Model and Operate Datacenter by Kubernetes at eBay 辛肖刚, Cloud Engineering Manager, ebay 梅岑恺, Senior Operation Manager, ebay Agenda About ebay Our fleet Kubernetes makes magic at ebay Model + Controller Controller How we model our datacenter Operation in large scale Q&A About ebay 177M Active buyers worldwide $22.7B Amount of eBay Inc. GMV $2.6B Reported revenue 62% International revenue 1.1B b e Converge & Reconcile WIRI: What it really is Kubernetes Core concept of Kubernetes - Declarative magic What is an application looks like? Replica Config LoadBalancer Rolling Update Quota

The Future of Cloud Native Applications with Open Application Model (OAM) and Dapr @markrussinovich Application models Describes the topology of your application and its components The way developers services and data stores Programming models Distributed Application Runtime (Dapr) Open Application Model (OAM) https://oam.dev State of Cloud Native Application Platforms Kubernetes for applications of concerns Application focused Application focused Container infrastructure Open Application Model Service Job Namespace Secret Volume Endpoint ConfigMap VolumeAttach CronJob Deployment

transformations specifying operations that process input streams and produce output streams. • Declarative languages specify the expected results of the computation rather than the execution flow. • selecting portions of a stream. • relation-to-stream: create streams through querying tables Declarative language: CQL 4 Vasiliki Kalavri | Boston University 2020 Select IStream(*) From S1 [Rows 5] be expressed using only non-blocking operators? 22 Vasiliki Kalavri | Boston University 2020 Model and formalization (I) A stream is a sequence of unbounded length, where tuples are ordered by their

Processing Systems Design Issues ▶ Continuous vs. micro-batch processing ▶ Record-at-a-Time vs. declarative APIs 3 / 79 Outline ▶ Spark streaming ▶ Flink 4 / 79 Spark Streaming 5 / 79 Contribution Contribution ▶ Design issues • Continuous vs. micro-batch processing • Record-at-a-Time vs. declarative APIs 6 / 79 Spark Streaming ▶ Run a streaming computation as a series of very small, deterministic batch Built on the Spark SQL engine. ▶ Perform database-like query optimizations. 56 / 79 Programming Model (1/2) ▶ Two main steps to develop a Spark stuctured streaming: ▶ 1. Defines a query on the input

Kubernetes Introduction § K8s is a production- grade container orchestration platform § Declarative management of objects using configuration files. § More introductions, go to • K8s official Open Tech Mini Academy @ IBM http://ibm.biz/opentech-ma Kubernetes Resource Model A common resource model can satisfy any deployment requirements § Config Maps § Daemon Sets § Deployments openwhisk-deploy-kube Technical details Deployment • A Deployment controller provides declarative updates for Pods and ReplicaSets. • Stands for a long running task, can be exposed as K8s

深度学习框架 – MXNet 概述 MXNet • 节省以及资源效率 • 工程中廉价的GPUs、较小的内存以及网络的限制 • 速度 • 线性的扩展能力 • 简单 • 混合了声明式（declarative）和命令式()代码的特点 为什么选择 MXNet ？ MXNet: 可扩展的深度学习框架 MXNet 框架的特点 命令式 NDArray API 声明式 Symbolic Executor Understanding Book Flight London Utterances Flight booking London Heathrow Intent / Slot model London Heathrow Origin Destination Departure Date Flight Booking “Book a flight to London” Understanding Book Flight London Utterances Flight booking London Heathrow Intent / Slot model London Heathrow Location Location Seattle Origin Destination Departure Date Flight Booking

orchestration • Automating deployment, scaling, and management of containerized applications • Declarative • Can be a mix of GPU or CPU nodes • Massive Scale • OpenAI dedicates up to 10k cores for a PyTorch, MXNet, Chainer, and more • JupyterHub to create and manage interactive Jupyter notebooks • Model serving – serve exported models with TF Serving or Seldon • Additional components for storage, workflow Demo: Run TensorFlow Training with Containers Demo: Serving the Model with TF Serving • Options for serving • Wrap model in a web framework (eg – Flask) • Tensorflow Serving • Seldon Demo:

contributors can get involved in the SIG. Kubernetes is in the process of moving to a new “out of tree” model, this effort spans all the touching points with the underlying infrastructure: compute, storage, have independent feature and patch release cycles, learn how SIG VMware is working to meet this new model on VMware platforms. Agenda 4 What is the VMware SIG Purpose, Projects managed, How to join Foundry. Cluster API provider for vSphere • The Cluster API is a Kubernetes project to bring declarative, Kubernetes-style APIs to cluster creation, configuration, and management. It provides optional

● Dial-tone security with Trust Domain ● L7 policy enforcement Step 1 Step 2 Step 3 Step 4 Declarative Intent Replace Hardware AZ Architecture Evolving Security Current Status #IstioCon Step ● Capture Traffic Management & Routing intent as “Access Point” Specs ○ Leverage Istio object model: Gateway, VirtualService, DestinationRules, etc. apiVersion: apps.cloud.io/v1 kind: AccessPoint

利用率並最終降低總體成本的工具。 Click to edit Master title style 13 GitOps 痛苦x甜密 • 有能力記錄叢集環境上的一切變化 • 使用宣告式(Declarative)的文件格式 來描述或是設定環境上要用到的所有 資源 • 所有的環境變化都可支援審核機制， 要通過審核才會往下運作 • 權限控管，控制誰有能力去對環境資 源進行更改 • 有辦法針對期望的狀態與運行的狀態 的好朋友 – xxxOperator • Operator 的目標是將 operation 知識 放入軟件中 • Operator 運行在 Kubernetes 集群內 並根據宣告式 (Declarative) 的 CRD 文件來自動化常見的 Day 1和 Day2 的活動。 15 Click to edit Master title style 16 Kube-Prometheus-stack