one click.
dsx-ai-factory/run-ai 项目深度剖析与商业评估报告
dsx-ai-factory/run-ai 项目深度剖析与商业评估报告
1. 核心定位与商业落地价值
1.1 核心痛点与行业背景 在以大语言模型(LLM)与生成式 AI 为核心的算力密集型时代,GPU 基础设施已成为企业最昂贵的资产之一。然而,原生 Kubernetes 体系在管理 AI 工作负载时面临多重天然结构性缺陷: 静态标量分配与低利用率:Kubernetes 原生仅支持将 nvidia.com/gpu 作为不可拆分的整数标量资源进行调度,导致交互式研发(JupyterLab)、轻量推理或测试任务严重浪费整卡显存与算力。 缺乏多租户层级化配额与公平共享:原生 ResourceQuota 仅提供平面化的 Namespace 级别硬限制,无法表达“组织(Org)- 部门(Department)- 项目组(Project)”的三层层级化弹性配额(Hierarchical Quotas),缺乏跨部门的闲置算力借用(Borrowing)与保障水位(Guaranteed vs. Overcommit)机制。 复杂批处理调度缺失:原生调度器缺乏全有或全无(Gang Scheduling/All-or-Nothing)、拓扑感知(NVLink/NVSwitch 域感知)以及基于业务优先级的智能抢占机制,极易导致分布式训练发生死锁(Deadlock)或通信效率低下。
dsx-ai-factory/run-ai(NVIDIA Run:ai)的开源,标志着企业级 GPU 编排与算力虚拟化层向开源云原生基础设施的关键演进。其核心定位是构建在标准 Kubernetes 之上的 AI 专属分布式工作负载调度与算力资源治理平台。
+-----------------------------------------------------------------------+ | Run:ai Control Plane | | (Multi-Cluster Governance / Identity / Hierarchical Quota / REST API)| +-----------------------------------------------------------------------+ ▲ State Sync & gRPC │ Workload Dispatch & Policies ▼ +-----------------------------------------------------------------------+ | Run:ai Cluster | | +-----------------------------------------------------------------+ | | | KAI Scheduler (Gang, Fair-share, Topology, Preemption) | | | +-----------------------------------------------------------------+ | | | Dynamic GPU Allocation (Fractional GPU, Memory Enforcement) | | | +-----------------------------------------------------------------+ | | | Workload Controllers & Webhooks (Workspaces / Training / KServe)| | | +-----------------------------------------------------------------+ | | Kubernetes Worker Nodes with NVIDIA GPUs | +-----------------------------------------------------------------------+
1.2 企业级商业落地场景 异构算力中心统一共享池化(Private AI Cloud):打破算法团队独占物理节点的壁垒,将分散的 DGX/HGX 集群聚合为统一弹性算力池,综合提升 GPU 总体利用率(从业界平均的 15%-25% 提升至 60%-75%)。 混合型 AI 工作负载共存:在同一集群内混合调度长周期分布式预训练(Training Operator/LeaderWorkerSet)、短周期批处理微调、交互式 Notebook 研发环境以及突发性高并发推理服务(KServe/vLLM)。 跨集群联合算力治理(Hybrid & Multi-Cluster AI):集中控制面允许平台团队跨本地 IDC(On-Premise)与公有云(AWS/GCP/Azure)执行统一的配额管控、审计策略下发与审计回溯。
1.3 竞品差异化与商业护城河 相比同类开源解决方案,Run:ai 构筑了差异化的竞争壁垒: 相较于 Volcano / Apache YuniKorn:Volcano 强于通用大数据与批处理调度,但在底层 GPU 硬件抽象、显存切分控制(Fractional GPU)、动态超分与交互式 AI 工作流(Workspaces 暂停/恢复)方面缺乏深度闭环。 相较于 Kubernetes Kueue:Kueue 是轻量级的资源队列系统,依赖底层调度器完成物理绑定,而 Run:ai 深度集成了专用自研的 KAI Scheduler,将排队配额、动态切片与底层拓扑编排(NUMA、PCIe Switch、NVLink)在单调度周期内原子化解决。 生态与硬件同构协同:在被 NVIDIA 深度整合后,Run:ai 拥有与 NVIDIA GPU Operator、Network Operator(InfiniBand/RoCE)、MIG(Multi-Instance GPU)硬件虚拟化技术的原生调优与发布协同能力。
---
在 README 中展示上榜徽章
[](https://github.fq6825.top/repo/dsx-ai-factory/run-ai)