Byzantine Generals A Lecture in CE Freshman Seminar Series:

Slides:



Advertisements
Similar presentations
Byzantine Generals. Outline r Byzantine generals problem.
Advertisements

Agreement: Byzantine Generals UNIVERSITY of WISCONSIN-MADISON Computer Sciences Department CS 739 Distributed Systems Andrea C. Arpaci-Dusseau Paper: “The.
Apr. 2007SatisfiabilitySlide 1 Satisfiability A Lecture in CE Freshman Seminar Series: Ten Puzzling Problems in Computer Engineering.
Teaser - Introduction to Distributed Computing
Apr. 2012Byzantine GeneralsSlide 1 Byzantine Generals A Lecture in CE Freshman Seminar Series: Ten Puzzling Problems in Computer Engineering.
? What is Computer Science and what can you do with it.
The Byzantine Generals Problem Boon Thau Loo CS294-4.
Prepared by Ilya Kolchinsky.  n generals, communicating through messengers  some of the generals (up to m) might be traitors  all loyal generals should.
CSE 486/586, Spring 2012 CSE 486/586 Distributed Systems Byzantine Fault Tolerance Steve Ko Computer Sciences and Engineering University at Buffalo.
DISTRIBUTED SYSTEMS II FAULT-TOLERANT AGREEMENT Prof Philippas Tsigas Distributed Computing and Systems Research Group.
Byzantine Generals Problem: Solution using signed messages.
 Two armies are camped on the outskirts of either side of an enemy city.
May 2015Binary SearchSlide 1 Binary Search A Lecture in CE Freshman Seminar Series: Ten Puzzling Problems in Computer Engineering.
June 2007Malfunction DiagnosisSlide 1 Malfunction Diagnosis A Lecture in CE Freshman Seminar Series: Ten Puzzling Problems in Computer Engineering.
Dec. 2006Agreement and AdjudicationSlide 1 Fault-Tolerant Computing Software Design Methods.
Copyright 2006 Koren & Krishna ECE655/ByzGen.1 UNIVERSITY OF MASSACHUSETTS Dept. of Electrical & Computer Engineering Fault Tolerant Computing ECE 655.
A Look at Byzantine Generals Problem R J Walters.
HandoutMay 2007Byzantine Generals A Lecture in CE Freshman Seminar Series: Ten Puzzling Problems in Computer Engineering.
The Island of Liars and Truth-Tellers
Last Class: Weak Consistency
Jan. 2005Logical ReasoningSlide 1 Logical Reasoning A Lesson in the “Math + Fun!” Series.
Apr. 2007Easy, Hard, Impossible!Slide 1 Easy, Hard, Impossible! A Lecture in CE Freshman Seminar Series: Ten Puzzling Problems in Computer Engineering.
May 2007Binary SearchSlide 1 Binary Search A Lecture in CE Freshman Seminar Series: Ten Puzzling Problems in Computer Engineering.
May 2012Malfunction DiagnosisSlide 1 Malfunction Diagnosis A Lecture in CE Freshman Seminar Series: Ten Puzzling Problems in Computer Engineering.
May 2007String MatchingSlide 1 String Matching A Lecture in CE Freshman Seminar Series: Ten Puzzling Problems in Computer Engineering.
Distributed Consensus Reaching agreement is a fundamental problem in distributed computing. Some examples are Leader election / Mutual Exclusion Commit.
DISTRIBUTED SYSTEMS II FAULT-TOLERANT AGREEMENT Prof Philippas Tsigas Distributed Computing and Systems Research Group.
1 Chapter 12 Consensus ( Fault Tolerance). 2 Reliable Systems Distributed processing creates faster systems by exploiting parallelism but also improve.
CSE 486/586 CSE 486/586 Distributed Systems Byzantine Fault Tolerance Steve Ko Computer Sciences and Engineering University at Buffalo.
Chap 15. Agreement. Problem Processes need to agree on a single bit No link failures A process can fail by crashing (no malicious behavior) Messages take.
UNIVERSITY of WISCONSIN-MADISON Computer Sciences Department
CSE 486/586, Spring 2013 CSE 486/586 Distributed Systems Byzantine Fault Tolerance Steve Ko Computer Sciences and Engineering University at Buffalo.
PROCESS RESILIENCE By Ravalika Pola. outline: Process Resilience  Design Issues  Failure Masking and Replication  Agreement in Faulty Systems  Failure.
CSE 486/586, Spring 2014 CSE 486/586 Distributed Systems Byzantine Fault Tolerance Steve Ko Computer Sciences and Engineering University at Buffalo.
CSE 486/586 Distributed Systems Byzantine Fault Tolerance
The Consensus Problem in Fault Tolerant Computing
reaching agreement in the presence of faults
Synchronizing Processes
COMP28112 – Lecture 15 Byzantine fault tolerance: dealing with arbitrary failures The Byzantine Generals’ problem (Byzantine Agreement) 26-Jan-18 COMP28112.
Splash Screen.
Coordination and Agreement
ECE 313 Probability with Engineering Applications Lecture 7
EECS 203 Lecture 19 Graphs.
CASCADE: AN ATTACK-RESISTANT DHT WITH MINIMAL HARD STATE
Algorithmic Problem Solving
When Is Agreement Possible
Distributed Computing
CSE 486/586 Distributed Systems Wrap-up
EECS 203 Lecture 20 More Graphs.
8.2. Process resilience Shreyas Karandikar.
C ODEBREAKER Class discussion.
COMP28112 – Lecture 14 Byzantine fault tolerance: dealing with arbitrary failures The Byzantine Generals’ problem (Byzantine Agreement) 13-Oct-18 COMP28112.
Satisfiability A Lecture in CE Freshman Seminar Series:
CSE 486/586 Distributed Systems Byzantine Fault Tolerance
Unsure about how to decide which sorority to join
COMP28112 – Lecture 13 Byzantine fault tolerance: dealing with arbitrary failures The Byzantine Generals’ problem (Byzantine Agreement) 19-Nov-18 COMP28112.
Distributed Consensus
Outline Introduction Background Distributed DBMS Architecture
Distributed Systems, Consensus and Replicated State Machines
Distributed Consensus
Consensus in Synchronous Systems: Byzantine Generals Problem
The Byzantine Generals Problem
COMP28112 – Lecture 13 Byzantine fault tolerance: dealing with arbitrary failures The Byzantine Generals’ problem (Byzantine Agreement) 22-Feb-19 COMP28112.
Sorting Networks A Lecture in CE Freshman Seminar Series:
Easy, Hard, Impossible! A Lecture in CE Freshman Seminar Series:
COMP28112 – Lecture 13 Byzantine fault tolerance: dealing with arbitrary failures The Byzantine Generals’ problem (Byzantine Agreement) 24-Apr-19 COMP28112.
Malfunction Diagnosis
Binary Search A Lecture in CE Freshman Seminar Series:
Byzantine Generals Problem
CSE 486/586 Distributed Systems Byzantine Fault Tolerance
Presentation transcript:

Byzantine Generals A Lecture in CE Freshman Seminar Series: Ten Puzzling Problems in Computer Engineering Apr. 2016 Byzantine Generals

About This Presentation This presentation belongs to the lecture series entitled “Ten Puzzling Problems in Computer Engineering,” devised for a ten-week, one-unit, freshman seminar course by Behrooz Parhami, Professor of Computer Engineering at University of California, Santa Barbara. The material can be used freely in teaching and other educational settings. Unauthorized uses, including any use for financial gain, are prohibited. © Behrooz Parhami Edition Released Revised First May 2007 Apr. 2008 Apr. 2009 Apr. 2010 Apr. 2011 Apr. 2012 Apr. 2015 Apr. 2016 Apr. 2016 Byzantine Generals

Reminder on ECE 1’s Theme and Direction Many engineering problems are puzzle-like (especially in CE) A puzzling problem: looks deceptively simple, but ... Each lecture starts with puzzles that we try to solve together I introduce you to CE problems that are related to the puzzles appears very difficult, or even impossible, but is readily tamed with the appropriate insight Topics thus far: Easy, Hard, Impossible (Collatz conjecture) Placement and Routing (houses & utilities) Satisfiability (making change) Cryptography (secret message) Byzantine Generals (liars and truth-tellers) Topics for the 2nd half: Binary Search (counterfeit coin) Task Scheduling (Sudoku) String Matching (word search) Sorting Networks (rearranging trains) Malfunction Diagnosis (logical reasoning)   Apr. 2016 Byzantine Generals

Do you realize how dead you will be if she realizes this is your weekly “bar meeting”? Apr. 2016 Byzantine Generals

The Island of Liars and Truth-Tellers Setting for puzzles in the next few slides: You are on an island populated by two tribes. Members of one tribe consistently lie. Members of the other tribe always tell the truth. Tribe members can recognize one another, but you can’t tell them apart. You run into a man on the island and ask him whether he is a truth-teller. A blaring siren prevents you from hearing his answer. You inquire, “Sorry, did you say you’re a truth-teller?” He responds: “No, I did not.” To which tribe does the man belong? He is a liar You meet a woman on the island. What single (yes/no) question can you ask her to determine whether she is a liar or a truth-teller? If I asked you whether you were a liar, what would your answer be?   Apr. 2016 Byzantine Generals

Meeting Two People on the Island You meet two people A and B on the island. A says, “Both of us are from the liars tribe.” Which tribe is A from? What about B? A: Liar, B: TT You meet two people, C and D on the island. C says, “Exactly one of us is from the liars tribe.” Which tribe is D from? D: Liar You meet two people E and F on the island. E says, “It is not the case that both of us are from the truth-tellers tribe.” Which tribe is E from? What about F? E: TT, F: Liar You meet two people, G and H on the island. Each of the two makes a statement. Which tribes are G and H from? G says: “We are from different tribes.” H says: “G is from the liars tribe.” G: TT, H: Liar   Apr. 2016 Byzantine Generals

Meeting Three People on the Island You meet three people A, B, C. You ask A, “How many among you are truth-tellers?” You don’t hear her answer, so you ask B, “What did she just say?” “She said one,” he replies. Then C adds, “Don’t believe him, he is lying!” What can you say about B’s or C’s tribe? A Liar TT B C Answer: B is a liar, C is a truth teller Solution by exhaustive search 1 2 3   Apr. 2016 Byzantine Generals

More Examples and a Variation Is it possible to ask the exact same question twice of a truth-teller and get two different answers? Yes, just use any question whose answer is time-dependent Or say, “Did I just ask you a question for the second time?” Twelve politicians from the island go to a city hall meeting. The 1st one says: “Not a single person in this room tells the truth.” The 2nd one says: “No more than one person in this room tells the truth.” The 3rd one says: “No more than two people in this room tell the truth.” . . . The 12th one says: “No more than 11 people in this room tell the truth.” What can you say about the composition of this group of politicians? Truth-tellers and nay-sayers: You are allowed only yes/no questions. One group of people answer truthfully and the other always answer “no” How can you tell a truth-teller apart from a nay-sayer?   Apr. 2016 Byzantine Generals

Other Interesting Variations Liars who lie selectively; for example, in answer to every other question or on certain days of the week Inhabitants of another island lie consistently on Tuesdays, Thursdays, and Saturdays, and they tell the truth on the other four days of the week. You have forgotten what day of the week it is, so you ask a passerby. “Saturday,” he answers. “And what day will it be tomorrow?” you inquire. “Wednesday,” he replies. Can you tell what day it is today? M Tu W Th F Su Sa Today cannot be M, W, F, Sa, or Su. So it is: Tu or Th Tomorrow cannot be M, F, or Su. So it is: T, W, Th, or Sa   Apr. 2016 Byzantine Generals

Liars, Randoms, and Truth-Tellers Setting for more complex puzzles: You are on an island populated by three tribes. Members of one tribe always tell the truth. Members of the second tribe choose to tell the truth or lie, completely at random. Members of the third tribe consistently lie. Tribe members can recognize one another, but you can’t tell them apart. Three people from the island, one representing each tribe, come to visit. How can you identify who is from which tribe by asking only three yes/no questions? Each question must be directed at only one person, but you can ask the same person multiple questions. P1 P2 P3 First question must be such that in one position R is excluded and neither subset associated with the two possible answers has more than four members. For example asking the leftmost person whether the letters corresponding to the two others are in alphabetical order left to right divides the set into {LTR, RLT, RTL, TLR} and {LRT, RLT, RTL, TRL}. The next two questions are asked of the person in the position holding only L or T. For example, one can ask whether the random person is in the leftmost position. Hint : There are 6 possibilities for P1, P2, P3: LRT, LTR, RLT, RTL, TLR, TRL Additional hint: Ask the leftmost person whether letters corresponding to the other two appear in alphabetical order. Then, regardless of the answer, you will have one position which excludes a random person.   Apr. 2016 Byzantine Generals

The Two Generals’ Paradox Troops led by two generals are camped on the outskirts of an enemy city The generals can only communicate via messengers who must travel through enemy territory and are thus subject to delays or capture The two generals have previously agreed on a plan of attack, but they must communicate to set up the attack time Not attacking together has dire results Unreliable commun. G1 G2 G1 decides to send the message, “Let’s attack at noon tomorrow” ATM: dispense cash Database: reduce balance G1 will not attack before getting an acknowledgment from G2 G2 will not attack before making sure that his acknowledgment was received by G1 (because he knows G1 would not attack otherwise), so he waits for an acknowledgment of his acknowledgment   Apr. 2016 Byzantine Generals

Liars, Randoms, and Truth-Tellers Stand for ... Sites communicating with one another to reach an agreement (e.g., to select a coordinating site, often called “leader”) Site status … Healthy: Gives the appropriate response to every message Truth-teller Crashed: Does not respond to any message Quiet Permanently failed: May respond identically to every message Nay-sayer Permanently failed: May give the wrong response consistently Liar Arbitrarily failed: May give an unpredictable response Random Maliciously failed: Gives a response that is calculated to do the maximum harm (adversary, worst-case failure) Byzantine   Apr. 2016 Byzantine Generals

The Byzantine Generals Problem A general and n – 1 lieutenant generals lead n divisions of the Byzantine army camped on the outskirts of an enemy city The n divisions can only communicate via messengers, who may be captured or arbitrarily delayed (due to the need to hide for a while) We seek a scheme for the generals to agree on a common plan of action (attack or retreat), even if some of the generals are traitors who will do anything to prevent loyal generals from reaching agreement G The problem is nontrivial even if messengers are totally reliable With unreliable messengers, the problem becomes very complex L2 L1   Apr. 2016 Byzantine Generals

Byzantine Generals with Reliable Messengers “Attack” Traitor G L2 L1 “Attack” “Retreat” “He said retreat” Traitor “He said retreat” With f Byzantine failures,  3f + 1 nodes needed to reach agreement G L3 L1 a Traitor L2 ? G L3 L1 a c Traitor L2 b   Apr. 2016 Byzantine Generals

3f + 1 Generals Needed with f Traitors By exchanging messages in multiple rounds, the 2f + 1 loyal generals can eventually reach a common plan of action which matches the order of the commanding general, provided the latter is loyal Loyal Some deem Byzantine faults very unlikely and not worth considering In “The Real Byzantine Generals,” the authors show why Byzantine faults are real and must be treated in both hardware and software http://ieeexplore.ieee.org/iel5/9579/30281/01390734.pdf “If a designer spent 50 hours per week, 52 weeks per year, for 35 years staring at one system, that would be less than 105 hours . . . far short of typical avionics requirements.” 50  52  35 = 91,000   Apr. 2016 Byzantine Generals

Without Malicious Faults, Voting Will Do Data fusion: Obtaining dependable results from potentially incorrect, inaccurate, or incomplete data Fuser y n Majority opinion Centralized voting with majority rule Approximate voting with imprecise inputs: E.g., temperature readings of 78.2, 45.5, 79.1, 78.7, 21.2, 120.0, 77.6 78.4 Mean of reasonable inputs Ask seven students to write down the time (hour and minutes) on sticky notes. Put them on the board in sorted order. Note the median value. Then, eliminate the top and bottom two, and average the remaining three. Median of all inputs 78.2 Distributed voting: Same concept, provided erroneous values are seen identically by the fusion processes at all sites Apr. 2016 Byzantine Generals

Voting Comes in Many Flavors Example: What time is it? Seven students write the exact time (hour and minute) on sticky notes Sort the sticky notes on the board Pick one of the following values: Fuser x1 x2 x3 x4 x7 x6 x5 x Derived opinion Majority, if a majority exists Plurality, if a plurality exists Interval voting: Each proposer supplies a range of values that is guaranteed to hold the correct value Median of all the values proposed Mean of all the values proposed Ask seven students to write down the time (hour and minutes) on sticky notes. Put them on the board in sorted order. Note the median value. Then, eliminate the top and bottom two, and average the remaining three. Mean of five values, after removing the largest and smallest of the seven Mean of three values, after removing the 2 largest and 2 smallest values Apr. 2016 Byzantine Generals

Mathematics of Voting Voting studied in several fields: Mathematics / Computing Political science Sociology (social choice theory) Economics Fuser x1 x2 x3 x4 x7 x6 x5 x Derived opinion No voting scheme is totally fool-proof Regular voting: Candidates A1, A2, B More qualified candidates A1 and A2 may split the votes, leading to the election of B (run-off helps in solving this problem, but creates others) Approval voting Vote for any number of candidates you like Borda voting comes pretty close to an ideal voting scheme Each participant ranks all candidates; tally votes by giving n points to each 1st-place choice, n – 1 points for 2nd place, ... , 1 point for nth place Apr. 2016 Byzantine Generals

With Signed Messages, Agreement is Easy L2 L1 “Attack: G” Traitor G L2 L1 “Attack: G” “Retreat: G” “Retreat: G: L2” Traitor Signature of G “Attack: G: L2” “Nonsense: L2” L2 may take two actions – Forward the signed message: This leads to correct outcome Send a different fake message that is recognized by L1 as fake (loyal generals ignore messages coming from known traitors) G L2 L1 “Attack: G” Traitor “Attack: G: L2” Apr. 2016 Byzantine Generals