Can Time-Travel Keep You From Blowing Up The Enterprise?

Transcript

1. CAN TIME-TRAVEL KEEP YOU FROM BLOWING UP THE ENTERPRISE? @DAVETRON5000

2. SCALING RAILS

3. SCALING RAILS ▸ Do stuff Faster? ▸ Do more stuff

   at once?

4. SCALING RAILS ▸ Do stuff Faster? ▸ Do more stuff

   at once? NO, IT'S ACTUALLY… ▸ Growing Team ▸ Growing Business ▸ Changing Business

5. MONOLITH? OR MICROSERVICES?

6. A FRAMING DEVICE CAUSE & EFFECT

7. DON'T BLOW UP THE ENTERPRISE

8. DON'T BLOW UP THE ENTERPRISE

9. DON'T BLOW UP THE ENTERPRISE

10. DON'T BLOW UP THE ENTERPRISE

11. DON'T BLOW UP THE ENTERPRISE

12. FIRST, SOME ASSUMPTIONS

13. FIRST, SOME ASSUMPTIONS ▸ Over time, more people will work

    on the code

14. FIRST, SOME ASSUMPTIONS ▸ Over time, more people will work

    on the code ▸ Over time, different people will work on the code

15. FIRST, SOME ASSUMPTIONS ▸ Over time, more people will work

    on the code ▸ Over time, different people will work on the code ▸ There will be increasing demands on the code

16. FAILURE: BUSINESS/PRODUCT NOT MEETING USER NEEDS ▸ Simple changes difficult

    ▸ Difficult changes impossible ▸ Low morale ▸ Brain drain

17. COMMON, BUT NOT UNIVERSAL ▸ Start-up ▸ New project on

    a larger company ▸ New longer-term consulting gig

18. COMMON, BUT NOT UNIVERSAL ▸ Start-up ▸ New project on

    a larger company ▸ New longer-term consulting gig GROWTH & CHANGE

19. NOTHIN' BEATS MONOLITH! GOOD 'OLE MONOLITH

20. SHIP EARLY, SHIP OFTEN INTEGRATED SYSTEM

21. MORE DEVELOPERS!

22. MORE DEVELOPERS!

23. MORE DEVELOPERS!

24. MORE HIDDEN COMPLEXITY HOMEPAGE ADMIN PURCHASING SEARCH REFERRALS

25. MORE HIDDEN COMPLEXITY ADMIN.CSS HOMEPAGE ADMIN PURCHASING SEARCH REFERRALS

26. MORE HIDDEN COMPLEXITY ADMIN.CSS HOMEPAGE ADMIN SEARCH REFERRALS

27. MORE HIDDEN COMPLEXITY ADMIN.CSS HOMEPAGE ADMIN SEARCH CACHE REFERRALS

28. MORE HIDDEN COMPLEXITY ADMIN.CSS HOMEPAGE ADMIN CACHE REFERRALS

29. ENTANGLEMENTS

30. ENTANGLEMENTS user.signed_up.chronologically.approved. active_orders.in_zip(:east_cost). unshipped.first.try(:first_name)

31. ENTANGLEMENTS

32. ENTANGLEMENTS after_create do |user| end

33. ENTANGLEMENTS after_create do |user| end UserMailer.send_welcome(user.id)

34. ENTANGLEMENTS after_create do |user| end UserMailer.send_welcome(user.id) PaymentProcessor.add_customer(user.id)

35. ENTANGLEMENTS after_create do |user| end UserMailer.send_welcome(user.id) PaymentProcessor.add_customer(user.id) MarketingTrackingSite.event!(user.id)

36. ENTANGLEMENTS after_create do |user| end UserMailer.send_welcome(user.id) PaymentProcessor.add_customer(user.id) MarketingTrackingSite.event!(user.id) OtherTrackingSite.event!(user.id)

37. OH, AND… ▸ Test suite is glacial ▸ Don't even

    think about looking in app/assets/javascripts ▸ /admin starts acting up -> www is dead, too ▸ Etc.

38. REMEMBER WHAT FAILURE LOOKS LIKE? ▸ Simple changes difficult ▸

    Difficult changes impossible ▸ Low morale ▸ Brain drain

39. REMEMBER WHAT FAILURE LOOKS LIKE? ▸ Simple changes difficult ▸

    Difficult changes impossible ▸ Low morale ▸ Brain drain

40. MICROSERVICES FROM DAY 1 NEVER AGAIN

41. BEFORE EVEN STARTING…

42. BEFORE EVEN STARTING… ▸ Deploy lots of apps

43. BEFORE EVEN STARTING… ▸ Deploy lots of apps ▸ Authentication

44. BEFORE EVEN STARTING… ▸ Deploy lots of apps ▸ Authentication

    ▸ Versioning

45. BEFORE EVEN STARTING… ▸ Deploy lots of apps ▸ Authentication

    ▸ Versioning ▸ JSON format

46. BEFORE EVEN STARTING… ▸ Deploy lots of apps ▸ Authentication

    ▸ Versioning ▸ JSON format ▸ Which HTTP library?

47. BEFORE EVEN STARTING… ▸ Deploy lots of apps ▸ Authentication

    ▸ Versioning ▸ JSON format ▸ Which HTTP library? ▸ Argue about REST and status codes

48. OH, AND WHAT SERVICES DO WE ACTUALLY NEED?

49. THREE MONTHS LATER

50. THREE MONTHS LATER ▸ React front-end

51. THREE MONTHS LATER ▸ React front-end ▸ Hits the Backend-for-frontends

    (BFF)

52. THREE MONTHS LATER ▸ React front-end ▸ Hits the Backend-for-frontends

    (BFF) ▸ BFF talks to the underlying auth service

53. THREE MONTHS LATER ▸ React front-end ▸ Hits the Backend-for-frontends

    (BFF) ▸ BFF talks to the underlying auth service ▸ Then hits the role service

54. THREE MONTHS LATER ▸ React front-end ▸ Hits the Backend-for-frontends

    (BFF) ▸ BFF talks to the underlying auth service ▸ Then hits the role service ▸ Then decorates with the consumer data

55. THREE MONTHS LATER ▸ React front-end ▸ Hits the Backend-for-frontends

    (BFF) ▸ BFF talks to the underlying auth service ▸ Then hits the role service ▸ Then decorates with the consumer data ▸ To render “Hello, #{first_name}!”

56. THREE MONTHS LATER ▸ React front-end ▸ Hits the Backend-for-frontends

    (BFF) ▸ BFF talks to the underlying auth service ▸ Then hits the role service ▸ Then decorates with the consumer data ▸ To render “Hello, #{first_name}!” ▸ Managed by Docker, Ansible, AWS, Terraform, Vault, Bash, and, of course, some Go

57. WHAT HAPPENED?

58. WHAT HAPPENED? ▸ Money has been spent

59. WHAT HAPPENED? ▸ Money has been spent ▸ Blog posts

    all over Hacker News

60. WHAT HAPPENED? ▸ Money has been spent ▸ Blog posts

    all over Hacker News ▸ Meetups have been attended

61. WHAT HAPPENED? ▸ Money has been spent ▸ Blog posts

    all over Hacker News ▸ Meetups have been attended ▸ No actual business value

62. WHAT HAPPENED? ▸ Money has been spent ▸ Blog posts

    all over Hacker News ▸ Meetups have been attended ▸ No actual business value ▸ Nothing to show

63. WHAT HAPPENED? ▸ Money has been spent ▸ Blog posts

    all over Hacker News ▸ Meetups have been attended ▸ No actual business value ▸ Nothing to show ▸ Is there any money left?

64. REMEMBER WHAT FAILURE LOOKS LIKE? ▸ Simple changes difficult ▸

    Difficult changes impossible ▸ Low morale ▸ Brain drain

65. REMEMBER WHAT FAILURE LOOKS LIKE? ▸ Simple changes difficult ▸

    Difficult changes impossible ▸ Low morale ▸ Brain drain

66. LET'S DO THIS OK MONOLITH

67. AS LONG AS WE _NOTICE_ THE MESS, WE CAN CLEAN

    IT UP —Pat Agîlecønßultånt

68. MORE DEVELOPERS!

69. MORE DEVELOPERS!

70. MORE HIDDEN COMPLEXITY HOMEPAGE ADMIN PURCHASING SEARCH REFERRALS

71. MORE HIDDEN COMPLEXITY ADMIN.CSS HOMEPAGE ADMIN CACHE REFERRALS NOW, IT’S

    TOO MESSY!
72. None

73. MORE HIDDEN COMPLEXITY HOMEPAGE ADMIN PURCHASING SEARCH REFERRALS

74. MORE HIDDEN COMPLEXITY HOMEPAGE ADMIN PURCHASING REFERRALS PROMO CODES

75. CLEANING THE AIRPORT BATHROOM ▸ Requires top engineers ▸ Grueling

    ▸ Small changes difficult ▸ Bigger changes become delayed

76. REMEMBER WHAT FAILURE LOOKS LIKE? ▸ Low morale ▸ Brain

    drain

77. REMEMBER WHAT FAILURE LOOKS LIKE? ▸ Low morale ▸ Brain

    drain

78. THERE _IS_ ANOTHER WAY NO BLUR,

79. WHILE IT'S IMPOSSIBLE TO PREDICT THE FUTURE, WE DON'T NEED

    TO ACT BLIND TO IT. Chris Reasonable

80. DON'T BE BLIND TO THE FUTURE ▸ More developers, more

    features, more code, more changes ▸ You'll want things decoupled; but not exactly what. ▸ Don't make a mess in the first place

81. ENABLE DECOUPLING ▸ One rails app per business function ▸

    No business logic in Active Record models ▸ Use the Rule of Three to guide decisions around abstraction, tooling, de- coupling, and general pattern identification.

82. RULE OF THREE ▸ 1st and 2nd times aren't a

    pattern. ▸ 3rd time is - use that to create needed tooling or process changes ▸ This is engineering culture ▸ Mature teams can graduate to the Rule of Two :) Effort 1ST 2ND 3RD 4TH 5TH 6TH

83. EXAMPLES ▸ Third Rails app - create it with a

    re-usable generator ▸ Third Ruby Gem - create it with a re-usable gem scaffold ▸ Third mess you make with jQuery - consider a front-end framework ▸ Third time customers report an outage - improve monitoring

84. WWW SHARED DATABASE

85. WWW CUSTOMER SERVICE SHARED DATABASE

86. WWW CUSTOMER SERVICE MARKETING SHARED DATABASE

87. WWW CUSTOMER SERVICE MARKETING SHARED DATABASE SHARED MODELS SHARED MODELS

    SHARED MODELS

88. WWW CUSTOMER SERVICE MARKETING SHARED DATABASE SHARED MODELS SHARED MODELS

    SHARED MODELS PROMO CODE LOGIC PROMO CODE LOGIC

89. WWW CUSTOMER SERVICE MARKETING SHARED DATABASE SHARED MODELS SHARED MODELS

    SHARED MODELS PROMO CODE LOGIC PROMO CODE LOGIC CACHE

90. WWW CUSTOMER SERVICE MARKETING SHARED DATABASE SHARED MODELS SHARED MODELS

    SHARED MODELS PROMO CODE LOGIC PROMO CODE LOGIC CACHE

91. WWW CUSTOMER SERVICE MARKETING SHARED DATABASE SHARED MODELS SHARED MODELS

    SHARED MODELS ANALYTICS PROMO CODE LOGIC PROMO CODE LOGIC CACHE

92. WWW CUSTOMER SERVICE MARKETING FINANCE SHARED DATABASE SHARED MODELS SHARED

    MODELS SHARED MODELS ANALYTICS PROMO CODE LOGIC PROMO CODE LOGIC CACHE

93. WWW CUSTOMER SERVICE MARKETING FINANCE SHARED DATABASE SHARED MODELS SHARED

    MODELS SHARED MODELS ANALYTICS PROMO CODES HTTP CACHE

94. WWW CUSTOMER SERVICE MARKETING FINANCE SHARED DATABASE SHARED MODELS SHARED

    MODELS SHARED MODELS ANALYTICS PROMO CODES HTTP CACHE

95. WWW CUSTOMER SERVICE MARKETING FINANCE SHARED DATABASE SHARED MODELS SHARED

    MODELS SHARED MODELS ANALYTICS PROMO CODES HTTP PRIVATE DATABASE CACHE

96. WWW CUSTOMER SERVICE MARKETING FINANCE SHARED DATABASE SHARED MODELS SHARED

    MODELS SHARED MODELS ANALYTICS PROMO CODES HTTP PRIVATE DATABASE PRIVATE DATABASE CACHE

97. YOUR ARCHITECTURE ISN'T EVER "DONE"

98. YOUR ARCHITECTURE PLANS SHOULD BE HOW IT CHANGES NOT WHAT

    IT SHOULD BE

99. ENTERPRISE BEGINS

100. WHAT DID WE LEARN EXACTLY? ▸ Don't be blind to

     the future (this is different than predicting it!) ▸ Write code to enable change ▸ Create a culture of change, based on the Rule of Three ▸ Your architecture must evolve and is never "done" ▸ Don't make a mess

101. 25% off CODE RailsConf16 FULL-STACK-RAILS.COM SWENG.ME NEED RANTS ABOUT BAD

     UX AND JAVASCRIPT? @DAVETRON5000 WANT TO WORK THIS WAY? TECH.STITCHFIX.COM/CAREERS/ FINAL SLIDE OF SELF-PROMOTION