Improve incremental speed 7. Switch to use a Maven plugin Download and install Docker Order of layers to optimize for cache hits github.com/GoogleContainerTools/jib What did we do? 1. Write first Improve incremental speed 7. Switch to use a Maven plugin Download and install Docker Order of layers to optimize for cache hits Use of multi-stage builds github.com/GoogleContainerTools/jib What Improve incremental speed 7. Switch to use a Maven plugin Download and install Docker Order of layers to optimize for cache hits Use of multi-stage builds Understanding Docker cache mechanism and quirks

where secrets are used vs managed Encryption at different layers (or turtles) disks file system etcd Recommendation: Use two-layers of encryption, e.g., full-disk & application-layer … then tries secret store Kubernetes secrets: summary ● Use encryption based on your threat model, e.g., two layers, like full-disk + application-layer ● Rotate keys regularly to limit the impact of a potential

er Efficient and reliable container release Ø Why not RollingUpdate ? Ø What are the advantages of batch gray release ? • more reliable and better control • More flexible • More efficient StatefulSetPlus

and test ML experiments with Helm • Manage training jobs and pipelines with Argo • Serve trained models for inference with TF Serving • Rapid prototyping with self-service Jupyter notebook from JupyterHub • JupyterHub to create and manage interactive Jupyter notebooks • Model serving – serve exported models with TF Serving or Seldon • Additional components for storage, workflow, etc. Artificial Intelligence Classic DevOps solutions: • Containers • CI/CD • Autoscaling • A/B testing and canary release of Models • Comparing Production accuracy vs expected accuracy when possible • Rolling-updates • … Resources

? ? ? Kubernetes Kops Kubernetes Operations Kubernetes Kops Kubernetes Operations “Kops models a cluster then realizes it in a cloud” - Kris Nova Kubernetes Kops Kubernetes Operations Statically dealing with YAML Kops 1.4 And we had a lot of it.. Kops 1.4 $: git checkout tags/v1.4.4 $: cd upup/models/cloudup $: find . | xargs wc -l | grep total total 1076 The next release Kops 1.5 wanted new

30% ������ �� ������ ������� ������ �� ��� ���� �������� �� ����������� Scheduling Isolation Efficient placement of service container and tasks When placed together, don’t affect each other Resource

functionality and maturity • Many scheduling algorithms • Reuse IPVS timer to reclaim eBPF map • More efficient and secure • V.S Taobao IPVS SNAT patch • Bypass conntrack/iptables • Less modification to kernel

On-premise Middleware / PaaS ● ● Cloud Native Hybrid ● ● ● Up Agile Portable Elastic Transform Efficient Secure Pay for use Out ● ● Migrate Move up, out or both? Google is a recognized leader in Open

like? Replica Config LoadBalancer Rolling Update Quota Docker Image Volume RBAC Kubernetes models applications How about Kubernetes itself? How about the fleet Kubernetes running on? Onboard