Microservices
6 likes1,193 views
This document discusses microservices architecture. It begins by outlining challenges with monolithic systems, such as everything failing if one part fails and inability to scale the database. The goals of microservices are then presented as scaling individual parts, reducing hardware costs, eliminating expensive support contracts, and allowing quick evaluation of new ideas. Several inspirations for microservices are cited from the 1960s onward. Key principles discussed include having independent, loosely coupled services that communicate through clean interfaces. The document also covers event storming, autonomous components, and other microservices patterns and best practices.
![GET /profile
{!!name: bob ross,!
! text: …!
! portrait: {!
! ! url: /m/f.jpg,!
! ! isLiked: true,!
! }!
! isOnline: true!
! publicPhotos: …
! privatePhotos: [{!
! ! url: /m/f.jpg!
! ! hasAccess: true!
! }]!
}](https://image.slidesharecdn.com/microservices-141003023554-phpapp01/85/Microservices-26-320.jpg)
Microservices
- 1. μServices
- 4. 28 000 000 page views per week *Google Analytics v.17 2014 18 MINUTER time spend per visit* 55% 45%
- 7. A monolithic design is characterised by such a tight coupling among modules that they have no independent existence.
- 8. challenges • If one things fails, everything fails • Can’t scale the database any further • Expensive hardware • Maintenance is scary • Platinum support contract costs
- 9. Goals • Scaling individual parts of the system • Lower hardware costs • No platinum support • Environment to quickly evaluate ideas
- 11. Top inspirations • Larry Constantine - Structured Design (1960) • Douglas McIlroy - Programs (1965) • Ken Thompson - Unix (1973) • Antony Williams - COM (1987) • Udi Dahan - Autonomous Components (2004) • Greg young - CQRS (2008) • Fred George - Micro Services (2013)
- 12. “Write simple parts connected by clean interfaces.” ! “Design programs to be connected to other programs. “ –Ken Thompson, Unix Philosophy
- 13. “Make each program do one thing well. To do a new job, build afresh rather than complicate old programs by adding new features.” ! “Expect the output of every program to become the input to another, as yet unknown, program.” –Mc Mcllroy, Unix Time-Sharing System Forward”. 1978
- 14. “A model can either be optimised for reading or for writing” –Greg Young
- 15. “Depend on others and fail together” –Rob Pike
- 16. simple: easily understood or done
- 18. coupling is the degree to which each module relies on each one of the other modules.
- 19. “μServices should be made as small as possible, but not smaller”
- 20. Event Storming is a workshop format for quickly exploring complex business domains.
- 22. where does online status come from?
- 24. “No μService should ever depend on another being available”
- 25. “every view should be served by a single to one μService”
- 26. GET /profile {!!name: bob ross,! ! text: …! ! portrait: {! ! ! url: /m/f.jpg,! ! ! isLiked: true,! ! }! ! isOnline: true! ! publicPhotos: … ! privatePhotos: [{! ! ! url: /m/f.jpg! ! ! hasAccess: true! ! }]! }
- 28. where does t he member live?
- 30. “No μService represents an business entity, but rather serves an expect of one”
- 31. μServices architecture subscribe to events service publish events GET /profile accept http requests event bus storage reads and write to own storage
- 32. “An μService is the smallest consistency boundary”
- 33. “μServices depends more on how the service fit into the environment than on how they are designed internally.”
- 35. “Polyglot developers still hate context switching”
- 36. “Clarity is better than cleverness”
- 40. “An μService is the smallest part of the system that can scale independently”
- 53. Shipping
- 55. “An μService doesn’t provide any value on its own”
- 56. “a μService is to μ to treat as a single product”
- 59. 7000 5250 3500 1750 0 v13.like.count v14.like.count 13:00 13:15 13:30 13:45
- 60. ETCD • Simple, curl-able API (HTTP + JSON) • Benchmarked 1000s of writes/ s per instance • Properly distributed using Raft protocol • Atomic test and set • Easily listen for changes to a prefix via HTTP long-polling
- 61. Summary
- 62. You've been a great audience! already knew that when creating this slide