Distributed Elixir

Presentation about some of the tools for distributed programming in Elixir

Transcript

1. It’s scary out there
   

   View Slide

2. Organisational Matters
   

   View Slide

3. View Slide

4. We’re 1 year old!
   

   View Slide

5. Summer break (probably)
   

   View Slide

6. We’re looking for speakers!
   

   View Slide

7. It’s scary out there
   Distributed Systems in Elixir
   Poznań Elixir Meetup #8
   

   View Slide

8. View Slide

9. Pid 1 Pid 2
   

   View Slide

10. Pid 1 Pid 2
    Node A Node B
    

    View Slide

11. The basics
    

    View Slide

12. iex --name n1@127.0.0.1 --cookie cookie -S mix
    

    View Slide

13. Node.connect(:’n1@127.0.0.1')
    

    View Slide

14. (DEMO)
    

    View Slide

15. #PID<0.94.0>
    

    View Slide

16. #PID<0.94.0>
    node identifier
    (relative to current node)
    

    View Slide

17. #PID<0.94.0>
    node identifier
    (relative to current node)
    0 =a local process
    

    View Slide

18. #PID<0.94.0>
    Process id
    node identifier
    (relative to current node)
    

    View Slide

19. How does it work?
    

    View Slide

20. Pid 1
    Node A
    Pid 2
    Node B
    

    View Slide

21. Pid 1
    Node A
    Pid 2
    Node B
    TCP Connection
    

    View Slide

22. send(pid2, msg)
    Pid 1
    Node A
    Pid 2
    Node B
    TCP Connection
    

    View Slide

23. send(pid2, msg)
    Pid 1
    Node A
    Pid 2
    Node B
    destination_node = node(pid)
    TCP Connection
    

    View Slide

24. send(pid2, msg)
    Pid 1
    Node A
    Pid 2
    Node B
    destination_node = node(pid)
    :erlang.term_to_binary(msg)
    TCP Connection
    

    View Slide

25. send(pid2, msg)
    Pid 1
    Node A
    Pid 2
    Node B
    destination_node = node(pid)
    :erlang.term_to_binary(msg)
    TCP Connection
    

    View Slide

26. send(pid2, msg)
    Pid 1
    Node A
    Pid 2
    Node B
    destination_node = node(pid)
    :erlang.term_to_binary(msg)
    TCP Connection
    :erlang.binary_to_term(encode)
    

    View Slide

27. send(pid2, msg)
    Pid 1
    Node A
    receive msg
    Pid 2
    Node B
    destination_node = node(pid)
    :erlang.term_to_binary(msg)
    TCP Connection
    :erlang.binary_to_term(encode)
    

    View Slide

28. Distributed Systems?
    

    View Slide

29. Distributed Systems?
    Solved!
    

    View Slide

30. Well, not exactly…
    

    View Slide

31. Difficulties
    

    View Slide

32. Node A
    Node B
    

    View Slide

33. Node A
    Node B
    Node C
    

    View Slide

34. Node A
    Node B
    Node C
    Node D
    

    View Slide

35. View Slide

36. A lot of messages
    

    View Slide

37. us-east-1
    us-west-2
    

    View Slide

38. 8 fallacies of
    distributed computing
    

    View Slide

39. fallacies of distributed computing
    1. The network is reliable
    2. Latency is zero
    3. Bandwidth is infinite
    4. The network is secure
    5. Topology doesn’t change
    6. The is one administrator
    7. Transport cost is zero
    8. The network is homogenous
    

    View Slide

40. CAP THEOREM
    

    View Slide

41. CAP THEOREM
    us-west-2
    us-east-1
    

    View Slide

42. CAP THEOREM
    us-west-2
    us-east-1
    Set X=5
    

    View Slide

43. CAP THEOREM
    us-west-2
    us-east-1
    Set X=5
    Read X
    

    View Slide

44. CAP THEOREM
    us-west-2
    us-east-1
    Set X=5
    Set X = 7
    

    View Slide

45. Consistency or Availability
    (under network partition)
    

    View Slide

46. Consistency or Speed
    In practice
    

    View Slide

47. Guarantees
    

    View Slide

48. Pid 1
    Pid 2
    Pid3
    Guarantees
    m1, m2, m3
    m4, m5, m6
    send(pid2, m1)
    send(pid2, m2)
    send(pid2, m3)
    send(pid2, m4)
    send(pid2, m5)
    send(pid2, m6)
    

    View Slide

49. Pid 1
    Pid 2
    Pid3
    Guarantees
    m1, m2, m3
    m4, m5, m6
    send(pid2, m1)
    send(pid2, m2)
    send(pid2, m3)
    send(pid2, m4)
    send(pid2, m5)
    send(pid2, m6)
    Ordering between two processes
    is preserved
    

    View Slide

50. Pid 1
    Pid 2
    Pid3
    Guarantees
    m4, m5, m6
    send(pid2, m1)
    send(pid2, m2)
    send(pid2, m3)
    send(pid2, m4)
    send(pid2, m5)
    send(pid2, m6)
    m1, m2, m3
    Delivery is not guaranteed
    

    View Slide

51. Pid 1
    Pid 2
    Pid3
    Guarantees
    m1, m2, m3
    m4, m5, m6
    send(pid2, m1)
    send(pid2, m2)
    send(pid2, m3)
    send(pid2, m4)
    send(pid2, m5)
    send(pid2, m6)
    Ordering between different processes
    is not guaranteed
    

    View Slide

52. [m1, m2, m3, m4, m5, m6]
    

    View Slide

53. [m1, m2, m3, m4, m5, m6]
    [m4, m5, m6, m1, m2, m3]
    

    View Slide

54. [m1, m2, m3, m4, m5, m6]
    [m4, m5, m6, m1, m2, m3]
    [m1, m4, m2, m5, m3, m6]
    

    View Slide

55. [m1, m2, m3, m4, m5, m6]
    [m4, m5, m6, m1, m2, m3]
    [m1, m4, m2, m5, m3, m6]
    [m1, m2, m3]
    

    View Slide

56. [m1, m2, m3, m4, m5, m6]
    [m4, m5, m6, m1, m2, m3]
    [m1, m4, m2, m5, m3, m6]
    [m1, m2, m3]
    [m1, m3, m5, m6]
    

    View Slide

57. [m1, m2, m3, m4, m5, m6]
    [m4, m5, m6, m1, m2, m3]
    [m1, m4, m2, m5, m3, m6]
    [m1, m2, m3]
    [m1, m3, m5, m6]
    []
    

    View Slide

58. [m1, m2, m3, m4, m5, m6]
    [m4, m5, m6, m1, m2, m3]
    [m1, m4, m2, m5, m3, m6]
    [m1, m2, m3]
    [m1, m3, m5, m6]
    []
    [m1, m3, m2, m4, m5, m6]
    

    View Slide

59. [m1, m2, m3, m4, m5, m6]
    [m4, m5, m6, m1, m2, m3]
    [m1, m4, m2, m5, m3, m6]
    [m1, m2, m3]
    [m1, m3, m5, m6]
    []
    [m1, m3, m2, m4, m5, m6]
    [M3, M3]
    

    View Slide

60. Phoenix
    Request A User Logged In
    

    View Slide

61. Phoenix
    Request A
    Phoenix
    Request B
    User Logged In
    User Logged OUT
    

    View Slide

62. Phoenix
    Request A
    Phoenix
    Request B
    User Logged In
    User Logged OUT
    This Can arrive first
    

    View Slide

63. Unfortunately, things tend to work fine
    locally
    

    View Slide

64. The Tools
    

    View Slide

65. :global
    

    View Slide

66. Pid 1
    Node A Node B
    Pid 2
    

    View Slide

67. Pid 1
    Node A Node B
    Pid 2
    :global.register_name(“global”, self())
    

    View Slide

68. Pid 1
    Node A Node B
    Pid 2
    :global.register_name(“global”, self())
    Register PId1 as “global”
    

    View Slide

69. Pid 1
    Node A Node B
    Pid 2
    :global.register_name(“global”, self())
    Register PId1 as “global”
    Sure
    

    View Slide

70. Pid 1
    Node A Node B
    Pid 2
    :global.register_name(“global”, self())
    Register PId1 as “global”
    Sure
    :global.whereis_name(“global”) = pid1
    

    View Slide

71. Pid 1
    Node A Node B
    Pid 2
    :global.register_name(“global”, self()) :global.register_name(“global”, self())
    ?
    

    View Slide

72. (DEMO)
    

    View Slide

73. :global
    • single process registration (if everything works OK)
    • Favours availability over consistency
    • Information stored locally (reading is fast)
    • Registration is blocking (may be slow)
    

    View Slide

74. :PG2
    

    View Slide

75. Pid1
    Pid3
    Pid2
    []
    []
    []
    

    View Slide

76. Pid1
    Pid3
    Pid2
    :pg2.create(“my_group”)
    []
    []
    []
    

    View Slide

77. Pid1
    Pid3
    Pid2
    []
    []
    []
    join
    join
    :pg2.join(“my_group”, self()
    

    View Slide

78. Pid1
    Pid3
    Pid2
    []
    [pid1]
    []
    Monitor
    Monitor
    :pg2.join(“my_group”, self()
    

    View Slide

79. Pid1
    Pid3
    Pid2
    [pid1]
    [pid1]
    [pid1]
    Monitor
    Monitor
    :pg2.join(“my_group”, self()
    

    View Slide

80. Pid1
    Pid3
    Pid2
    [pid1]
    [pid1]
    [pid1]
    

    View Slide

81. Pid1
    Pid3
    Pid2
    :pg2.join(“my_group”, self()
    [pid1]
    [pid1, pid2]
    [pid1]
    

    View Slide

82. Pid1
    Pid3
    Pid2
    join
    :pg2.join(“my_group”, self()
    join
    [pid1, pid2]
    [pid1, pid2]
    [pid1, pid2]
    

    View Slide

83. Pid1
    Pid3
    Pid2
    [pid1]
    [pid2]
    [pid1]
    

    View Slide

84. Pid1
    Pid3
    Pid2
    [pid1]
    [pid2]
    [pid1]
    

    View Slide

85. Pid1
    Pid3
    Pid2
    [pid1]
    [pid2]
    [pid1]
    

    View Slide

86. Pid1
    Pid3
    Pid2
    [pid1, pid2]
    [pid1, pid2]
    [pid1, pid2]
    

    View Slide

87. It will heal, but the state in inconsistent for
    some time
    

    View Slide

88. What does it matter?
    

    View Slide

89. Node A
    Pg2
    Pg2
    Pg2
    Node B Node C
    

    View Slide

90. Node A
    Pg2
    Pg2
    Pg2
    Node B Node C
    Phoenix Channels
    

    View Slide

91. Node A
    Pg2
    Pg2
    Pg2
    Node B Node C
    Phoenix Presence
    

    View Slide

92. Node A
    Pg2
    Pg2
    Pg2
    Node B Node C
    Phoenix Channels
    

    View Slide

93. :pg2
    • Process groups
    • Favours availability over consistency
    • Information stored locally (reading is fast)
    • Registration is blocking (may be slow)
    

    View Slide

94. Strongly consistent Solutions
    

    View Slide

95. Strongly consistent Solutions
    • Consensus - Raft, Paxos, ZAB
    • Two-phase commit/THree-phase commit (2PC/3PC)
    • Read/Write quorums
    • Single database as a source of truth
    

    View Slide

96. Summary
    

    View Slide

97. Distributed Systems
    

    View Slide

98. Well, not exactly…
    

    View Slide

99. Asynchronous messages
    Distributed systems are all about
    

    View Slide

100. Really, there’s no magic
     

     View Slide

101. Just asynchronous messages between nodes
     

     View Slide

102. Just asynchronous messages between nodes
     & node failures
     

     View Slide

103. Just asynchronous messages between nodes
     & node failures
     & Communication failures
     

     View Slide

104. Just asynchronous messages between nodes
     & node failures
     & Communication failures
     & Network partitions
     

     View Slide

105. Tradeoffs
     Distributed systems are all about
     

     View Slide

106. Where to go next
     

     View Slide

107. Worth looking at
     • Riak_core
     • RAFT
     • Two-Phase Commit (2PC) / Three-Phase Commit (3PC)
     • CRDTs
     • LASP and Partisan
     

     View Slide

108. Free online (click!)
     Elixir / Erlang
     

     View Slide

109. Free PDF (Click!)
     Distributed Systems
     

     View Slide

110. Theory (The hard stuff)
     

     View Slide

111. • https://raft.github.io/ (Raft Consensus)
     • http://learnyousomeerlang.com/distribunomicon
     • https://www.rgoarchitects.com/Files/fallacies.pdf (Fallacies of distributed computing)
     • https://dzone.com/articles/better-explaining-cap-theorem (CAP Theorem)
     • https://medium.com/learn-elixir/message-order-and-delivery-guarantees-in-elixir-
     erlang-9350a3ea7541 (Elixir message delivery guarantees)
     • https://lasp-lang.readme.io/ (LASP)
     • https://arxiv.org/pdf/1802.02652.pdf (Partisan Paper)
     • https://bravenewgeek.com/tag/three-phase-commit/ (3PC)
     

     View Slide

112. We’re looking for speakers!
     

     View Slide

113. Thank You!
     Poznań Elixir Meetup #8
     

     View Slide