stackconf 2020 | The path to a Serverless-native era with Kubernetes by Paolo Mainardi
0 likes133 views
The document discusses the concept of serverless computing and its implementation on Kubernetes, emphasizing the developer experience and cost efficiency of serverless architectures. It introduces key components like Functions-as-a-Service (FaaS) and Backend-as-a-Service (BaaS), and highlights tools such as Knative for managing serverless workloads on Kubernetes. Additionally, it covers the flexibility and scalability of serverless solutions, as well as the convergence of serverless platforms with Kubernetes in cloud environments.
![Knative - Openshift serverless
Openshift Serverless
1. GA product since 30rd of April
2. Based on Knative Serving + Eventing + Tekton pipeline
3. Fully integrated with the Openshift ecosystem.
Pivotal Riff
1. Deep Kubernetes integration using CRDs
2. Pluggable application runtimes [Core, Knative...]
3. A custom builder based on buildpacks.io (source to container)
4. A custom cli tool “riff” to manage the entire lifecycle](https://image.slidesharecdn.com/paolomainardi-thepathtoaserverless-nativeerawithkubernetes-200626065424/85/stackconf-2020-The-path-to-a-Serverless-native-era-with-Kubernetes-by-Paolo-Mainardi-69-320.jpg)
stackconf 2020 | The path to a Serverless-native era with Kubernetes by Paolo Mainardi
- 1. The path to a Serverless-native era with Kubernetes
- 2. Who i am Paolo Mainardi @paolomainardi - https://www.paolomainardi.com ➔ ➔ CTO of Sparkfabrik - https://www.sparkfabrik.com/en/ ➔ CKA(D) certified, GCP Cloud Architect ➔ We are hiring :)
- 3. The session ● What is the hype cycle ● Serverless concepts (FaaS, BaaS, *aaS) ● Serverless on Kubernetes
- 7. WHAT IS SERVERLESS ?
- 8. “Serverless computing refers to the concept of building and running applications that do not require server management” CNCF Serverless Whitepaper v1.0 - https://github.com/cncf/wg-serverless/tree/master/whitepapers/serverless-overview
- 9. 1. Functions-as-a-Service (FaaS) which typically provides event-driven computing. Developers run and manage application code with functions that are triggered by events or HTTP requests. Developers deploy small units of code to the FaaS, which are executed as needed as discrete actions, scaling without the need to manage servers or any other underlying infrastructure. 2. Backend-as-a-Service (BaaS) which are third-party API-based services that replace core subsets of functionality in an application. Because those APIs are provided as a service that auto-scales and operates transparently, this appears to the developer to be serverless. Serverless computing platform
- 10. SERVERLESS IS ALL ABOUT THE DEVELOPER EXPERIENCE
- 12. 1. Cost and resource efficiency — scale down to zero 2. Pay as you go — scale up on demand 3. Automation — of scaling, failure handling, and recovery SERVERLESS IS ALL ABOUT THE DEVELOPER EXPERIENCE©
- 13. FaaS Functions as a Service
- 14. Use case - Translation service FaaS - Translation service Business requirement Translate text files from Italian to German, both on-demand and batch uploads. Non-functional requirements: ● Scale up on demand ● Highly available >= 99.5% ● Scale down to zero ● Pay per use
- 15. Use case - Translation service FaaS - Translation service Architecture overview ● NodeJS ● Google Cloud Platform ○ Cloud services ○ Translate API
- 16. Translation service implementation FaaS - Translation service
- 17. Non-functional requirements: ● Scale up on demand ● Highly available >= 99.5% ● Scale down to zero ● Pay as you go FaaS - Cloud architecture #0
- 18. FaaS - Cloud architecture #1 Non-functional requirements: ● Scale up on demand ● Highly available >= 99.5% ● Scale down to zero ● Pay as you go
- 19. Non-functional requirements: ● Scale up on demand ❌ ● Highly available >= 99.5% ❌ ● Scale down to zero ❌ ● Pay as you go ❌ FaaS - Cloud architecture #1 Service Instances Cost per month Compute engine (n1-standard-1) 1 30$ HTTP(s) load balancer 1 18$ Fixed costs TOTAL: 48$
- 20. FaaS - Cloud architecture #2 Non-functional requirements: ● Scale up on demand ✅ ● Highly available >= 99.5% ❌ ● Scale down to zero ❌ ● Pay as you go ❌ Service Instances Cost per month Compute engine (n1-standard-1) 1 30$ HTTP(s) load balancer 1 18$ Fixed costs TOTAL: 48$
- 21. FaaS - Cloud architecture #3 Non-functional requirements: ● Scale up on demand ✅ ● Highly available >= 99.5% ✅ ● Scale down to zero ❌ ● Pay as you go ❌ Service Instances Cost per month Compute engine (n1-standard-1) 1 90$ HTTP(s) load balancer 1 18$ Fixed costs TOTAL: 108$
- 22. FaaS - Functions as a service TRADITIONAL SERVERLESShttps://cloud.google.com/functions/docs/writing Serverless refactoring
- 23. FaaS - Serverless Cloud architecture Non-functional requirements: ● Scale up on demand ✅ ● Highly available >= 99.5% ✅ ● Scale down to zero ✅ ● Pay as you go ✅ Service Instances Cost per month - - - Fixed costs TOTAL: 0$
- 24. Credits: https://www.gocd.org/2017/06/26/serverless-architecture-continuous-delivery/ Key properties of FaaS: ● Stateless and ephemeral ● Scalable from 0 to N ● Event-triggered ● Fully managed by a third party FasS - Functions as a service
- 25. Functions on public cloud The state of FaaS on public cloud ● Limited language runtimes (NodeJS, Python, Java, C#) ● Vendor-proprietary mechanisms to package and deploy the functions
- 26. ● Locked-in language runtime ● Proprietary Event format ● The state of FaaS on public cloud FaaS on public cloud
- 27. LOCK-IN
- 28. Lock-in Lockin use-case Convert CSV files to JSON on GCP {JSON}
- 29. Lock-in Convert CSV files to JSON coming from multiple cloud providers {JSON}
- 30. Lock-in Convert CSV files to JSON coming from multiple cloud providers {JSON}
- 32. Lock-in {JSON}
- 33. Lock-in {JSON}
- 34. CloudEvents CLOUDEVENTS IS A SPECIFICATION FOR DESCRIBING EVENT DATA IN A COMMON WAY
- 35. CloudEvents
- 36. CloudEvents CloudEvents on public cloud https://azure.microsoft.com/de-de/blog/announcing-first -class-support-for-cloudevents-on-azure/ Contributors https://github.com/cloudevents/spec/blo b/master/community/contributors.md Azure Event Grid
- 37. WHAT IS KUBERNETES ?
- 38. What is Kubernetes What is Kubernetes https://kubernetes.io/docs/concepts/overview/what-is-kubernetes/ Kubernetes is a portable, extensible, open-source platform for managing containerized workloads and services, that facilitates both declarative configuration and automation. The name Kubernetes originates from Greek, meaning helmsman or pilot. Google open-sourced the Kubernetes project in 2014. Kubernetes combines over 15 years of Google’s experience running production workloads at scale with best-of-breed ideas and practices from the community.
- 39. What is Kubernetes What is Kubernetes
- 40. What is Kubernetes - Kubelet Kubelet An agent that runs on each node in the cluster. It makes sure that containers are running in a Pod
- 41. What is Kubernetes - Pods Pods A Pod is a group of one or more containers with shared storage/network, and a specification for how to run the containers
- 43. Serverless & Kubernetes Developer experience on Kubernetes Have to do Want to do Write code
- 44. Serverless & Kubernetes Developer experience on Kubernetes Have to do Want to do Write code Build docker image
- 45. Serverless & Kubernetes Have to do Want to do Write code Build docker image Push docker image to registry Developer experience on Kubernetes
- 46. Serverless & Kubernetes Have to do Want to do Write code Build docker image Push docker image to registry Write yaml manifests Developer experience on Kubernetes
- 47. Serverless & Kubernetes Have to do Want to do Write code Build docker image Push docker image to registry Write yaml manifests kubectl apply -f myapp/ Developer experience on Kubernetes
- 48. Serverless & Kubernetes Have to do Want to do Write code Write code Build docker image Push docker image to registry Write yaml manifests kubectl apply -f myapp/ Developer experience on Kubernetes
- 49. Serverless & Kubernetes Have to do Want to do Write code Write code Build docker image Ship it Push docker image to registry Write yaml manifests kubectl apply -f myapp/ Developer experience on Kubernetes
- 50. Serverless & Kubernetes Developing on Kubernetes is not easy 1. It’s not the right abstractions for end-developer experience (eg: deployment, service, ingress, hpa, monitoring) 2. But it’s a great platform to build a PaaS on top of it: Primitives, Extensibility
- 51. BUILD A SERVERLESS DEVELOPMENT EXPERIENCE ON KUBERNETES
- 52. ● Something to wake-up workloads on request ● Something to scale-up and scale back to zero ● Something to turn the source code into a container ● An eventing system with configurable sources and subscribers Serverless on Kubernetes to-do list:
- 53. Knative Kubernetes-based platform to build deploy and manage modern serverless workloads. Knative offers features like scale-to-zero, autoscaling and eventing framework https://knative.dev KNATIVE
- 56. Knative Knative serving 1. Scale-up and scale back to zero 2. Fine-grained traffic routing 3. Automatic metrics and telemetry 4. Point-in-time snapshot of code and configurations (revisions)
- 58. Knative Knative eventing 1. A complete eventing system for event-driven apps. 2. Easy for developers to consume events (CloudEvents spec) 3. Pluggable, Customizable Event Sources.
- 61. Knative Knative build “Deprecating Knative Build would resolve this confusion as it would make it clear that the "building the service(s) image" part is not Knative's responsability and user can use whatever tools is capable of doing that — preferably Tekton Pipelines 👼.”
- 62. Knative Tekton 1. The Tekton project provides Kubernetes-style resources for declaring CI/CD-style pipelines 2. Builds are ran completely within Kubernetes (eg: Kaniko, Buildkit) and using standard formats like buildpacks 3. Openly-governed under the Continuous Delivery Foundation (CDF)
- 63. Knative ● As a developer: Use it directly to deploy your stuff in a more streamlined way than raw Kubernetes. What you can do with Knative
- 64. Knative ● As a developer: Use it directly to deploy your stuff in a more streamlined way than raw Kubernetes. ● As an operator: Put a level of abstractions between Kubernetes and your devs; Keep the costs under control thanks to auto-scaling features. What you can do with Knative
- 65. Knative ● As a developer: Use it directly to deploy your stuff in a more streamlined way than raw Kubernetes. ● As an operator: Put a level of abstractions between Kubernetes and your devs; Keep the costs under control thanks to auto-scaling features. ● As a platform architect: Build your own PaaS/FaaS platform, relying on the Knative primitives. What you can do with Knative
- 67. Knative - Google Cloud Run Google CloudRun 1. Fully GCP managed Serverless platform based on Knative apis 2. Event-driven architecture (HTTP + Pub/Sub) 3. Cloud Run + Cloud run for Anthos 4. A sort of managed Knative + Istio Credits: https://twitter.com/ahmetb/status/1116041166359654400
- 68. Knative - Google Cloud Run Google CloudRun 1. Fully GCP managed Serverless platform based on Knative apis 2. Event-driven architecture (HTTP + Pub/Sub) 3. Cloud Run + Cloud run for Anthos 4. A sort of managed Knative + Istio Credits: https://ahmet.im/blog/cloud-run-is-a-knative/
- 69. Knative - Openshift serverless Openshift Serverless 1. GA product since 30rd of April 2. Based on Knative Serving + Eventing + Tekton pipeline 3. Fully integrated with the Openshift ecosystem. Pivotal Riff 1. Deep Kubernetes integration using CRDs 2. Pluggable application runtimes [Core, Knative...] 3. A custom builder based on buildpacks.io (source to container) 4. A custom cli tool “riff” to manage the entire lifecycle
- 71. Keda Keda is an open-source CNCF sandbox project from Microsoft and Redhat KEDA provides a lightweight auto-scaling infrastructure that allows to autoscale applications based on custom criteria KEDA - Kubernetes Event Driven Autoscaling https://keda.sh
- 72. Keda KEDA
- 73. Keda KEDA Custom metrics server for HPA
- 74. Keda KEDA ➔ Watching for new ScaledObjects ➔ In charge of scaling from 0 to 1 or vice-versa, from 1 to N remains in charge the standard HPA
- 75. Keda KEDA
- 76. Kubernetes-native serverless platforms ➔ OSS platforms ➔ Flexible language runtimes ➔ Custom CLIs ➔ Written in GO (OpenWhisk in Scala) ➔ Backed by big vendors (Oracle, IBM, Vmware) Kubernetes fully fledged serverless platforms
- 77. VIRTUAL KUBELET
- 78. Virtual Kubelet Virtual Kubelet Virtual Kubelet is an open-source Kubernetes kubelet implementation that masquerades as a kubelet. This allows Kubernetes nodes to be backed by Virtual Kubelet providers such as serverless cloud container platforms https://virtual-kubelet.io
- 81. Virtual Kubelet Virtual Kubelet virtual kubelet Virtual kubelet register itself as a “node” and allows developers to deploy pods and containers with their own apis.
- 82. Virtual Kubelet Virtual Kubelet virtual kubelet AZURE CONTAINER SERVICE HASHICORP NOMAD AWS FARGATE ALIBABA CLOUD ECI YOUR CUSTOM PROVIDER
- 83. Serverless & Kubernetes Serverless on Kubernetes:
- 84. Serverless & Kubernetes Serverless on Kubernetes: ● Flexible language runtimes based on containers
- 85. ● Flexible language runtimes based on containers ● Run on multi/hybrid cloud and on your laptop Serverless & Kubernetes Serverless on Kubernetes:
- 86. Serverless & Kubernetes Serverless on Kubernetes: ● Flexible language runtimes based on containers ● Run on multi/hybrid cloud and on your laptop ● Mix and match stateless and stateful workloads
- 87. Serverless & Kubernetes Serverless on Kubernetes: ● Flexible language runtimes based on containers ● Run on multi/hybrid cloud and on your laptop ● Mix and match stateless and stateful workloads ● Extensibility and a better development experience
- 89. In the cloud platforms, Serverless & k8s will essentially converge and we’ll deploy just by asking: “Kubernetes, just run my code” Alexis Richardson, Founder and CEO, Weaveworks