KubeCon + CloudNativeCon + Open Source Summit + AI_dev China 2024

Explore China Merchants Bank's (CMB), one of China's largest retail banks, transformative journey through cloud migration, cloud-native transformation, and platform engineering over the past three years. Despite challenges such as increased complexity in cloud technology and management, and potential risks to developer productivity and continuous assurance of financial services, CMB successfully leveraged KubeVela, OpenFeature, Envoy, Clilum, and OpenTelemetry to build the Next-Gen FinTech IDP. This led to the management of 70% of applications within a year and improved developer experience, covering thousands of R&D engineers. We'll discuss the strategic thinking, 'Golden Path' implementation, struggles, trade-offs, and key success metrics with platform engineering maturity model. This session provides a blueprint and reference architecture for financial organizations undergoing similar transformations.

在KubeCon的会议描述中，探索中国招商银行（CMB）作为中国最大的零售银行之一，在过去三年中通过云迁移、云原生转型和平台工程的变革之旅。尽管面临诸如云技术和管理复杂性增加、开发人员生产力和金融服务持续保障的潜在风险等挑战，CMB成功利用KubeVela、OpenFeature、Envoy、Clilum和OpenTelemetry构建了下一代金融科技IDP。这导致了一年内管理了70%的应用程序，并改善了开发人员体验，涵盖了数千名研发工程师。我们将讨论战略思维、“黄金路径”实施、挣扎、权衡和关键成功指标，以及平台工程成熟度模型。本场演讲提供了金融机构进行类似转型的蓝图和参考架构。

In the new AI coming age, it has become inevitable for any organizations to embrace the multi-cloud approach. Managing applications across multiple clouds can present various challenges, including resilience, performance, security, cost, and deployment management. How well did you prepare yourself and your services for that new coming age? This presentation will introduce Karmada and PipeCD, two powerful tools designed to support organizations in effectively addressing these challenges and achieving seamless multi-cloud management. Karmada is a multi-cloud container orchestration, while PipeCD is a multi-cloud continuous delivery solution. Both tools are built based on extensive experience in managing applications at scale across multiple clouds. We will delve into the key features and benefits of Karmada and PipeCD, and how they can simplify multi-cloud management. Together, we can unlock the true potential of multi-cloud systems and empower organizations to thrive in the era of AI.

在新的人工智能时代，任何组织都不可避免地需要采用多云方法。在多个云上管理应用程序可能会带来各种挑战，包括弹性、性能、安全性、成本和部署管理。您为新时代做好了多少准备？本次演讲将介绍Karmada和PipeCD，这两款强大的工具旨在支持组织有效应对这些挑战，实现无缝的多云管理。Karmada是一个多云容器编排工具，而PipeCD是一个多云持续交付解决方案。这两款工具都是基于在多个云上管理应用程序的丰富经验构建的。我们将深入探讨Karmada和PipeCD的关键特性和优势，以及它们如何简化多云管理。让我们一起释放多云系统的真正潜力，赋予组织在人工智能时代蓬勃发展的力量。

Crossplane has become the foundation of many Internal Developer Platforms (IDPs). A requirement for any IDP in production is the ability to make changes and upgrades to the platform with confidence. This talk will cover testing and release patterns based on our experience building production-ready environments across a range of Crossplane users. We’ll cover the lifecycle of a Crossplane Composition upgrade, from local commit to pull request to target customer environment, end-to-end testing tools, handling API changes, and how to control updates to customer environments. For quite a while, testing Crossplane Compositions meant relying exclusively on costly end-to-end layers. In this talk, we're unveiling new unit testing capabilities that allow you to evaluate and test your Composition code in complete isolation.

Crossplane已成为许多内部开发者平台（IDPs）的基础。在生产中，任何IDP的要求都是能够有信心地对平台进行更改和升级。 本次演讲将涵盖基于我们在跨多个Crossplane用户构建生产就绪环境的经验，讨论测试和发布模式。 我们将介绍Crossplane Composition升级的生命周期，从本地提交到拉取请求再到目标客户环境，端到端测试工具，处理API更改以及如何控制对客户环境的更新。 相当长一段时间以来，测试Crossplane Compositions意味着完全依赖昂贵的端到端层。在本次演讲中，我们将揭示新的单元测试功能，使您能够在完全隔离的环境中评估和测试您的Composition代码。

With the rise of generative AI technology, more and more applications are starting to integrate with the capabilities of generative AI. However, the high costs of training and inference can be daunting for developers. In this talk, we will discuss the issues and solutions that need additional consideration when using elastic scaling in generative AI scenarios, including: ● How to enhance the elastic startup efficiency of generative AI ● How to address the efficiency of inference when separating compute and storage in generative AI ● How to reduce the costs of training and inference ● How to solve the interruption problem in AI training scenarios using Spot instances ● How to address the issue of capacity elasticity in LLM scenarios Finally, we will introduce the practical experience of the world's leading generative AI service provider: HaiYi (seaart.ai), allowing more developers to understand the architectural methods of elastic cloud AI infrastructure.

随着生成式人工智能技术的兴起，越来越多的应用程序开始与生成式人工智能的能力集成。然而，训练和推理的高成本可能会让开发人员望而却步。在这次演讲中，我们将讨论在生成式人工智能场景中使用弹性扩展时需要额外考虑的问题和解决方案，包括： ● 如何提高生成式人工智能的弹性启动效率 ● 如何在生成式人工智能中分离计算和存储时解决推理效率的问题 ● 如何降低训练和推理的成本 ● 如何使用Spot实例解决AI训练场景中的中断问题 ● 如何解决LLM场景中的容量弹性问题 最后，我们将介绍世界领先的生成式人工智能服务提供商海艺（seaart.ai）的实际经验，让更多开发人员了解弹性云AI基础设施的架构方法。

DP Technology provides scientific simulation platforms for research in biomedicine, energy, materials and other industries. Science simulation workflows are inherently complex and resource-intensive, and manual deployment is often prone to errors. After adopting Argo workflows to orchestrate science simulation, we get productivity 100% improvement. In this talk, we will introduce why chose Argo Workflow, how to orchestrate large-scale tasks of science simulation, how to make whole system scalability and reliability. Specially, we will share best practice about how manage super large workflow (thousands of tasks), how to do reasonable workflow retry, how to use memorization to reduce runtime and compute cost, how to interact with HPC systems. We also made contributions to Argo community to enhance functionalities and improve reliability. Additionally, we'll introduce DFlow, our open-source Python SDK designed for the seamless orchestration of scientific simulations with Argo Workflows.

DP Technology为生物医药、能源、材料等行业的研究提供科学模拟平台。科学模拟工作流程本质上复杂且资源密集，手动部署往往容易出错。采用Argo工作流程来编排科学模拟后，我们的生产力提高了100%。在本次演讲中，我们将介绍为什么选择Argo工作流程，如何编排大规模科学模拟任务，如何实现整个系统的可扩展性和可靠性。特别是，我们将分享如何管理超大型工作流程（数千个任务），如何合理重试工作流程，如何使用记忆化来减少运行时间和计算成本，如何与HPC系统交互。我们还为Argo社区做出了贡献，以增强功能性和提高可靠性。此外，我们还将介绍DFlow，我们的开源Python SDK，旨在与Argo工作流程无缝协同编排科学模拟。

In today's rapidly evolving landscape of software development, the role of developer portals has become indispensable. This presentation delves into the experiences of developers over the course of a year, exploring the transformative impact of Backstage developer portal on their workflows, collaboration, and overall productivity based on case studies from existing adopters of Backstage. Through a comprehensive exploration of real-world scenarios, this talk offers insights into the daily challenges faced by developers and how Backstage empowers them to overcome these hurdles. From streamlined onboarding processes to simplified access to internal services and documentation, attendees will gain a deeper understanding of the multifaceted benefits that Backstage brings to developer teams. Moreover, we'll discuss best practices for leveraging Backstage to foster a culture of innovation, collaboration, and continuous improvement.

在当今快速发展的软件开发领域，开发者门户的作用变得不可或缺。本次演讲将深入探讨开发者在一年时间内的经验，通过现有Backstage采用者的案例研究，探讨Backstage开发者门户对他们的工作流程、协作和整体生产力的转变影响。 通过对现实场景的全面探讨，本次演讲将为参与者提供洞察开发者面临的日常挑战，以及Backstage如何赋予他们克服这些障碍的能力。从简化入职流程到简化访问内部服务和文档，参与者将更深入地了解Backstage为开发团队带来的多方面好处。此外，我们还将讨论利用Backstage促进创新、协作和持续改进文化的最佳实践。

Kubernetes provides StatefulSet to manage stateful services, but it is far from enough to run enterprise stateful applications in practice. For example: how does Zookeeper accomplish leader election, and how does MQ implement configuration hot loading? How to do daily operation and maintenance of the database? Many practitioners resort to operators that manages  pod directly e.g. KubeBlocks,  for specific applications e.g. database, yet  they are not general enough for  other stateful applications.   OpenKruise provides several stateful features that are missing in native StatefulSet, such as in-place resource and volume resizing, progressive Configmap & Secret hot update and container operation channel. Teams from Alibaba and Xiaohongshu will share their lessons to build operators and platforms for general stateful apps and containerize database and middleware with a scale of hundreds of thousands of pods.

Kubernetes提供了StatefulSet来管理有状态服务，但实际上要运行企业级有状态应用还远远不够。例如：Zookeeper如何完成领导者选举，MQ如何实现配置热加载？如何进行数据库的日常运维？许多从业者借助直接管理pod的运营商，例如KubeBlocks，针对特定应用程序，例如数据库，但它们并不足够通用以适用于其他有状态应用程序。 OpenKruise提供了一些在原生StatefulSet中缺失的有状态功能，例如原地资源和卷大小调整，渐进式Configmap和Secret热更新以及容器操作通道。来自阿里巴巴和小红书的团队将分享他们构建运营商和平台以适用于通用有状态应用程序，并将数据库和中间件容器化的经验，规模达数十万个pod。

Sandbox API is released in containerd 1.7 and will be stable in containerd 2.0. It provides a clean way to implement a sandbox oriented container runtime. Container is more a set of API specifications than a single technology now, with the introduction of different kinds of isolation techiques as sandboxes, We need a clear and abstract definition of Sandbox API, to make it easy to integrate different kinds of sandboxing techiniques to become a container runtime. In this sharing, We will: 1. Make an introduction of Sandbox API of containerd, and why we need it. 2. Show how we build our container runtimes based on the Sandobx API and the benefits comes with it. 3. We will show the demostration of different kinds of sandboxed containers created by Kuasar, a container runtime framework based on the new Sandbox API, currently supports sandboxes of VMM, UserMode Kernel, WebAssembly and Runc.

在KubeCon的会议描述中，我们将介绍Sandbox API在containerd 1.7中发布，并将在containerd 2.0中稳定。它提供了一种清晰的方式来实现面向沙箱的容器运行时。随着不同类型的隔离技术（如沙箱）的引入，容器现在更多地是一组API规范，而不是单一技术。我们需要对Sandbox API进行清晰和抽象的定义，以便轻松集成不同类型的沙箱技术，使其成为容器运行时。 在这次分享中，我们将： 1. 介绍containerd的Sandbox API，以及为什么我们需要它。 2. 展示我们如何基于Sandbox API构建我们的容器运行时以及带来的好处。 3. 我们将展示由基于新Sandbox API的容器运行时框架Kuasar创建的不同类型的沙箱容器的演示，目前支持VMM、UserMode Kernel、WebAssembly和Runc的沙箱。

With one year's effort, the kubernetes community has made great progress on final approval of the cluster inventory API project. The project has gained a lot of attention and interest from different companies and open source projects, with many new use cases being explored. This panel discussion brings together maintainers from different multicluster management projects who bootstraps this project. We will share what is cluster inventory API, and how we get there. We will also introduce the ongoing work and emerging use cases on this project, and our vision for the future plan. During the panel discussion, attendees will gain a comprehensive understanding of the use cases, eg, how to support multi-cluster AI workload scheduling using inventory API, and challenges, eg how to migrate a cluster manager tool to another seamlessly. We will shed light on the collaborative efforts to standardize cluster inventory APIs and how it evolves from a small group discussion to the community effort.

经过一年的努力，Kubernetes社区在最终批准集群清单API项目方面取得了巨大进展。该项目受到了不同公司和开源项目的关注和兴趣，许多新的用例正在被探索。本次小组讨论将汇集来自不同多集群管理项目的维护者，他们启动了这个项目。我们将分享什么是集群清单API，以及我们是如何实现的。我们还将介绍该项目的正在进行的工作和新兴用例，以及我们对未来计划的愿景。在小组讨论期间，与会者将全面了解用例，例如如何使用清单API支持多集群AI工作负载调度，以及挑战，例如如何无缝迁移集群管理工具。我们将阐明协作努力以标准化集群清单API，并介绍它是如何从一个小组讨论演变为社区努力的。

In the cloud-native realm, eBPF's versatility has led to scalable solutions in observability and security by attaching to system event checkpoints without kernel code modification. This concept has paved the way for extending business applications non-invasively and flexibly without altering the original code. In this session, we'll use Harbor, the cloud-native artifact registry, to showcase how WASM (WebAssembly) extends Harbor's functionalities without code modification. Here, Harbor is analogous to the Linux kernel, and WASM to user-provided eBPF programs. Harbor provides mounting points for various events, such as pre-pull requests, enabling users to filter requests with custom WASM programs. This facilitates fine-grained permission control and artifact security auditing before a user pulls the artifacts, with more features to discover.

在云原生领域，eBPF 的多功能性使得它能够通过附加到系统事件检查点而无需修改内核代码，从而实现可扩展的可观测性和安全性解决方案。这一概念为在不改变原始代码的情况下非侵入性和灵活地扩展业务应用程序铺平了道路。 在本场演讲中，我们将使用 Harbor，云原生制品注册表，展示如何使用 WASM（WebAssembly）在不修改代码的情况下扩展 Harbor 的功能。在这里，Harbor 类似于 Linux 内核，而 WASM 则类似于用户提供的 eBPF 程序。Harbor 提供了各种事件的挂载点，例如预拉取请求，使用户能够使用自定义的 WASM 程序过滤请求。这有助于在用户拉取制品之前进行细粒度的权限控制和制品安全审计，还有更多功能等待您去发现。

In JDCloud, the federated Serverless service is based on the federated management model and Serverless application model, providing JDOS application container control services for federated application container deployment, elastic scaling, and fault migration capabilities. It manages multiple clusters with over 10,000 nodes. Unify management of multiple sub-clusters to improve overall resource utilization. Reduce the complexity of multi-cluster management, scheduling, and distribution on the platform. End users can use our platform just like the native Kubernetes API. Throughout the process, we will address numerous technical challenges, including: 1. Multi-cluster management and distribution practice 2. Efficient cross-cluster elastic scaling solution 3. Problems encountered in production and sharing

在京东云中，联邦Serverless服务基于联邦管理模型和Serverless应用模型，为联邦应用容器部署、弹性扩展和故障迁移提供JDOS应用容器控制服务。它管理超过10,000个节点的多个集群。统一管理多个子集群，提高整体资源利用率。减少平台上多集群管理、调度和分发的复杂性。最终用户可以像使用本机Kubernetes API一样使用我们的平台。在整个过程中，我们将解决许多技术挑战，包括： 1. 多集群管理和分发实践 2. 高效的跨集群弹性扩展解决方案 3. 在生产和分享中遇到的问题