docker images # REPOSITORY TAG IMAGE ID CREATED SIZE # hello-node v1 353fc3541cea 2 minutes ago 912MB 创建 Deployment Kubernetes Pod 是⼀个或多个容器组合在⼀起得共享资源，Kubernetes Deployment

rbind��� /mnt/c! 1. ��A� RO �����RBD volume: /var/lib/kubelet/plugins/xxxx! 2. node-exporter � make-private ����������! 3. ��A������RBD volume����rbd umap����node-exporter� � RO ��! 4. ��B� RW �������RBD K8S�������������! ����! • node-exporter! • node-exporter�K8S daemonset������������������! • K8S���mount make-private��������volume��������! • ceph��! • ceph rbd client����������! • ��lsblk����������! • K8S ��������������! • K8S! • ��K8S�volume���������PR����! • node-exporter���������������make-shared������! • K8S�����mount —rbind —make-private���5�����PR���8������! ����! ������! AI�����! AI�������! ����K8S

environments such as battlefields. The an- alytics enabled and performed by Smart- Edge allow battalions to make real-time, data-driven decisions which dramatically improve operational outcomes and in- crease Capitalizing on open source solutions like K3s, Booz Allen’s SmartEdge solution allows bat- talions to make real-time, data-driven decisions which dramatically improve operational outcomes and increase the For the team, the priority is to deploy SmartEdge in as many scenarios as pos- sible, where it can make a real and tangible difference. From an innovation point-of- view, the digital solutions business

Suggestion: • Run the test against the same cluster during near time • Make CPU the bottleneck • 1 CPU handles 500,000 pps • Make LB the bottleneck • The target server and client server shall be powerful

Kubernetes Take control Easy to manage and scale Experience KOPS: Installation + Create AWS user + Make a CI host + wget and install kops and kubectl + Configure AWS cli + Create S3 Bucket + Export

sequential: huge time saver for large trainings Kubeflow • Machine Learning Toolkit for Kubernetes • To make ML workflows on Kubernetes simple, portable, and scalable • Training controllers – simplify and

is to create a tool that is easy to use and scheduler agnostic. Technical Issues • Trying to make a scheduler-agnostic deployment tool is hard and can lead to abstraction that’s awkward for one

– Probe: readiness probe and liveness probe Component Deployment Topology • Use affinity to make deployment topology policies for different component. E.g. controller node and DB node may not

embed in .go file import “text/template” parse at runtime How do we develop ? ? ? No Yes make go-bindata go build ./kops change work? Kops 1.4 fin Kops 1.4 We were really good at dealing

Annotator scores node periodically. Ø Dynamic-Scheduler query node score directly from node to make decisions during predicate and priority stage. Dynamic Scheduler Kube-scheduler Dynamic scheduler