Serverless - The Future of the Cloud?!
9 likes1,847 views
The document explores the evolution of computing towards a serverless model, emphasizing its advantages such as automatic scaling and cost efficiency. It outlines the distinctions between serverless functions (FaaS) and traditional platform as a service (PaaS), highlighting the benefits of reduced server management and pay-per-use pricing. Additionally, it discusses various cloud implementations, specifically AWS Lambda, and addresses potential drawbacks such as vendor lock-in and security concerns.
Serverless - The Future of the Cloud?!
- 1. by Bert Ertman Those who stand for nothing, fall for anything - Alexander Hamilton Serverless The Future of the Cloud?! @BertErtman JAVAZONE 2017
- 2. • Fellow at Luminis • Background in all things Java since 1995 • Java Champion, JavaOne Rockstar Speaker, and a Duke’s Choice Award Winner • Involved in architecting and implementing dozens of large scale systems over the past 20 years or so • Book author for O’Reilly, speaker at many conferences About me
- 3. The Evolution of Compute Physical Virtualization Cloud Compute Containers Serverless
- 4. Why?!
- 5. The case for Serverless • So far, the cloud is just someone else’s computer • Servers should be treated as cattle, not pets • PAYGO? or PAYGO&aAYDG?
- 6. “No server is easier to manage than no server” - Werner Vogels CTO, Amazon
- 7. Serverless - what’s in a name? • Mostly describes what its not… • Also known as Functions as a Service (FaaS) • or Function PaaS (fPaaS) as defined by Gartner • Some people refer to it as Back-end as a Service (BaaS)
- 8. Wait! Did someone just say “Back-end as a Service”?
- 9. Back-end as a Service??
- 10. Serverless Manifesto • Functions are the unit of deployment and scaling • No Machines, VMs, or Containers are visible in the programming model • Permanent storage lives elsewhere • Scales per request. Users cannot over- or under-provision capacity • Never pay for idle (no cold servers/containers or their costs) • Implicitly fault-tolerant because functions can run anywhere • BYOC - Bring Your Own Code • Metrics and Logging are a universal right
- 11. Main Benefits • No servers to administer • Pay for code execution only • Automatic Scaling
- 12. In Other Words… • We don’t (have to) care about Application Servers • We don’t (have to) care about Docker
- 13. And Best of All • Very short Time-to-Market • From development to production in matter of seconds • Very affordable (PAYG only, no up-front costs)
- 14. So far so good, right?
- 16. PaaS vs FaaS • With PaaS, you still manage “applications” • With PaaS, you take care of scaling • With PaaS, you manage runtime environment configuration • With PaaS, you’ll pay for all of the above too • With FaaS, you have neither
- 17. Serverless implementations • Several Cloud vendors have implementations: • AWS Lambda • Google CloudFunctions • Azure Functions • IBM BlueMix OpenWhisk 2015 2016 2017
- 19. AWS Lambda • Event-driven, serverless computing platform provided by Amazon • Runs code in response to events and automatically manages the compute resources required by that code • First introduced in Nov 2014 • Part of the Amazon Web Services offerings
- 21. Runtime Support • Python 2.7 • NodeJS 4.3 • Java 8 • C#
- 22. How it works Upload Trigger Run Pay
- 23. lambda Erhm..
- 24. AWS Lambda Pricing • Suits the PAYG model better than running idle compute instances • Unlike EC2 which is priced by the hour, Lambdas are priced per 100 ms (and memory footprint per second) • AWS Free Tier: 1,000,000 free requests per month* *) Restrictions apply, see AWS Lambda Pricing Details online
- 26. Many Event Sources API Gateway Amazon S3 Amazon DynamoDB Amazon Aurora Amazon Simple Notification Service Amazon Simple Email Service Amazon Cognito Amazon CloudWatch Amazon Kinesis Streams AWS CodeCommit AWS CloudFormation AWS Config Amazon Lex
- 27. Many Event Sources API Gateway Amazon S3 Amazon DynamoDB Amazon Aurora Amazon Simple Notification Service Amazon Simple Email Service Amazon Cognito Amazon CloudWatch Amazon Kinesis Streams AWS CodeCommit AWS CloudFormation AWS Config Amazon Lex HTTP(s) requests CRUD events on data sources messaging events log/stream processing commit hooks / AWS services setup configuration mgmt voice & text
- 29. Rethinking Traditional Architectural Concepts Client (browser) Web Application (server-side) Database
- 30. Client (browser) API Gateway Purchase Database Authentication Service Product Catalog (Database) Search function Purchase function CDN 1 2 3 6 74 5 Rethinking Traditional Architectural Concepts
- 31. Just because you can might not always be the right reason
- 32. Better Example: BaaS • Backend-as-a-Service for a (mobile) app • Responds to incoming HTTP GET/POST requests • Stateless
- 33. Typical Scenario Mobile App Web application API Gateway Lambda function DynamoDB
- 34. demo
- 35. Typical Scenario Mobile App Web application API Gateway Lambda function DynamoDB Microservice?
- 36. Are Lambda functions Microservices? • Similarities: • Do one thing, and one thing well • Event-based interaction == choreography model • Differences: • One Lambda is equal to one action == NanoService • Microservice == bounded context of actions with autonomous storage
- 37. Example: Event-based processing • Respond to incoming data, such as an S3 Bucket insert • Useful for data/stream processing, MapReduce, or batch processing
- 38. Typical Scenario Application S3 Bucket S3 BucketLambda function
- 39. demo
- 40. What Expedia is doing with Lambda source: AWS re:invent 2016: Serverless Computing Patterns at Expedia (SVR306) beyond Hello World example
- 42. Yeah, nice and all, but all of this $**t is running in the cloud…
- 43. OpenWhisk • Runs in IBM’s BlueMix Cloud, but can also run on-premise • OpenWhisk is Apache licensed and Open Source • Currently supports: JavaScript, Java, Python, and Swift(!) • Possible to run functions in provided Docker images
- 44. But there are some more drawbacks…
- 45. Drawbacks • Vendor control and lock-in • Multi-tenancy • Security concerns (increasing the attack surface) • Loss of server optimizations • Execution time is limited • Start-up latency • Testing • Discovery
- 46. summary
- 47. Serverless • ..is rapidly being embraced by major cloud players • ..is promoting functions as first class citizens • ..is event-based, stateless, and transient • ..is infinitely scalable (in theory) • ..is different from traditional deployment models • ..is giving the cloud a run for its money • ..is lots of bang for the buck • ..is still very much proprietary, so lock-in is your choice!
- 48. Serverless, the future of the Cloud! Thank you! @BertErtman