Serverless - When to FaaS?
13 likes3,181 views
This document discusses serverless computing and function as a service (FaaS). It provides an overview of what serverless and FaaS are, how they work, common use cases, limitations, and the future impact. Key points include that FaaS provides fully managed compute with no servers to provision and pays only for actual usage, common triggers are HTTP requests and events, and use cases span web backends, mobile apps, file processing, bots, and more. The document also outlines limitations like statelessness and cold starts, as well as emerging practices around business models, architectures, operations, and tooling.
Serverless - When to FaaS?
- 1. The Rise of Serverless - When to FaaS? Photo by Shane Perry Icon by Serverless
- 2. ◉ Cloud Architecture & Development Consulting ◉ Serverless Enthusiast ◉ Ex- Find me at @benikbauer I am Benny Bauer Shalom!
- 3. EvolutionGenesis Custom Built Product (+ rental) Commodity (+ utility) Cloud evolution in 60 sec Data center Hosting Virtualization IaaS Icons made by freepik from www.flaticon.com & unlimicon from Noun Project On Premise Compute Platform PaaS Serverless
- 4. What is Serverless FaaS BaaS B2D SaaS
- 5. What is FaaS - Function as a Service Fully-managed compute Provisioning, patching, scaling, monitoring, logging are provided out-of-the- box Deploy your code Just package and upload the code Pay for actual usage Getting charged only upon code execution, per 100ms 100% UTILIZATION LESS OPS
- 6. How it works Deploy your code Define triggers Code execution (ephemeral instance) EVENT-DRIVEN AUTO SCALING + AVAILABILITY
- 7. Event-driven on steroids ◉ HTTP requests ◉ Storage (e.g. file upload) ◉ DB (e.g. row insert) ◉ Events stream ◉ Queue/Messaging ◉ Scheduled job ◉ Monitoring ◉ Infrastructure events ◉ Code build events ◉ Many more...
- 8. Distributed architecture on steroids Monolith Microservices Serverless Source: http://ryanjbaxter.com/2015/07/15/using-microservices-to-build-cloud-native-applications-part-1/
- 9. Bad ◉ Monitoring ◉ Debugging ◉ Deployment ◉ Security Distributed architecture on steroids Good ◉ Scaling ◉ Fault tolerance ◉ Ownership ◉ Cost (better utilization)
- 10. Let’s combine these building blocks to something useful Use Cases
- 12. Use case: Web backend
- 13. Use case: Web backend Source: https://www.linkedin.com/pulse/serverless-architecture-cloud-guru-ryan-kroonenburg
- 15. Use case: Mobile backend
- 16. Use case: Mobile backend Async request
- 18. Use case: File processing
- 19. Use case: File processing
- 21. Use case: Bot by
- 22. Use case: Bot
- 24. Use case: Distributed compute import pywren def my_function(x): return x + 7 wrenexec = pywren.default_executor() futures = wrenexec.map(my_function, range(10)) pywren.get_all_results(futures) pywren
- 26. Use case: Stream processing ◉ Log processing ◉ Analytics ◉ etc.
- 28. Use case: Scheduled tasks ◉ Health checks ◉ Monitoring ◉ Periodical digest ◉ etc.
- 30. Use case: Operations Policy enforcement on AWS resources. Examples: ● Unmanaged instances handling ● Low utilized instances resizing Triggered by: ● CloudWatch events ● Scheduled
- 32. Use case: CI Source: https://github.com/lambci/lambci
- 33. Use case: CI Source: https://github.com/lambci/lambci LambCI (Travis alternative)
- 35. Use case: IoT backend Source: https://www.hackster.io/jose-troche/alexa-drone-pilot-7f2deb
- 36. Use case: IoT backend Source: https://medium.com/serverless-stories/building-a-serverless-garden-monitoring-system-with-lambda-c72c54113df2 Garden Monitoring System
- 38. Use case: CDN Source: https://www.slideshare.net/AmazonWebServices/new-launch-bringing-aws-lambda-to-the-edge ◉ Purpose: ○ Better UX ○ Reduce load from backend ◉ Use cases: ○ Request validation ○ A/B testing ○ Content customizations according to user-agent
- 39. Use case: CDN Source: https://www.slideshare.net/AmazonWebServices/new-launch-bringing-aws-lambda-to-the-edge
- 42. Use case: @Device - example Source: https://www.slideshare.net/AmazonWebServices/aws-welcome-to-reinvent-recap-20161214
- 44. Use case: More
- 45. But isn’t it limitationless?? Mind The Limitations
- 46. Limitations: Stateless ◉ Instances are ephemeral (memory, storage) but not necessarily disposable ◉ No sticky sessions ◉ Workarounds: ○ Websockets through IoT (MQTT) ○ Manage state yourself: ■ Pass on request ■ Store in cache/db
- 47. Execution time: 5 min Memory: 1.5GB Disk: 500MB Package (zip): 50MB Limitations: Provider limits
- 48. ◉ Ephemeral => cold start on first or infrequent calls ◉ Inconsistent Limitations: Latency Source: http://www.yusp.com/blog/cold-start-problem-recommender-systems/
- 49. Limitations: Latency Source: https://blog.newrelic.com/2017/01/11/aws-lambda-cold-start-optimization/ Frequency: 1 minute 10 minutes 30 minutes 60 minutes
- 51. Limitations: Costs AWS Lambda EC2 Duration Per 100 ms Per 1 hour Reduced cost for RI Cost of ownership Low High Additional costs Requests Redundancy Storage Cost effective For low or changing load For high utilization
- 53. Limitations: Vendor lock-in ◉ The real lock-in is the integrated services ◉ A tradeoff: control vs. convenience Source: https://www.slideshare.net/AmazonWebServices/reinvent-recap-session-1-whats-new-with-aws-lambda
- 56. Know the challenges ◉ Development ○ Local development ○ Integration tests ○ Debugging ◉ Granularity ○ Deployment ○ Monitoring ○ Logging ○ Security ○ Service discovery
- 58. Tools
- 59. The FaaS effect New emerging practices Photo by TimOve (https://flic.kr/p/4aTMMx)
- 60. Business model ◉ Worth Based Development (“FinDev”) https://hackernoon.com/why-the-fuss-about-serverless-4370b1596da0#.yct5knd7h by Simon Wardley
- 61. Business model ◉ API Marketplace
- 62. Architecture ◉ More distributed and pluggable ◉ More event-driven ◉ Frameworks are changing
- 63. Operations DevOps is changing ◉ Less about provisioning, capacity planning ◉ More about distributed systems: ○ Monitoring & visibility ○ Debuggability ○ Graceful degradation & Resiliency ○ CI/CD ○ Costs optimisations
- 64. Summary What was it all about?
- 65. Predicting the future ◉ Lower costs ◉ Less limitations ◉ More use cases ◉ More tooling
- 66. Takeaways ◉ Deliver faster with lower costs for many use cases ◉ Consider limitations ◉ Get ready for new emerging practices
- 67. Any questions? Thanks! Slides template by SlidesCarnival You can find me at ◉ @benikbauer ◉ bennybauer@gmail.com This work is licensed under a CC Attribution 4.0 International License.