Scale your node application...forget the infrastructure

Delivered at CodeMash 2017 - You don’t have to look far to find yet another article on scaling your NodeJS app to handle large numbers of users. But the techniques covered usually involve becoming an expert in deploying hardware or a guaranteed minimum outlay of cash to handle your expected load. But what if there were a way to scale almost infinitely without having to worry about the infrastructure to run any of it. Using platforms like AWS Lambda, Azure Functions, and IBM OpenWhisk, you can focus on providing scalable functionality in your NodeJS application without having to think about any of infrastructure details. In this session, Matt Williams will show you how to get started building a complex Node application on AWS Lambda from scratch. Starting with the standard CLI, we move to other frameworks like Node Lambda and Serverless to add more functionality to serve your users. We will consider some key architectural decisions that affect how the application is designed. And all the way along we look at ways to monitor the application to help find the bottlenecks. By the end of the session, you will be eager to start working on your next application on AWS Lambda.

Transcript

1. SCALE YOUR NODE APPLICATION... ...FORGET THE INFRASTRUCTURE Matt Williams Evangelist @ Datadog mattw@datadoghq.com @technovangelist
2. IMAGINE THIS...
3. YOU HAVE BUILT AN APP (using node of course)
4. IT'S AMAZING
5. PEOPLE NOTICE
6. IT GETS POPULAR
7. AND NOT HOW IT SCALES BUT YOU FOCUSED ON THE APP
8. HOW DO YOU FIX THIS? Do Less Tweak & Tune Throw Hardware at it Refactor Rearchitect
9. NEXT TIME AROUND...
10. YOU BUILD AN APP (using node of course)
11. You have spent as much time on the app as on the infrastructure THIS TIME IS DIFFERENT
12. YOU STARTED WITH AN EXTENSIBLE PLATFORM
13. YOU HAVE CM READY to scale up and out
14. AND THE APPLICATION IS AMAZING...
15. FOR ALL 3 USERS WHO SHOWED UP
16. t2.large t2.xl t2.2xl m4.large m4.2xl c4.large r4.large 1 67 135 270.72 77.76 310.32 72 95.76 2 135 271 541 155 621 144 192 3 203 406 812 233 931 216 287
17. SO LETS LOOK AT AN ALTERNATIVE
18. SERVERLESS
19. AWS Lambda Azure Functions Google Cloud Functions IBM OpenWhisk SERVERLESS PLATFORMS
22. SERVERLESS ...AS FAR AS YOU ARE CONCERNED
23. WHAT IS LAMBDA? Your executable code No restrictions Priced based on: execution time quantity memory
24. LAMBDA FREE PLAN/MONTH
            
    
25. RUN THE NUMBERS Run a Lambda function every 5 minutes s/m 1s 5s 10s 128 3,200,000 370 74 37 256 1,600,000 185 37 18 512 800,000 93 18 9 1024 400,000 46 9 5 1536 266,666 31 6 3 0.0025 cents / second thereafter for 1.5G
26. GETTING STARTED WITH LAMBDA
27. HOW TO CREATE A LAMBDA FUNCTION 1. Write the code 2. Setup on AWS 3. Debug
28. LAMBDA DEMO
29. FRAMEWORKS
30. NATIVE GOTCHAS Lambda runs on an Amazon Linux container... ...and your dev box is probably not Amazon Linux... ...so native NPM modules are problematic
31. NATIVE NPM 1. Launch Amazon Linux ec2 2. Use a node version manager 3. Install node 4.3 4. Install the native modules you need 5. Package the modules and dl to your box 6. Deploy
32. BUILDING THE DEMO WITH NODE-LAMBDA Four files: 1. package.json 2. .env 3. event.json 4. index.js
33. LAMBDA BEST PRACTICES Minimize startup Core logic into module Monitor Delete old functions
34. MONITORING 101 the tl;dr version
35. COLLECTING DATA IS CHEAP
36. COLLECTING DATA IS CHEAP NOT HAVING IT WHEN YOU NEED IT IS EXPENSIVE
38. THE FRAMEWORK
39. THE FRAMEWORK
40. THE FRAMEWORK
41. THE FRAMEWORK
45. MONITORING LAMBDA
46. WHAT ARE THE LAMBDA METRICS aws.lambda.duration, min, max, sum aws.lambda.errors aws.lambda.invocations aws.lambda.throttles
47. 3 OPTIONS FOR MONITORING LAMBDA 1. Add a custom metric to CloudWatch 2. Add metric directly to monitoring application 3. Add a line to the CloudWatch logs
48. HOW WE DO IT Add a line to the CloudWatch logs:
             !"
49. MONITORING EXAMPLE
50. MONITORING EXAMPLE
51. MONITORING EXAMPLE
53. MATT WILLIAMS EVANGELIST @ DATADOG MATTW@DATADOGHQ.COM | @TECHNOVANGELIST