feat(blog): add recap for ContainerDays Singapore 2026 with sessions on GPU sharing and AI Factory

blog/containerdays-singapore-2026-recap/index.md

@@ -0,0 +1,67 @@
+---
+title: "HAMi at ContainerDays Singapore 2026: From GPU Sharing to AI Factory"
+date: "2026-10-30"
+description: "A recap of five ContainerDays Singapore 2026 sessions covering HAMi: hands-on GPU scheduling, Neocloud GPU utilization, DRA integration, AI Factory architecture, and AI-native infrastructure."
+tags: ["ContainerDays", "GPU", "Kubernetes", "AI", "Singapore"]
+authors: [hami_community]
+---
+
+[ContainerDays & AI Context Singapore 2026](https://www.containerdays.io/containerdays-singapore-2026/) was held on October 27-28 at Suntec Singapore, bringing together practitioners in cloud native, AI infrastructure, and platform engineering. HAMi Community Manager [Jimmy Song](https://github.com/rootsongjc) and HAMi Contributor [Reza Jelveh](https://github.com/fishman) (both at Dynamia) presented a hands-on workshop and a conference talk, and HAMi appeared in three other sessions at the event.
+
+![ContainerDays Singapore 2026](/img/containerdays-singapore-2026-recap/banner.png)
+
+<!-- truncate -->
+
+## Workshop: From Zero to Shared GPUs
+
+The workshop **From Zero to Shared GPUs: Hands-on GPU Scheduling in Kubernetes with HAMi** walked participants through GPU sharing with a local lab built on kind. A mock device plugin registered two simulated NVIDIA A100-40GB GPUs, so participants ran the real HAMi Scheduler and observed requests, device selection, and Pod placement without physical GPUs.
+
+![Workshop lab at ContainerDays Singapore 2026](/img/containerdays-singapore-2026-recap/workshop-demo.jpg)
+
+One experiment made a key point visible: three Pods each requesting 15,000 MB of GPU memory and 30% of compute were placed across the two simulated GPUs, while a Pod requesting 45,000 MB stayed Pending. Even though the two cards had enough free memory combined, the scheduler must check whether a single device can satisfy the request, not just the node's free capacity.
+
+The lab continued with a comparison of **binpack** and **spread** placement, and an overview of how HAMi combines with the **KAI Scheduler**, **Volcano**, and **Kubernetes DRA**, with HAMi-core providing runtime resource limits in each combination. Simulated GPUs are sufficient for validating scheduling logic; runtime isolation still needs to be verified on real hardware.
+
+The [workshop slides](/resources/events/2026-containerdays-singapore/containerdays-singapore-2026-workshop.pdf) and the [lab guide](https://github.com/Project-HAMi/hami-workshop) are available online, with details on the [event page](/landing/containerdays-singapore).
+
+## Talk: 10X Your Neocloud's GPU Utilization
+
+Reza Jelveh's session **10X Your Neocloud's GPU Utilization with HAMi** addressed a practical question for Neocloud platforms: when GPUs must serve multiple tenants and workload types, how do you reduce idle capacity?
+
+Surveys cited in the talk reported GPU utilization mostly between 5% and 20% across teams: workloads queue for GPUs while allocated GPUs sit underused. A time-series inference task needing about 1 GB of memory can still occupy an entire 141 GB H200 if allocation is whole-card only.
+
+![Inference replicas sharing one GPU, scaling utilization while keeping p95 latency stable](/img/containerdays-singapore-2026-recap/inference-replicas.png)
+
+For inference, adding replicas on the same card scales utilization: in the demo, ten replicas reached 100% GPU SM utilization at the same p95 latency. The "10X" in the title is not a universal promise, as results depend on the workload mix, resource configuration, and platform baseline. Large training jobs keep whole GPUs, untrusted tenants call for MIG or dedicated devices, and heterogeneous devices such as AMD GPUs and Ascend NPUs need per-vendor resource management, which HAMi supports alongside schedulers like Volcano, Kueue, and the KAI Scheduler. The talk also distinguished HAMi's two metric sources: allocation metrics from the Scheduler (port 9395) and usage metrics from vGPUmonitor (port 9394).
+
+The [talk slides](/resources/events/2026-containerdays-singapore/containerdays-singapore-2026-10x.pdf) are available, with details on the [event page](/landing/containerdays-singapore-10x).
+
+## DRA Allocates, But Who Enforces?
+
+In **DRA Allocates, But Who Enforces? Closing the GPU Sharing Gap with HAMi**, Ivan Porta (Todea) examined the gap between DRA's resource allocation and runtime enforcement. DRA can describe and allocate devices through ResourceClaims, but the Kubernetes API does not by itself stop a container from exceeding its GPU quota at runtime.
+
+The session showed how the HAMi components close that gap: the mutating webhook adjusts scheduling information, the scheduler extender participates in placement decisions, the device plugin registers and assigns devices, and HAMi-core intercepts CUDA calls in the container runtime to enforce memory and compute limits. HAMi-DRA can convert `nvidia.com/gpu` requests into DRA ResourceClaims, keeping the Pod's resource interface stable while the underlying mechanism evolves.
+
+![Ivan Porta presenting on DRA and HAMi](/img/containerdays-singapore-2026-recap/ivan-porta.jpg)
+
+## Will Everyone Build an AI Factory?
+
+Saiyam Pathak (vCluster Labs) explored the path from a single GPU on a personal computer to an enterprise AI Factory. His live demo ran multiple teams' isolated Kubernetes clusters and inference workloads on the same GPU, using open-source projects including vCluster and HAMi.
+
+![Saiyam Pathak presenting on AI Factory architecture](/img/containerdays-singapore-2026-recap/saiyam-pathak.jpg)
+
+The takeaway: an AI Factory is not just a pile of GPUs. It is a software platform that organizes hardware into manageable compute environments, where Kubernetes handles orchestration and GPU sharing lets limited accelerator capacity serve more teams.
+
+## From Cloud-Native to AI-Native
+
+Xu Chen (DaoCloud) discussed Southeast Asia's shift from cloud-native application modernization toward AI-native platforms, covering open cloud-native AI platforms and the role of open-source ecosystems. HAMi was included in the stack as the project for sharing and managing heterogeneous compute devices in Kubernetes, alongside other cloud-native components for AI workloads.
+
+![Xu Chen presenting on cloud-native to AI-native transformation](/img/containerdays-singapore-2026-recap/xu-chen.jpg)
+
+## Common Threads
+
+The five sessions approached different problems, but kept returning to the same questions: how resources are allocated, how limits are enforced at runtime, whether the scheduling policy fits the workload, and whether the platform can observe both allocated and used capacity.
+
+GPU sharing is not a uniform answer for every workload, and finer partitioning is not automatically better. Choosing an allocation model means weighing workload characteristics, performance targets, isolation boundaries, and the platform's overall utilization.
+
+The HAMi community will continue working on GPU sharing, heterogeneous compute management, Kubernetes scheduling, and runtime isolation. To get involved, visit the [community page](/community) or the [HAMi repository](https://github.com/Project-HAMi/HAMi).

i18n/zh/docusaurus-plugin-content-blog/containerdays-singapore-2026-recap/index.md

@@ -0,0 +1,67 @@
+---
+title: "HAMi 亮相 ContainerDays 新加坡 2026：从 GPU 共享到 AI Factory"
+date: "2026-10-30"
+description: "回顾 ContainerDays 新加坡 2026 上与 HAMi 相关的五场分享：GPU 共享动手实验、Neocloud 算力利用率、DRA 集成、AI Factory 架构与 AI 原生基础设施。"
+tags: ["ContainerDays", "GPU", "Kubernetes", "AI", "Singapore"]
+authors: [hami_community]
+---
+
+[ContainerDays & AI Context Singapore 2026](https://www.containerdays.io/containerdays-singapore-2026/) 于 10 月 27 至 28 日在新加坡新达城会展中心举行，议题覆盖云原生、AI 基础设施与平台工程。HAMi Community Manager [宋净超（Jimmy Song）](https://github.com/rootsongjc)与 HAMi Contributor [Reza Jelveh](https://github.com/fishman)（两位均来自密瓜智能）带来了一场 Workshop 和一场技术分享，HAMi 还出现在大会的另外三场议程中。
+
+![ContainerDays 新加坡 2026](/img/containerdays-singapore-2026-recap/banner.png)
+
+<!-- truncate -->
+
+## Workshop：从零开始实践共享 GPU
+
+Workshop **From Zero to Shared GPUs: Hands-on GPU Scheduling in Kubernetes with HAMi** 通过基于 kind 的本地实验环境带领参与者实践 GPU 共享。实验使用模拟设备插件注册了两张虚拟的 NVIDIA A100-40GB，参与者无需物理 GPU，即可运行真实的 HAMi Scheduler，观察资源申请、设备选择与 Pod 调度的完整过程。
+
+![ContainerDays 新加坡 2026 的 Workshop 现场](/img/containerdays-singapore-2026-recap/workshop-demo.jpg)
+
+一组实验直观展示了调度中的关键问题：三个各申请 15,000 MB 显存与 30% 计算资源的 Pod 被分配到两张模拟 GPU 上，而一个申请 45,000 MB 显存的 Pod 始终处于 Pending 状态。即使两张卡的剩余显存之和能够满足请求，调度器也必须判断单个设备能否满足任务需求，而不仅是检查节点的空闲容量。
+
+实验随后对比了 **binpack** 与 **spread** 两种放置策略的效果，并介绍了 HAMi 与 **KAI Scheduler**、**Volcano** 以及 **Kubernetes DRA** 的协作方式：无论选择哪种调度体系，都可以由 HAMi-core 提供运行时资源限制。模拟 GPU 适合验证调度逻辑，运行时隔离仍需在真实硬件上验证。
+
+[工作坊讲义](/resources/events/2026-containerdays-singapore/containerdays-singapore-2026-workshop.pdf)与[实验手册](https://github.com/Project-HAMi/hami-workshop)已在线发布，详情见[活动专页](/zh/landing/containerdays-singapore)。
+
+## 分享：让 Neocloud 的 GPU 利用率提升十倍
+
+Reza Jelveh 的分享 **10X Your Neocloud's GPU Utilization with HAMi** 关注 Neocloud 平台的实际问题：当 GPU 需要同时服务多个租户和不同类型的工作负载时，如何减少资源闲置？
+
+演讲引用的多组数据显示，不同团队的 GPU 利用率大多在 5% 至 20% 之间：一部分任务还在排队等待 GPU，已分配出去的 GPU 却没有被充分使用。一个只需约 1 GB 显存的时序推理任务，在整卡分配模式下仍可能独占一张 141 GB 的 H200。
+
+![推理副本共享同一张 GPU，利用率随副本数上升而 p95 延迟保持稳定](/img/containerdays-singapore-2026-recap/inference-replicas.png)
+
+对于推理服务，在同一张卡上增加副本可以提升利用率：演示中，10 个副本的 GPU SM 利用率达到 100%，p95 延迟保持不变。标题中的“10X”并非对所有平台的承诺，实际效果取决于工作负载构成、资源配置与平台基线。大型训练任务通常保留整卡，不可信租户需要结合 MIG 或独占设备评估隔离要求；面对 AMD、昇腾等异构设备，HAMi 提供对应的资源管理与限制机制，并可与 Volcano、Kueue、KAI Scheduler 等协同。演讲还区分了 HAMi 的两类监控指标：Scheduler 暴露的资源分配指标（端口 9395）与 vGPUmonitor 暴露的资源使用指标（端口 9394）。
+
+[演讲幻灯片](/resources/events/2026-containerdays-singapore/containerdays-singapore-2026-10x.pdf)已在线发布，详情见[活动专页](/zh/landing/containerdays-singapore-10x)。
+
+## DRA 负责分配，谁来负责执行？
+
+在 **DRA Allocates, But Who Enforces? Closing the GPU Sharing Gap with HAMi** 这场分享中，Ivan Porta（Todea）讨论了 DRA 的资源分配与运行时限制之间的衔接问题。DRA 可以通过 ResourceClaims 描述和分配设备，但 Kubernetes API 本身并不会阻止容器在运行时超出 GPU 资源额度。
+
+演讲介绍了 HAMi 各组件如何补足这一环节：Mutating Webhook 调整调度相关信息，Scheduler Extender 参与调度决策，Device Plugin 负责设备注册与分配，HAMi-core 在容器运行时拦截 CUDA 调用并执行显存与计算资源限制。HAMi-DRA 还可以将 `nvidia.com/gpu` 资源请求转换为 DRA ResourceClaims，在保持 Pod 资源接口稳定的同时，让底层资源管理方式随 Kubernetes 环境演进。
+
+![Ivan Porta 演讲现场](/img/containerdays-singapore-2026-recap/ivan-porta.jpg)
+
+## 人人都会构建 AI Factory 吗？
+
+Saiyam Pathak（vCluster Labs）分享了从个人电脑上的单张 GPU 到企业级 AI Factory 的演进路径。现场演示让多个团队各自隔离的 Kubernetes 集群与推理工作负载运行在同一张 GPU 上，其中使用了 vCluster 与 HAMi 等开源项目。
+
+![Saiyam Pathak 演讲现场](/img/containerdays-singapore-2026-recap/saiyam-pathak.jpg)
+
+这场分享的要点在于：AI Factory 不只是把更多 GPU 集中起来，而是通过软件平台将硬件组织为可管理的计算环境，其中 Kubernetes 负责工作负载编排，GPU 共享让有限的加速计算资源服务更多团队。
+
+## 从云原生到 AI 原生
+
+Xu Chen（DaoCloud）讨论了东南亚数字化转型中从云原生应用现代化走向 AI 原生平台建设的路径，内容涵盖开放式 Cloud-Native AI 平台与开源生态的作用。HAMi 作为 Kubernetes 中异构计算设备共享与管理的项目被纳入这一技术版图，与其他云原生组件共同支撑 AI 工作负载。
+
+![Xu Chen 演讲现场](/img/containerdays-singapore-2026-recap/xu-chen.jpg)
+
+## 共同的主线
+
+五场分享切入的角度不同，但反复回到同样的问题：资源如何分配、运行时如何执行限制、调度策略是否适合工作负载，以及平台能否同时观察资源的分配情况与实际使用情况。
+
+GPU 共享不是所有工作负载的统一答案，切分粒度也并非越细越好。选择资源分配方式，需要结合任务特点、性能目标、隔离边界与平台整体利用率综合判断。
+
+HAMi 社区将继续围绕 GPU 共享、异构算力管理、Kubernetes 调度与运行时隔离开展探索。欢迎访问[社区页面](/zh/community)或 [HAMi 仓库](https://github.com/Project-HAMi/HAMi)参与协作。

i18n/zh/docusaurus-plugin-content-blog/tags.yml

@@ -13,6 +13,9 @@
 "CNCF":
   label: "CNCF"
   permalink: "/cncf"
+"ContainerDays":
+  label: "ContainerDays"
+  permalink: "/container-days"
 "DRA":
   label: "DRA"
   permalink: "/dra"
@@ -61,6 +64,9 @@
 "China":
   label: "China"
   permalink: "/china"
+"Singapore":
+  label: "Singapore"
+  permalink: "/singapore"
 "KAI Scheduler":
   label: "KAI Scheduler"
   permalink: "/kai-scheduler"