Slide 1
Slide 1 text
WHAT WE TALK ABOUT
WHEN WE TALK ABOUT
DISTRIBUTED
SYSTEMS
ALVARO VIDELA - RABBITMQ
Slide 2
Slide 2 text
DISTRIBUTED
SYSTEMS FOR
THE IKEA FAMILY
ALVARO VIDELA - RABBITMQ
Slide 3
Slide 3 text
DISTRIBUTED
SYSTEMS
Slide 4
Slide 4 text
“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
Slide 5
Slide 5 text
No content
Slide 6
Slide 6 text
No content
Slide 7
Slide 7 text
Google: define jargon
Slide 8
Slide 8 text
DISTRIBUTED
SYSTEMS
Slide 9
Slide 9 text
DISTRIBUTED
SYSTEMS
• Many entities trying to solve a problem
(nodes, processes)
Slide 10
Slide 10 text
DISTRIBUTED
SYSTEMS
• Many entities trying to solve a problem
(nodes, processes)
• Partial Knowledge
Slide 11
Slide 11 text
DISTRIBUTED
SYSTEMS
• Many entities trying to solve a problem
(nodes, processes)
• Partial Knowledge
• Uncertainty
Slide 12
Slide 12 text
DEEP RABBIT
HOLE
Slide 13
Slide 13 text
WHAT TO
READ?
Slide 14
Slide 14 text
WHICH
PAPERS?
Slide 15
Slide 15 text
No content
Slide 16
Slide 16 text
No content
Slide 17
Slide 17 text
No content
Slide 18
Slide 18 text
No content
Slide 19
Slide 19 text
No content
Slide 20
Slide 20 text
WHICH BOOKS?
Slide 21
Slide 21 text
No content
Slide 22
Slide 22 text
WHY?
Slide 23
Slide 23 text
http://tobielangel.com
Slide 24
Slide 24 text
THE PROBLEM
Slide 25
Slide 25 text
DIFFERENT MODELS
Slide 26
Slide 26 text
DIFFERENT MODELS
• Timing Model
Slide 27
Slide 27 text
DIFFERENT MODELS
• Timing Model
• Inter Process Communication Used (IPC
method)
Slide 28
Slide 28 text
DIFFERENT MODELS
• Timing Model
• Inter Process Communication Used (IPC
method)
• Failure Modes
Slide 29
Slide 29 text
TIMING MODEL
Slide 30
Slide 30 text
TIMING MODEL
• Synchronous Model
Slide 31
Slide 31 text
TIMING MODEL
• Synchronous Model
• Asynchronous Model
Slide 32
Slide 32 text
TIMING MODEL
• Synchronous Model
• Asynchronous Model
• Semi-synchronous Model
Slide 33
Slide 33 text
INTERPROCESS
COMMUNICATION
Slide 34
Slide 34 text
INTERPROCESS
COMMUNICATION
• Message Passing
Slide 35
Slide 35 text
INTERPROCESS
COMMUNICATION
• Message Passing
• Shared Memory
Slide 36
Slide 36 text
FAILURE MODES
Slide 37
Slide 37 text
FAILURE MODES
• Crash-stop
Slide 38
Slide 38 text
FAILURE MODES
• Crash-stop
• Crash-recovery
Slide 39
Slide 39 text
FAILURE MODES
• Crash-stop
• Crash-recovery
• Omission Faults
Slide 40
Slide 40 text
FAILURE MODES
• Crash-stop
• Crash-recovery
• Omission Faults
• Arbitrary Failures Mode (Byzantine)
Slide 41
Slide 41 text
LIVENESS AND
SAFETY
Slide 42
Slide 42 text
LIVENESS AND SAFETY
PROPERTIES OF ALGORITHMS
Slide 43
Slide 43 text
SAFETY
Some “bad” thing does not
happens during execution
Slide 44
Slide 44 text
SAFETY
“Communication links should not
invent messages out of thin air”
Slide 45
Slide 45 text
LIVENESS
A “good” thing happens during
execution
Slide 46
Slide 46 text
LIVENESS
“A destination process eventually
delivers the message”
Slide 47
Slide 47 text
LET’S TAKE A LOOK
AT FLP1
1 - Fischer, Lynch, Paterson
Slide 48
Slide 48 text
No content
Slide 49
Slide 49 text
IMPOSSIBILITY OF DISTRIBUTED
CONSENSUS WITH ONE FAULTY
PROCESS
Slide 50
Slide 50 text
IMPOSSIBILITY OF DISTRIBUTED
CONSENSUS WITH ONE FAULTY
PROCESS
Slide 51
Slide 51 text
IMPOSSIBILITY OF DISTRIBUTED
CONSENSUS WITH ONE FAULTY
PROCESS
Slide 52
Slide 52 text
IMPOSSIBILITY OF DISTRIBUTED
CONSENSUS WITH ONE FAULTY
PROCESS
Slide 53
Slide 53 text
IMPOSSIBILITY OF DISTRIBUTED
CONSENSUS WITH ONE FAULTY
PROCESS
Slide 54
Slide 54 text
WHAT’S CONSENSUS
ANYWAY?
Slide 55
Slide 55 text
“THE CONSENSUS
PROBLEM IS A PARADIGM
OF AGREEMENT
PROBLEMS”
https://dl.acm.org/citation.cfm?id=1052796.1052806
Slide 56
Slide 56 text
PROPERTIES OF
CONSENSUS
Slide 57
Slide 57 text
PROPERTIES OF
CONSENSUS
• C-Termination: Every correct process eventually decides on some value
Slide 58
Slide 58 text
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
Slide 59
Slide 59 text
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
Slide 60
Slide 60 text
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.
Slide 61
Slide 61 text
WE NEED CONSENSUS
WHEN:
A SET OF PROCESSES
HAVE TO AGREE TO TAKE
A COMMON ACTION
Slide 62
Slide 62 text
WE NEED CONSENSUS
WHEN:
A SET OF PROCESSES
HAVE TO AGREE TO TAKE
A COMMON ACTION
Atomic
Broadcast
Slide 63
Slide 63 text
WE NEED CONSENSUS
WHEN:
A SET OF PROCESSES
HAVE TO AGREE TO TAKE
A COMMON ACTION
Atomic
Broadcast
Group
Membership
Slide 64
Slide 64 text
ATOMIC BROADCAST
“CORRECT PROCESSES
DELIVER THE SAME SET OF
MESSAGES IN THE SAME
ORDER”
Slide 65
Slide 65 text
FLP TELLS US THAT IF
CONSENSUS CANNOT BE
ACHIEVED, THEN ATOMIC
BROADCAST OR GROUP
MEMBERSHIP CANNOT BE
ACHIEVED EITHER
Slide 66
Slide 66 text
SO, WE PACK OUR BAGS
AND GO?
NOTHING TO SEE HERE?
Slide 67
Slide 67 text
STUMBLING OVER
CONSENSUS RESEARCH:
MISUNDERSTANDING AND
ISSUES
Marcos K. Aguilera
Slide 68
Slide 68 text
FAILURE
DETECTORS
Slide 69
Slide 69 text
No content
Slide 70
Slide 70 text
FAILURE DETECTORS
Slide 71
Slide 71 text
FAILURE DETECTORS
• External process
Slide 72
Slide 72 text
FAILURE DETECTORS
• External process
• Provides information about suspected processes
Slide 73
Slide 73 text
FAILURE DETECTORS
• External process
• Provides information about suspected processes
• Completeness property (crashed processes are
detected)
Slide 74
Slide 74 text
FAILURE DETECTORS
• External process
• Provides information about suspected processes
• Completeness property (crashed processes are
detected)
• Accuracy (correct process are never suspected)
Slide 75
Slide 75 text
“RUB SOME PERFECT
FAILURE DETECTOR
ON IT”
Slide 76
Slide 76 text
http://www.amazon.com/Introduction-Reliable-Secure-
Distributed-Programming/dp/3642152597
PERFECT FAILURE
DETECTOR
Slide 77
Slide 77 text
EVENTUALLY ACCURATE
FAILURE DETECTOR
Slide 78
Slide 78 text
EVENTUALLY ACCURATE
FAILURE DETECTOR
• Strong Completeness: Eventually, every
process that crashes is permanently
suspected by every correct process.
Slide 79
Slide 79 text
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.
Slide 80
Slide 80 text
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
Slide 81
Slide 81 text
No content
Slide 82
Slide 82 text
QUORUMS
Slide 83
Slide 83 text
TL;DR:
INTERSECTING
SETS
Slide 84
Slide 84 text
“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
Slide 85
Slide 85 text
“IF F < N/2 PROCESSES FAIL BY
CRASHING, THERE IS ALWAYS AT
LEAST ONE QUORUM OF
NONCRASHED PROCESSES IN SUCH
SYSTEMS”
QUORUMS
Slide 86
Slide 86 text
CONSISTENCY
Slide 87
Slide 87 text
No content
Slide 88
Slide 88 text
CONCURRENT
FIFO QUEUE
Slide 89
Slide 89 text
CONSISTENCY
CONDITIONS
Slide 90
Slide 90 text
CONSISTENCY
CONDITIONS
• Atomic Consistency (Linearizabilty)
Slide 91
Slide 91 text
CONSISTENCY
CONDITIONS
• Atomic Consistency (Linearizabilty)
• Sequential Consistency
Slide 92
Slide 92 text
CONSISTENCY
CONDITIONS
• Atomic Consistency (Linearizabilty)
• Sequential Consistency
• Causal Consistency
Slide 93
Slide 93 text
CONSISTENCY
CONDITIONS
• Atomic Consistency (Linearizabilty)
• Sequential Consistency
• Causal Consistency
https://aphyr.com/posts/313-strong-consistency-
models
Slide 94
Slide 94 text
LINEARIZABILTY
http://www.amazon.com/Distributed-Algorithms-Message-Passing-Systems-Michel/dp/
3642381227/
Slide 95
Slide 95 text
LINEARIZABILTY
http://www.amazon.com/Distributed-Algorithms-Message-Passing-Systems-Michel/dp/
3642381227/
Slide 96
Slide 96 text
SOME BOOKS
Slide 97
Slide 97 text
http://www.amazon.com/Distributed-Algorithms-
Message-Passing-Systems-Michel/dp/3642381227/
Slide 98
Slide 98 text
http://www.amazon.com/Fault-tolerant-Agreement-
Synchronous-Message-passing-Distributed/dp/
1608455254/
Slide 99
Slide 99 text
http://www.amazon.com/Communication-Abstractions-
Fault-tolerant-Asynchronous-Distributed/dp/160845293X/
Slide 100
Slide 100 text
http://www.amazon.com/Distributed-Algorithms-
Kaufmann-Management-Systems/dp/1558603484/
Slide 101
Slide 101 text
http://www.amazon.com/Introduction-Reliable-Secure-
Distributed-Programming/dp/3642152597
Slide 102
Slide 102 text
http://www.amazon.com/Guide-Reliable-Distributed-
Systems-High-Assurance/dp/1447124154/
Slide 103
Slide 103 text
http://www.amazon.com/Replication-Practice-Lecture-
Computer-Theoretical/dp/3642112935/
Slide 104
Slide 104 text
FINDING NON
PAYWALLED PAPERS
Slide 105
Slide 105 text
CONCLUSION
Slide 106
Slide 106 text
CONCLUSION
• Deep Rabbit Hole
Slide 107
Slide 107 text
CONCLUSION
• Deep Rabbit Hole
• Computing Science where Science is Still a Thing™
Slide 108
Slide 108 text
CONCLUSION
• Deep Rabbit Hole
• Computing Science where Science is Still a Thing™
• History of the Field Matters
Slide 109
Slide 109 text
CONCLUSION
• Deep Rabbit Hole
• Computing Science where Science is Still a Thing™
• History of the Field Matters
• Read, read, read
Slide 110
Slide 110 text
THANKS!
@old_sound