When to use Serverless? When to use Kubernetes?

When to use Serverless?
When to use Kubernetes?
Niklas Heidloff
Developer Advocate, IBM
@nheidloff
heidloff.net
H1 / 2018

Agenda, Disclaimers and Credits
Open Source solutions only
Focus is on development related topics
Focus is on compute, not on storage
Thanks a lot to Ansgar Schmidt for the preparation of this
session which we gave together at OOP 2018, Munich !
@nheidloff
Code
Intro
Test
Deploy
Run
Summary

Serverless
“Serverless computing is a cloud computing execution model in
which the cloud provider dynamically manages the allocation of
machine resources. Pricing is based on the actual amount of
resources consumed by an application.” (via Wikipedia)
Sample scenario: Serverless Web Application
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Apache OpenWhisk
“Apache OpenWhisk is a serverless, open source cloud platform
that executes functions in response to events without developer
concern for managing the lifecycle or operations of the
containers that execute the code.” (via Apache)
Supported runtimes: Node.js, Python, PHP, Swift and Docker
Deployment: On-premises or in the cloud
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Kubernetes
“Kubernetes is an open-source system for automating
deployment, scaling, and management of containerized
applications.” (via kubernetes.io)
Open source projected initiated by Google and governed by
the Cloud Native Computing Foundation
Main functionality: Self-healing, automatic binpacking, service
discovery and load balancing, horizontal scaling, automated
rollouts
Offered as computing option by all major cloud providers
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Istio
Code
Intro
Test
Deploy
Run
Summary
“An open platform to connect, manage, and
secure microservices.” (via istio)
Open source projected initiated by IBM and Google and
governed by the Cloud Native Computing Foundation
Main functionality: Trafﬁc
management, observability,
policy enforcement, service
identity
@nheidloff

Code
Intro
Test
Deploy
Run
Summary
Kubernetes
Local development via standard tooling and IDEs
Minikube to run Kubernetes locally
Docker support for Kubernetes (beta)
@nheidloff

Serverless
Supported runtimes: Node.js, Python, PHP, Swift and Docker
Development
Local development via standard tooling and IDEs
Web tooling provided by cloud providers
Orchestration via Composer
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 1 – Serverless APIs
Building simple CRUD APIs via Node.js
GET https://..../person?id=x
Types of functions
Single file (synchronous and asynchronous)
Multi-file functions
Docker functions
Debugging single file Nodes.js functions
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 1 – Function Code
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 1 – Asynchronous Function (successful)
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 1 – Multiple File Function
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 1 – Debugging simple Node.js Function
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 2 – Debugging Docker
Developing serverless functions via Docker
Debugging containers locally
For both Serverless and Kubernetes
With Visual Studio Code
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 2 – Nodes.js Function in Docker
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 2 – Debugging Nodes.js Function in Docker
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Composer
Composer is a new programming model for composing
functions, built on Apache OpenWhisk
Declarative and programmatic
approach to deﬁne sophisticated
ﬂows between functions in
cloud-native apps
State management between
‘stateless’ functions
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Code
Intro
Test
Deploy
Run
Summary
Kubernetes
Testing can be done via standard testing tools and services
Istio provides capabilities to easily do A/B testing to test new
features in production environments
@nheidloff

Code
Intro
Test
Deploy
Run
Summary
Demo 3- A/B Testing
Book reviews sample application
https://istio.io/docs/guides/bookinfo.html
Deployments of new service versions to subsets of users
@nheidloff

Demo 3 – Book Info App
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 3 – Book Info App: Different Review Versions
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Code
Intro
Test
Deploy
Run
Summary
Demo 3 - A/B Testing: 3 Versions of Review Service
@nheidloff

Code
Intro
Test
Deploy
Run
Summary
Demo 3 - A/B Testing: Version 2 only for Jason
@nheidloff

Code
Intro
Test
Deploy
Run
Summary
Demo 3 - A/B Testing: Version 3 for 50% of Users
@nheidloff

Serverless
No good way to do A/B Testing
Tests cause costs when running functions in the cloud
Testing doesn’t become easier automatically
“Functions” doesn’t mean necessarily “Pure Functions”
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Code
Intro
Test
Deploy
Run
Summary
Serverless
Deployments can be automated via CD/CI pipelines
No notion of staging environments for functions in the cloud
Serverless Framework: Toolkit for deploying and operating
serverless architectures
@nheidloff

Code
Intro
Test
Deploy
Run
Summary
Kubernetes
Deployments can be done via kubectl CLI
Deployments can be automated via CD/CI pipelines
Helm is a tool for managing Kubernetes
charts which are packages of pre-conﬁgured
Kubernetes resources
@nheidloff

Demo 4 – Deployments
Code
Intro
Test
Deploy
Run
Summary
Deployment of a Function via wsk CLI
Deployment of Kubernetes applications
Via kubectl CLI
Via Helm
@nheidloff

Demo 1 – Deployment of Function
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 1 – Function Invocation via CLI
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 4 – Function Invocation via HTTP
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 4 – Deployment of nginx on Kubernetes
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 4 – Kubernetes Scalability
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Demo 4 – Jenkins Installation via Helm
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Serverless
Limited out of the box monitoring right now, especially for
Composer ﬂows
Monitoring via Grafana
Serverless doesn’t always mean administrator-less
On-premises deployments require operations
Management of Docker containers, e.g. vulnerabilities
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Kubernetes
Lots of monitoring tools available
Part of Kubernetes: kubectl proxy
Community and third parties: Grafana, Zipkin and many more
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Kubernetes Dashboard (kubectl proxy)
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Kubernetes: Grafana
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Kubernetes: Zipkin
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Kubernetes: ServiceGraph
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Summary
Use Serverless for ...
Variable and irregular loads (can lead to lower costs)
Developer productivity for ﬁrst time experience
Inherent auto-scalability
Use Kubernetes for ...
Mature deployment options, e.g. A/B testing, monitoring
Minimal response latency
High performance computing without resource restrictions
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Resources
OpenWhisk
https://openwhisk.apache.org
https://github.com/ibm-functions/composer
https://github.com/nheidloff/openwhisk-debug-nodejs
Kubernetes
https://kubernetes.io
https://istio.io
IBM Cloud
http://ibm.biz/oop2018
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Abstract
There is a lot of debate whether to use Serverless or Kubernetes
to build cloud-native apps. Both have their advantages and
unique capabilities which developers should take into
consideration when planning new projects. We will throw some
light on the topics ease of use, maturity, types of scenarios,
developer productivity and debugging, supported languages,
DevOps and monitoring, performance, community and pricing.
Cloud-native architectures shift the complexity from within an
application to orchestrations of Microservices. Both Kubernetes
and Serverless have their strengths which we will discuss.
Besides the core development topics, developers should also
understand operational aspects how complicated it is to
maintain your own systems versus using managed platforms.
@nheidloff

API Management with IBM Cloud
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

Serverless – Demo 1
Code
Intro
Test
Deploy
Run
Summary
@nheidloff

When to use Serverless? When to use Kubernetes?

Empfohlen

Empfohlen

Weitere ähnliche Inhalte

Was ist angesagt?

Was ist angesagt? (20)

Ähnlich wie When to use Serverless? When to use Kubernetes?

Ähnlich wie When to use Serverless? When to use Kubernetes? (20)

Mehr von Niklas Heidloff

Mehr von Niklas Heidloff (20)

Kürzlich hochgeladen

Kürzlich hochgeladen (20)

When to use Serverless? When to use Kubernetes?