https://aik8s.run/ GPU、大数据、调度、训练、推理、RAG、Agent 与生产运维 zh-CN Tue, 11 Aug 2026 09:01:41 +0000 https://aik8s.run/assets/practices/gpu-notebook-platform-evolution/README/ https://aik8s.run/assets/practices/gpu-notebook-platform-evolution/README/ Tue, 11 Aug 2026 09:01:01 +0000 The three article figures are generated by scripts/generategpunotebookfigures.py. assets https://aik8s.run/ai-k8s/practices/gpu-notebook-platform-evolution/ https://aik8s.run/ai-k8s/practices/gpu-notebook-platform-evolution/ Tue, 11 Aug 2026 09:01:01 +0000 从 JupyterHub/KubeSpawner、code-server 和组级八卡共享，走到 Ceph RBD、CephFS、Workspace Operator 与 KubeVirt 根盘的真实演进 practices https://aik8s.run/ai-k8s/practices/ https://aik8s.run/ai-k8s/practices/ Tue, 11 Aug 2026 09:01:01 +0000 用可复现的实验、性能数字、事故证据和决策树回答 AI/LLM on Kubernetes 的高频问题 practices https://aik8s.run/ai-k8s/development/gpu-notebook-platform/ https://aik8s.run/ai-k8s/development/gpu-notebook-platform/ Tue, 11 Aug 2026 09:01:01 +0000 从 JupyterHub、Kubeflow 和托管 Workbench，到整卡、MIG、共享 GPU、用户 Home、对象存储与本地缓存的生产选型 development https://aik8s.run/ai-k8s/rag-agent/openclaw-enterprise-agent-platform/ https://aik8s.run/ai-k8s/rag-agent/openclaw-enterprise-agent-platform/ Tue, 11 Aug 2026 00:00:00 +0000 分析 OpenClaw 的可复用能力、企业平台缺口、安全风险、推荐架构和采用决策 rag-agent https://aik8s.run/ai-k8s/agentic-workloads/ https://aik8s.run/ai-k8s/agentic-workloads/ Mon, 10 Aug 2026 16:16:11 +0000 Agent 工作负载的运行时隔离、网络、身份、工具权限和审计设计 ai-k8s https://aik8s.run/ai-k8s/ https://aik8s.run/ai-k8s/ Mon, 10 Aug 2026 16:16:11 +0000 从加速器、大数据和集群，到训练、推理、RAG、Agent 与生产运维的完整技术地图 ai-k8s https://aik8s.run/ https://aik8s.run/ Mon, 10 Aug 2026 16:16:11 +0000 面向平台工程师、SRE、数据平台和模型服务团队的 AI/LLM 基础设施工程文档 首页 https://aik8s.run/ai-k8s/rag-agent/agent-landscape-2026/ https://aik8s.run/ai-k8s/rag-agent/agent-landscape-2026/ Mon, 10 Aug 2026 14:47:14 +0000 主流编程、浏览器与企业 Agent 的产品格局、实现原理、开发框架、风险和演进趋势 rag-agent https://aik8s.run/ai-k8s/practices/deepseek-v4-flash-h20-evaluation/ https://aik8s.run/ai-k8s/practices/deepseek-v4-flash-h20-evaluation/ Mon, 10 Aug 2026 12:13:31 +0000 记录 DeepSeek-V4-Flash-0731 在单机八卡 H20 上的资源条件、vLLM 启动流程、性能基线、同 Pod PD 分离实测与 OpenWebUI 对接方法 practices https://aik8s.run/ai-k8s/practices/rbg-existing-cluster/ https://aik8s.run/ai-k8s/practices/rbg-existing-cluster/ Mon, 10 Aug 2026 11:58:43 +0000 在 Kubernetes 1.30 集群部署 RoleBasedGroup，实测角色依赖、服务发现、扩缩、自愈、Ray 两机推理和 NIXL P/D 分离，并以相同镜像与模型对比 RBG 和 AIBrix practices https://aik8s.run/ai-k8s/practices/aibrix-gpu-multinode-pd-production/ https://aik8s.run/ai-k8s/practices/aibrix-gpu-multinode-pd-production/ Sat, 08 Aug 2026 13:22:34 +0000 在生产 Kubernetes 集群使用 NVIDIA L20、AIBrix v0.7.0、RayClusterFleet、StormService 和 vLLM，验证两机模型并行、NIXL P/D 分离、八节点 Qwen3-235B FP8，以及 DeepSeek 70B 的四节点单卡、两节点双卡和单节点四卡拓扑 practices https://aik8s.run/ai-k8s/practices/comfyui-minimax-h3-gpu/ https://aik8s.run/ai-k8s/practices/comfyui-minimax-h3-gpu/ Sat, 08 Aug 2026 12:42:42 +0000 使用 NVIDIA GPU、只读 CephFS、ComfyUI extra_model_paths、Init Container 模型别名及双层 Ingress，在受限网络 Kubernetes 环境提供 MiniMax-H3 工作流页面 practices https://aik8s.run/ai-k8s/practices/aibrix-existing-cluster/ https://aik8s.run/ai-k8s/practices/aibrix-existing-cluster/ Sat, 08 Aug 2026 02:16:47 +0000 在 Kubernetes 1.30 集群安装 AIBrix v0.7.0，用 CPU mock 跑通模型路由、P/D、StormService、自动扩缩容、Prometheus 指标和 Higress 两层网关串联 practices https://aik8s.run/ai-k8s/practices/gpu-failure-atlas/ https://aik8s.run/ai-k8s/practices/gpu-failure-atlas/ Sat, 08 Aug 2026 02:02:56 +0000 用 XID、ECC、NVLink、掉卡、NCCL、RDMA 和 kubelet 证据定位 GPU 故障 practices https://aik8s.run/ai-k8s/inference/higress-ai-gateway/ https://aik8s.run/ai-k8s/inference/higress-ai-gateway/ Fri, 07 Aug 2026 21:59:40 +0000 在既有 Kubernetes 集群隔离安装 Higress，理解 Controller、Gateway、Console、AI Proxy 与可观测插件，并设计 Higress 和 AIBrix 的同集群及跨集群链路 inference https://aik8s.run/ai-k8s/practices/sglang-model-gateway-cpu-lab/ https://aik8s.run/ai-k8s/practices/sglang-model-gateway-cpu-lab/ Fri, 07 Aug 2026 09:53:15 +0000 在没有 GPU 的 Kubernetes 1.30 集群部署 SGLang Router 和两个 OpenAI-Compatible Mock Worker，实测动态发现、轮询、摘除恢复与 Prometheus 指标 practices https://aik8s.run/ai-k8s/inference/engines/ https://aik8s.run/ai-k8s/inference/engines/ Fri, 07 Aug 2026 09:53:15 +0000 对比 vLLM、SGLang、TensorRT-LLM、Triton、llama.cpp 等引擎的能力、边界和 Kubernetes 集成方式 inference https://aik8s.run/ai-k8s/inference/distributed-serving/ https://aik8s.run/ai-k8s/inference/distributed-serving/ Fri, 07 Aug 2026 09:53:15 +0000 设计多机模型副本、LeaderWorkerSet、Prefill/Decode 分离和 KV 传输，并对比 AIBrix、llm-d、KServe、Dynamo、Ray Serve 与 vLLM Production Stack inference https://aik8s.run/ai-k8s/inference/gateway-routing/ https://aik8s.run/ai-k8s/inference/gateway-routing/ Thu, 06 Aug 2026 12:12:15 +0000 区分 API Gateway、Gateway API Inference Extension 和模型请求调度，设计认证、配额、缓存感知和发布策略 inference https://aik8s.run/ai-k8s/practices/gpu-resource-bank-tidal-platform/ https://aik8s.run/ai-k8s/practices/gpu-resource-bank-tidal-platform/ Wed, 05 Aug 2026 10:14:41 +0000 用 Kueue 或 Volcano 实现业务组 GPU 配额互借，结合 KServe、AIBrix、KEDA 与 vLLM 构建实时和定时弹性，并建立从硬件健康到单位 Token 成本的可观测闭环 practices https://aik8s.run/ai-k8s/practices/kubevirt-local-desktop-lab/ https://aik8s.run/ai-k8s/practices/kubevirt-local-desktop-lab/ Wed, 05 Aug 2026 07:07:09 +0000 在没有 Ceph 的 Kubernetes 集群中，把 KubeVirt VM 限制到单个节点，使用本地盘保存完整系统环境，并通过受限 noVNC 网关从浏览器访问桌面 practices https://aik8s.run/ai-k8s/practices/kubevirt-rbd-notebook/ https://aik8s.run/ai-k8s/practices/kubevirt-rbd-notebook/ Wed, 05 Aug 2026 00:17:27 +0000 为每个用户提供带独立 RBD 根盘的持久 Linux 工作站，覆盖宿主机准备、KubeVirt/CDI、GPU 直通、快照、故障恢复和生产验收 practices https://aik8s.run/ai-k8s/practices/spot-checkpoint/ https://aik8s.run/ai-k8s/practices/spot-checkpoint/ Tue, 04 Aug 2026 08:14:28 +0000 测量抢占通知、Checkpoint、重排队和训练恢复的真实成本收益 practices https://aik8s.run/ai-k8s/practices/pd-break-even/ https://aik8s.run/ai-k8s/practices/pd-break-even/ Tue, 04 Aug 2026 08:14:28 +0000 用双池容量、KV 传输和真实负载判断 P/D 分离是否值得 practices https://aik8s.run/ai-k8s/practices/ollama-to-production/ https://aik8s.run/ai-k8s/practices/ollama-to-production/ Tue, 04 Aug 2026 08:14:28 +0000 把本地模型验证迁移为可压测、可灰度、可观测的 vLLM/KServe 服务 practices https://aik8s.run/ai-k8s/practices/model-memory-calculator/ https://aik8s.run/ai-k8s/practices/model-memory-calculator/ Tue, 04 Aug 2026 08:14:28 +0000 估算模型权重、KV Cache、运行时开销、最大并发和单位 Token 成本 practices https://aik8s.run/ai-k8s/practices/model-distribution-100-nodes/ https://aik8s.run/ai-k8s/practices/model-distribution-100-nodes/ Tue, 04 Aug 2026 08:14:28 +0000 为 TB 级模型和百节点推理池设计 Registry、对象存储、P2P 与节点缓存实验 practices https://aik8s.run/ai-k8s/practices/low-gpu-utilization/ https://aik8s.run/ai-k8s/practices/low-gpu-utilization/ Tue, 04 Aug 2026 08:14:28 +0000 从数据、CPU、通信、内核、批处理和平台指标定位 GPU 空转 practices https://aik8s.run/ai-k8s/practices/kueue-vs-volcano/ https://aik8s.run/ai-k8s/practices/kueue-vs-volcano/ Tue, 04 Aug 2026 08:14:28 +0000 用相同训练任务验证准入队列、Gang、公平共享、抢占与拓扑能力 practices https://aik8s.run/ai-k8s/practices/kubernetes-vs-slurm/ https://aik8s.run/ai-k8s/practices/kubernetes-vs-slurm/ Tue, 04 Aug 2026 08:14:28 +0000 从训练、推理、拓扑、队列、生态和运维边界选择 AI 调度平台 practices https://aik8s.run/ai-k8s/practices/inference-engine-benchmark/ https://aik8s.run/ai-k8s/practices/inference-engine-benchmark/ Tue, 04 Aug 2026 08:14:28 +0000 固定模型、硬件和负载，公平比较主流 LLM 推理引擎 practices https://aik8s.run/ai-k8s/practices/incident-review/ https://aik8s.run/ai-k8s/practices/incident-review/ Tue, 04 Aug 2026 08:14:28 +0000 用时间线、故障域、恢复证据和行动项复盘训练与推理事故 practices https://aik8s.run/ai-k8s/practices/gpu-pending/ https://aik8s.run/ai-k8s/practices/gpu-pending/ Tue, 04 Aug 2026 08:14:28 +0000 从队列准入、调度、拓扑、设备、存储和弹性逐层定位 GPU Pending practices https://aik8s.run/ai-k8s/practices/domestic-accelerators/ https://aik8s.run/ai-k8s/practices/domestic-accelerators/ Tue, 04 Aug 2026 08:14:28 +0000 用平台契约接入昇腾及其他国产加速器，管理驱动、资源名、镜像、调度和可观测差异 practices https://aik8s.run/ai-k8s/practices/cloud-gpu-selection/ https://aik8s.run/ai-k8s/practices/cloud-gpu-selection/ Tue, 04 Aug 2026 08:14:28 +0000 按库存、拓扑、网络、存储、Spot、出流和软件栈评估云上 GPU practices https://aik8s.run/ai-k8s/practices/air-gapped-ai-platform/ https://aik8s.run/ai-k8s/practices/air-gapped-ai-platform/ Tue, 04 Aug 2026 08:14:28 +0000 在无公网环境同步镜像、Chart、模型、驱动、软件包和安全元数据 practices https://aik8s.run/ai-k8s/practices/agent-sandbox-red-team/ https://aik8s.run/ai-k8s/practices/agent-sandbox-red-team/ Tue, 04 Aug 2026 08:14:28 +0000 用文件、凭据、网络、资源耗尽和持久化测试验证 Agent 隔离边界 practices https://aik8s.run/ai-k8s/cluster/open-source-management/ https://aik8s.run/ai-k8s/cluster/open-source-management/ Tue, 04 Aug 2026 07:44:36 +0000 从 kubectl、Web UI 和 GitOps，到 Rancher、Cluster API、Gardener、Karmada 与 OCM 的分层选型 cluster https://aik8s.run/ai-k8s/cluster/cloud-managed-kubernetes/ https://aik8s.run/ai-k8s/cluster/cloud-managed-kubernetes/ Tue, 04 Aug 2026 07:44:36 +0000 对比阿里云 ACK、腾讯云 TKE、华为云 CCE、AWS EKS、Google GKE 和 Azure AKS 的托管边界与 AI/GPU 能力 cluster https://aik8s.run/cases/ai-cluster-component-checklist/ https://aik8s.run/cases/ai-cluster-component-checklist/ Tue, 04 Aug 2026 06:02:41 +0000 按最小闭环、训练、大数据、Ray、在线推理和高级推理场景列出 Kubernetes 集群需要的组件、引入理由与验收条件 cases https://aik8s.run/ai-k8s/data/model-artifacts/ https://aik8s.run/ai-k8s/data/model-artifacts/ Tue, 04 Aug 2026 06:02:41 +0000 管理模型格式、OCI 制品、跨地域复制、P2P 分发、节点缓存、流式加载、签名和冷启动 data https://aik8s.run/cases/ https://aik8s.run/cases/ Tue, 04 Aug 2026 04:19:40 +0000 记录真实 Kubernetes 与 AI 基础设施环境中的实施过程、故障处理、验证证据和可复用经验 cases https://aik8s.run/ai-k8s/reference/glossary/ https://aik8s.run/ai-k8s/reference/glossary/ Tue, 04 Aug 2026 04:19:40 +0000 GPU、大数据、调度、训练、推理、网络、Agent 沙箱、模型制品和可靠性常用术语速查 reference https://aik8s.run/ai-k8s/ray-llm-platform/ https://aik8s.run/ai-k8s/ray-llm-platform/ Tue, 04 Aug 2026 04:19:40 +0000 Ray Core、Data、Train、Tune、Serve、Serve LLM 与 KubeRay 在大模型数据处理、训练、后训练和在线推理中的边界 ai-k8s https://aik8s.run/ai-k8s/llm-inference/ https://aik8s.run/ai-k8s/llm-inference/ Tue, 04 Aug 2026 04:19:40 +0000 在线大模型服务的控制面、运行时、路由、伸缩和容量设计 ai-k8s https://aik8s.run/ai-k8s/inference/serving-frameworks/ https://aik8s.run/ai-k8s/inference/serving-frameworks/ Tue, 04 Aug 2026 04:19:40 +0000 对比 vLLM、KServe、AIBrix、Ray Serve、BentoML、NVIDIA NIM 与应用编排框架的层次、组合方式和选型边界 inference https://aik8s.run/ai-k8s/distributed-training/ https://aik8s.run/ai-k8s/distributed-training/ Tue, 04 Aug 2026 04:19:40 +0000 训练控制器、Gang Scheduling、集合通信、容错和作业生命周期 ai-k8s https://aik8s.run/ai-k8s/data/big-data-on-kubernetes/ https://aik8s.run/ai-k8s/data/big-data-on-kubernetes/ Tue, 04 Aug 2026 04:19:40 +0000 Spark、Flink、Kafka、Trino、Lakehouse、对象存储、Operator、队列调度，以及与大模型训练和 RAG 的结合 data https://aik8s.run/ai-k8s/data-storage/ https://aik8s.run/ai-k8s/data-storage/ Tue, 04 Aug 2026 04:00:13 +0000 对象存储、共享文件、本地 NVMe、数据加载和多级缓存设计 ai-k8s