PyData London 2015 - Getting Started with Cloud Foundry for Data Science

Slide 1

Slide 1 text

Getting Started with Cloud Foundry •  Ian Huston, Data Scientist Ian Huston, Data Scientist PyData London 2015

Slide 2

Slide 2 text

2 © Copyright 2015 Pivotal. All rights reserved. Who am I?   Ian Huston   @ianhuston   www.ianhuston.net   Talk resources:  https://github.com/ihuston/ python-cf-examples   Data Scientist   Use PyData stack for predictive analytics and machine learning   Previously a theoretical physicist using Python for numerical simulations & HPC

Slide 3

Slide 3 text

Slide 4

Slide 4 text

4 © Copyright 2015 Pivotal. All rights reserved. Aims for this tutorial Goals:   Understand how to use CF as a developer/data scientist   Understand how to push PyData apps and bind to services Anti-goals:   Understand CF from an operations viewpoint   Install & run a full Cloud Foundry installation

Slide 5

Slide 5 text

5 © Copyright 2015 Pivotal. All rights reserved. Plan   What is Cloud Foundry?   Getting started with a hosted CF instance   Pushing your ﬁrst app   What are buildpacks?   Using PyData packages   Using data services like Redis   Putting it all together

Slide 6

Slide 6 text

Slide 7

Slide 7 text

What is Cloud Foundry? http://cloudfoundry.org Open Source Multi-Cloud Platform Simple App Deployment, Scaling & Availability

Slide 8

Slide 8 text

Cloud Applications Haiku Here is my source code Run it on the cloud for me I do not care how. -  Onsi Fakhouri @onsijoe

Slide 9

Slide 9 text

How can data scientists use CF?

Slide 10

Slide 10 text

Data Services Easy control of incoming data

Slide 11

Slide 11 text

Distributed computation

Slide 12

Slide 12 text

Data Driven Applications

Slide 13

Slide 13 text

Slide 14

Slide 14 text

Slide 15

Slide 15 text

Slide 16

Slide 16 text

16 © Copyright 2015 Pivotal. All rights reserved. Getting Started on Pivotal Web Services   Go to http://run.pivotal.io   Trial lasts for 60 days, no credit card required   Click ‘Sign Up’ and enter your email   Open confirmation email and verify registration   Complete SMS verification check   Choose an organisation name (your name/handle is fine)

Slide 17

Slide 17 text

17 © Copyright 2015 Pivotal. All rights reserved. Downloading the Command Line Interface   Register at http://run.pivotal.io   In the CF management website click ‘Tools’ and download appropriate package   Or go to https://github.com/cloudfoundry/cli   Or use homebrew on OSX: $ brew tap pivotal/tap $ brew install cloudfoundry-‐cli   Provides the cf command, your interface to Cloud Foundry

Slide 18

Slide 18 text

Slide 19

Slide 19 text

19 © Copyright 2015 Pivotal. All rights reserved. Logging in to PWS   Choose PWS API endpoint:   $ cf api https://api.run.pivotal.io   Login: $ cf login   Enter your email address and password   You should see output like:  API endpoint: https://api.run.pivotal.io (API version:2.28.0) User: YOUR_USER_NAME Org: YOUR_ORG_NAME Space: development

Slide 20

Slide 20 text

20 © Copyright 2015 Pivotal. All rights reserved. Challenge 1: Pushing your first application   Choose PWS API endpoint: $ cf api https://api.run.pivotal.io   Login: $ cf login   Code directory: 01-‐simple-‐python-‐app   Deploy with $ cf push   Check it worked at http://myapp-RANDOM-WORDS.cfapps.io   Turn off your app when finished with cf stop myapp (but not yet!)

Slide 21

Slide 21 text

21 © Copyright 2015 Pivotal. All rights reserved. Simple Flask App Demo   Simple one page “Hello World” web app   Video: https://www.youtube.com/watch?v=QOfD6tnoAB8   Demonstrates: –  Installation of requirements –  Scaling properties   Need to Provide: –  App files –  Dependencies listed in requirements.txt file –  Optional manifest.yml file with configuration for deployment

Slide 22

Slide 22 text

C F R O" U" T" E" R 2. Set up domain Cloud Controller Instance 1. Upload code 4. Copy app into containerised instances 3. Install Python & Dependencies 5. Start app and accept connections Send request to URL WHAT JUST HAPPENED? Source Code Instance $ cf push Browser 5. Load balance between instances

Slide 23

Slide 23 text

23 © Copyright 2015 Pivotal. All rights reserved. What just happened? 1.  Application code is uploaded to CF 2.  Domain URL is set up ready for routing 3.  Cloud controller builds application in container: –  Python interpreter selected –  Dependencies installed with pip 4.  Container is replicated to provide instances 5.  App starts and Router load balances requests   See what’s happening using logs: $ cf logs myapp -‐-‐recent

Slide 24

Slide 24 text

24 © Copyright 2015 Pivotal. All rights reserved. Python on Cloud Foundry   First class language (with Go, Java, Ruby, Node.js, PHP)   Automatic app type detection –  Looks for requirements.txt or setup.py   Buildpack takes care of –  Detecting that a Python app is being pushed –  Installing Python interpreter –  Installing packages in requirements.txt using pip –  Starting web app as requested (e.g. python myapp.py)

Slide 25

Slide 25 text

25 © Copyright 2015 Pivotal. All rights reserved. Scaling memory and instances   We can scale our application as needed in terms of memory and number of instances: $ cf scale myapp –i 5 $ cf scale myapp –m 256M   Check app in browser to see diﬀerent ports being used.   Scale back down with $ cf scale myapp –i 1

Slide 26

Slide 26 text

26 © Copyright 2015 Pivotal. All rights reserved. How does this work?   Containerised application is cached and ready to be deployed.   Scaling number of instances replicates container and load balances requests across all instances.   Scaling memory requires restarting app.   Auto-scaling is also possible.

Slide 27

Slide 27 text

Slide 28

Slide 28 text

28 © Copyright 2015 Pivotal. All rights reserved. What are buildpacks?   Idea and format from Heroku   Responsible for doing whatever is necessary to get your app running.   Buildpacks take care of –  Detecting which type of application is being pushed –  Installing the appropriate run-time –  Installing required dependencies or other artifacts –  Starting the application as requested   Oﬃcial buildpacks for Python, Java, Node.js, Go, Ruby, PHP & for static websites and running binaries

Slide 29

Slide 29 text

29 © Copyright 2015 Pivotal. All rights reserved. Containers vs Buildpacks runtime layer OS image application layer Container (e.g. Docker) system brings ﬁxed host OS Kernel * Devs may bring a custom buildpack runtime layer* OS image application layer Buildpack App container System Provides Dev Provides system brings ﬁxed host OS Kernel

Slide 30

Slide 30 text

30 © Copyright 2015 Pivotal. All rights reserved. Custom buildpacks   Instead of using an oﬃcial buildpack you can use any custom buildpack installed on your CF or available on Github.   Only 3 shell scripts needed: – detect – compile – release   Specify buildpack with -‐b or in manifest.yml

Slide 31

Slide 31 text

31 © Copyright 2015 Pivotal. All rights reserved. Community Buildpacks https://github.com/cloudfoundry-community/ cf-docs-contrib/wiki/Buildpacks   Lots of languages:   Clojure, Haskell, .NET, Erlang, Elixir, etc.   RShiny app buildpack:   https://github.com/alexkago/cf-buildpack-r   Can also use some Heroku buildpacks without modiﬁcation.

Slide 32

Slide 32 text

32 © Copyright 2015 Pivotal. All rights reserved. Official Python buildpack ü  Great for simple pip based requirements ü  Well tested and officially maintained ü  Covers both Python 2 and 3 ✗ Suffers from the Python Packaging Problem: -  Hard to build packages with C, C++ or Fortran extensions -  Complicated local configuration of libraries and paths needed -  Takes a long time to build main PyData packages from source

Slide 33

Slide 33 text

33 © Copyright 2015 Pivotal. All rights reserved. Using conda for package management   http://conda.pydata.org   Beneﬁts: –  Uses precompiled binary packages –  No ﬁddling with Fortran or C compilers and library paths –  Known good combinations of main package versions –  Really simple environment management (better than virtualenv) –  Easy to run Python 2 and 3 side-by-side Go try it out if you haven’t already!

Slide 34

Slide 34 text

34 © Copyright 2015 Pivotal. All rights reserved. How to use the conda buildpack https://github.com/ihuston/python-conda-buildpack   Specify as a custom buildpack when pushing app with manifest or -‐b command line option.   Export your current environment to a environment.yml ﬁle   Or write requirements.txt (pip) and conda_requirements.txt   Send me feedback & pull requests!

Slide 35

Slide 35 text

35 © Copyright 2015 Pivotal. All rights reserved. Challenge 2: Pushing a PyData app   Code directory: 02-‐pydata-‐spyre-‐app   Spyre – Adam Hajari https://github.com/adamhajari/spyre   This app uses the Pydata buildpack to install Matplotlib, NumPy and more.   Spyre provides a simple way to build interactive web based visualisations similar to Rshiny.

Slide 36

Slide 36 text

Slide 37

Slide 37 text

37 © Copyright 2015 Pivotal. All rights reserved. Services These are available on Pivotal CF (Pivotal’s packaged Cloud Foundry oﬀering). See http://run.pivotal.io for the services available on Pivotal Web Services.

Slide 38

Slide 38 text

38 © Copyright 2015 Pivotal. All rights reserved. Cloud Native Applications   Suitable for deployment on cloud platforms   Can scale up easily without changes   Follow these rules: http://12factor.net   In a nutshell: –  Apps are stateless processes –  Easy to create and destroy apps with no side-eﬀects –  Persistent state handled through backing services –  Interact with services through port binding

Slide 39

Slide 39 text

39 © Copyright 2015 Pivotal. All rights reserved. Challenge 3: Binding a service to an app   Create a free Redis service: $ cf cs rediscloud 30mb myredis   Bind to our earlier app: $ cf bind-‐service myapp myredis   See how apps ﬁnd credentials in environmental variables: $ cf env myapp

Slide 40

Slide 40 text

40 © Copyright 2015 Pivotal. All rights reserved. Challenge 3: Using Services in an app   Code directory: 03-‐services-‐redis   This app binds to a Redis service if available and counts the number of hits on the app homepage.   Add keys and values by adding /keyname/value to URL.   Data is persisted in Redis, will survive deleting and restarting the app.   Multiple instances all access the same service.

Slide 41

Slide 41 text

User Provided Services How to add User Provided Services: Standalone Hadoop or Apache Spark cluster, " Big Data System, RDBMS etc. $ cf cups SERVICE_INSTANCE -‐p "host, port, username, password" Data Service App App App App App App

Slide 42

Slide 42 text

Slide 43

Slide 43 text

R E S T A P I Send data as JSON Data Ingest Model Create Model Redis Kicking off periodic retraining Save training data Save model object Send JSON data without label Receive prediction from trained model instance Deployed at: http://dsoncf.cfapps.io Code: https://github.com/alexkago/ds-cfpylearning PREDICTION API ARCHITECTURE $ cf create-‐service rediscloud PLAN_NAME INSTANCE_NAME

Slide 44

Slide 44 text

44 © Copyright 2015 Pivotal. All rights reserved. Final challenge: Build your own predictive API   Code directory: 04-‐learning-‐api   Simple Flask + scikit-learn based machine learning API   Push the application and go to the app URL to see instructions on how to use.   Model is built in scikit-learn and persisted in Redis   Simpliﬁed version of this project by Alex Kagoshima:  https://github.com/alexkago/ds-cfpylearning

Slide 45

Slide 45 text

45 © Copyright 2015 Pivotal. All rights reserved. Resources   Docs: http://docs.cloudfoundry.org   CF Summit videos: http://cfsummit.com   Join the CF community: http://cloudfoundry.org   CF meetups: http://cloud-foundry.meetup.com   Don’t forget to stop your apps with cf stop myapp.