D1-3 Alvaro Videla - What We Talk About When We Talk About Distributed Systems

Distributed Systems are a complex topic. There’s abundant research about it but sometimes it is hard for a beginner to know where to start. I would like to outline the main concepts of distributed systems, so the interested person can have a clear path on how to start their own research as well. In this talk I will review the different models: asynchronous vs. synchronous distributed systems; message passing vs shared memory communication; failure detectors and leader election problems; consensus and different kinds of replication.

I will also review a series of books on distributed systems in order to recommend the best one according to the topics we would like to learn about, or the problems we would like to solve. The goal of the talk is to set a good foundation for people interested in learning more about distributed systems.

Talk objectives: When learning about Distributed Systems there are lot of books and papers to chose from, with many of them having titles that are hard to understand. It’s difficult then to judge their relevance to our interests if we don’t know the topic already. The goal of the talk is to lay a common ground for Distributed Systems so everyone can benefit from the current research on the topic.

Transcript

1. WHAT WE TALK ABOUT WHEN WE TALK ABOUT DISTRIBUTED SYSTEMS ALVARO VIDELA
2. DISTRIBUTED SYSTEMS FOR THE IKEA FAMILY ALVARO VIDELA
3. HTTP://BIT.LY/DIST-SYS101
4. @HINTJENS
5. DISTRIBUTED SYSTEMS
6. “A DISTRIBUTED SYSTEM IS ONE IN WHICH THE FAILURE OF A COMPUTER YOU DID NOT EVEN KNOW EXISTED CAN RENDER YOUR OWN COMPUTER UNUSABLE” Leslie Lamport
9. Google: define jargon
10. DISTRIBUTED SYSTEMS
11. DISTRIBUTED SYSTEMS • Many entities trying to solve a problem (nodes, processes)
12. DISTRIBUTED SYSTEMS • Many entities trying to solve a problem (nodes, processes) • Partial Knowledge
13. DISTRIBUTED SYSTEMS • Many entities trying to solve a problem (nodes, processes) • Partial Knowledge • Uncertainty
14. DEEP RABBIT HOLE
15. WHAT TO READ?
16. WHICH PAPERS?
22. WHICH BOOKS?
24. WHY?
25. http://tobielangel.com
26. THE PROBLEM
27. DIFFERENT MODELS
28. DIFFERENT MODELS • Timing Model
29. DIFFERENT MODELS • Timing Model • Inter Process Communication Used (IPC method)
30. DIFFERENT MODELS • Timing Model • Inter Process Communication Used (IPC method) • Failure Modes
31. TIMING MODEL
32. TIMING MODEL • Synchronous Model
33. TIMING MODEL • Synchronous Model • Asynchronous Model
34. TIMING MODEL • Synchronous Model • Asynchronous Model • Semi-synchronous Model
35. INTERPROCESS COMMUNICATION
36. INTERPROCESS COMMUNICATION • Message Passing
37. INTERPROCESS COMMUNICATION • Message Passing • Shared Memory
38. FAILURE MODES
39. FAILURE MODES • Crash-stop
40. FAILURE MODES • Crash-stop • Crash-recovery
41. FAILURE MODES • Crash-stop • Crash-recovery • Omission Faults
42. FAILURE MODES • Crash-stop • Crash-recovery • Omission Faults • Arbitrary Failures Mode (Byzantine)
43. LIVENESS AND SAFETY
44. LIVENESS AND SAFETY PROPERTIES OF ALGORITHMS
45. SAFETY Some “bad” thing does not happens during execution
46. SAFETY “Communication links should not invent messages out of thin air”
47. LIVENESS A “good” thing happens during execution
48. LIVENESS “A destination process eventually delivers the message”
49. LET’S TAKE A LOOK AT FLP1 1 - Fischer, Lynch, Paterson
51. IMPOSSIBILITY OF DISTRIBUTED CONSENSUS WITH ONE FAULTY PROCESS
52. IMPOSSIBILITY OF DISTRIBUTED CONSENSUS WITH ONE FAULTY PROCESS
53. IMPOSSIBILITY OF DISTRIBUTED CONSENSUS WITH ONE FAULTY PROCESS
54. IMPOSSIBILITY OF DISTRIBUTED CONSENSUS WITH ONE FAULTY PROCESS
55. IMPOSSIBILITY OF DISTRIBUTED CONSENSUS WITH ONE FAULTY PROCESS
56. WHAT’S CONSENSUS ANYWAY?
57. “THE CONSENSUS PROBLEM IS A PARADIGM OF AGREEMENT PROBLEMS” https://dl.acm.org/citation.cfm?id=1052796.1052806
58. PROPERTIES OF CONSENSUS
59. PROPERTIES OF CONSENSUS • C-Termination: Every correct process eventually decides on some value
60. PROPERTIES OF CONSENSUS • C-Termination: Every correct process eventually decides on some value • C-Validity: If a process decides v, then v was proposed by some process
61. PROPERTIES OF CONSENSUS • C-Termination: Every correct process eventually decides on some value • C-Validity: If a process decides v, then v was proposed by some process • C-Agreement: No two correct processes decide differently
62. PROPERTIES OF UNIFORM CONSENSUS • C-Termination: Every correct process eventually decides on some value • C-Validity: If a process decides v, then v was proposed by some process • C-Agreement: No two correct processes decide differently • C-Uniform Agreement: No two processes (correct or not) decide differently.
63. WE NEED CONSENSUS WHEN: A SET OF PROCESSES HAVE TO AGREE TO TAKE A COMMON ACTION
64. WE NEED CONSENSUS WHEN: A SET OF PROCESSES HAVE TO AGREE TO TAKE A COMMON ACTION Atomic Broadcast
65. WE NEED CONSENSUS WHEN: A SET OF PROCESSES HAVE TO AGREE TO TAKE A COMMON ACTION Atomic Broadcast Group Membership
66. ATOMIC BROADCAST “CORRECT PROCESSES DELIVER THE SAME SET OF MESSAGES IN THE SAME ORDER”
67. FLP TELLS US THAT IF CONSENSUS CANNOT BE ACHIEVED, THEN ATOMIC BROADCAST OR GROUP MEMBERSHIP CANNOT BE ACHIEVED EITHER
68. SO, WE PACK OUR BAGS AND GO? NOTHING TO SEE HERE?
69. STUMBLING OVER CONSENSUS RESEARCH: MISUNDERSTANDING AND ISSUES Marcos K. Aguilera
70. FAILURE DETECTORS
72. FAILURE DETECTORS
73. FAILURE DETECTORS • External process
74. FAILURE DETECTORS • External process • Provides information about suspected processes
75. FAILURE DETECTORS • External process • Provides information about suspected processes • Completeness property (crashed processes are detected)
76. FAILURE DETECTORS • External process • Provides information about suspected processes • Completeness property (crashed processes are detected) • Accuracy (correct process are never suspected)
77. “RUB SOME PERFECT FAILURE DETECTOR ON IT”
78. http://www.amazon.com/Introduction-Reliable-Secure- Distributed-Programming/dp/3642152597 PERFECT FAILURE DETECTOR
79. EVENTUALLY ACCURATE FAILURE DETECTOR
80. EVENTUALLY ACCURATE FAILURE DETECTOR • Strong Completeness: Eventually, every process that crashes is permanently suspected by every correct process.
81. EVENTUALLY ACCURATE FAILURE DETECTOR • Strong Completeness: Eventually, every process that crashes is permanently suspected by every correct process. • Eventual Weak Accuracy: There is a time after which some correct process is never suspected by the correct processes.
82. EVENTUALLY ACCURATE FAILURE DETECTOR • Strong Completeness: Eventually, every process that crashes is permanently suspected by every correct process. • Eventual Weak Accuracy: There is a time after which some correct process is never suspected by the correct processes. http://dl.acm.org/citation.cfm?id=1052806
84. QUORUMS
85. TL;DR: INTERSECTING SETS
86. “A QUORUM IN A SYSTEM WITH N CRASH-FAULT PROCESS ABSTRACTIONS […] IS ANY MAJORITY OF PROCESSES, I.E., ANY SET OF MORE THAN N/2 PROCESSES” QUORUMS
87. “IF F < N/2 PROCESSES FAIL BY CRASHING, THERE IS ALWAYS AT LEAST ONE QUORUM OF NONCRASHED PROCESSES IN SUCH SYSTEMS” QUORUMS
88. QUORUMS A - B - C - D - E
89. QUORUMS A - B - C - D - E
90. QUORUMS A - B’ - C - D’ - E
91. QUORUMS A - B’ - C - D’ - E
92. QUORUMS A - B’ - C’ - D’ - E’
93. QUORUMS A - B’ - C’ - D’ - E’
94. CONSISTENCY
96. CONCURRENT FIFO QUEUE
97. CONSISTENCY CONDITIONS
98. CONSISTENCY CONDITIONS • Atomic Consistency (Linearizabilty)
99. CONSISTENCY CONDITIONS • Atomic Consistency (Linearizabilty) • Sequential Consistency
100. CONSISTENCY CONDITIONS • Atomic Consistency (Linearizabilty) • Sequential Consistency • Causal Consistency
101. CONSISTENCY CONDITIONS • Atomic Consistency (Linearizabilty) • Sequential Consistency • Causal Consistency https://aphyr.com/posts/313-strong-consistency- models
102. LINEARIZABILTY http://www.amazon.com/Distributed-Algorithms-Message-Passing-Systems-Michel/dp/ 3642381227/
103. LINEARIZABILTY http://www.amazon.com/Distributed-Algorithms-Message-Passing-Systems-Michel/dp/ 3642381227/
104. SOME BOOKS
105. http://www.amazon.com/Distributed-Algorithms- Message-Passing-Systems-Michel/dp/3642381227/
106. http://www.amazon.com/Fault-tolerant-Agreement- Synchronous-Message-passing-Distributed/dp/ 1608455254/
107. http://www.amazon.com/Communication-Abstractions- Fault-tolerant-Asynchronous-Distributed/dp/160845293X/
108. http://www.amazon.com/Distributed-Algorithms- Kaufmann-Management-Systems/dp/1558603484/
109. http://www.amazon.com/Introduction-Reliable-Secure- Distributed-Programming/dp/3642152597
110. http://www.amazon.com/Guide-Reliable-Distributed- Systems-High-Assurance/dp/1447124154/
111. http://www.amazon.com/Replication-Practice-Lecture- Computer-Theoretical/dp/3642112935/
112. FINDING NON PAYWALLED PAPERS
113. CONCLUSION
114. CONCLUSION • Deep Rabbit Hole
115. CONCLUSION • Deep Rabbit Hole • Computing Science where Science is Still a Thing™
116. CONCLUSION • Deep Rabbit Hole • Computing Science where Science is Still a Thing™ • History of the Field Matters
117. CONCLUSION • Deep Rabbit Hole • Computing Science where Science is Still a Thing™ • History of the Field Matters • Read, read, read
118. THANKS! @old_sound