DOCSLIB.ORG

Mastering Agreement Problems in Distributed Systems

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Mastering Agreement Problems in Distributed Systems focusfault tolerance Mastering Agreement Problems in Distributed Systems Michel Raynal, IRISA Mukesh Singhal, Ohio State University istributed systems are difficult to design and implement because of the unpredictability of message transfer delays and process speeds (asynchrony) and of failures. Consequently, systems de- D signers have to cope with these difficulties intrinsically associated with distributed systems. We present the nonblocking atomic commitment (NBAC) agreement problem as a case study in the context of both synchro- nous and asynchronous distributed systems where processes can fail by Overcoming agreement crashing. This problem lends itself to fur- lem, a concurrency control conflict, and so problems in ther exploration of fundamental mecha- on) a participant cannot locally commit the distributed nisms and concepts such as time-out, multi- transaction, it votes No. A Yes vote means systems is a cast, consensus, and failure detector. We the participant commits to make local up- will survey recent theoretical results of dates permanent if required. The second primary studies related to agreement problems and phase pronounces the order to commit the challenge to show how system engineers can use this in- transaction if all participants voted Yes or systems formation to gain deeper insight into the to abort it if some participants voted No. designers. These challenges they encounter. Of course, we must enrich this protocol sketch to take into account failures to cor- authors focus on Commitment protocols rectly implement the failure atomicity prop- practical In a distributed system, a transaction erty. The underlying idea is that a failed solutions for a usually involves several sites as participants. participant is considered to have voted No. well-known At a transaction’s end, its participants must enter a commitment protocol to commit it Atomic commitment protocols agreement (when everything goes well) or abort it Researchers have proposed several atomic problem—the (when something goes wrong). Usually, this commitment protocols and implemented nonblocking protocol obeys a two-phase pattern (the them in various systems.1,2 Unfortunately, atomic two-phase commit, or 2PC). In the first some of them (such as the 2PC with a main phase, each participant votes Yes or No. If coordinator) exhibit the blocking property in commitment. for any reason (deadlock, a storage prob- some failure scenarios.3,4 “Blocking” means 40 IEEE SOFTWARE July/August 2001 0740-7459/01/$10.00 © 2001 IEEE Authorized licensed use limited to: University of Minnesota. Downloaded on January 28, 2010 at 14:57 from IEEE Xplore. Restrictions apply. The Two-Phase Commit Protocol Can Block A coordinator-based two-phase commit protocol obeys the where noncrashed participants cannot progress because of following message exchanges. First, the coordinator requests a participant crash occurrences. vote from all transaction participants and waits for their an- swers. If all participants vote Yes, the coordinator returns the decision Commit; if even a single participant votes No or Reference failed, the coordinator returns the decision Abort. 1. Ö. Babao˘glu and S. Toueg, “Non-Blocking Atomic Commitment,” Distrib- 1 uted Systems, S. Mullender, ed., ACM Press, New York, 1993, pp. Consider the following failure scenario (see Figure A). Par- 147–166. ticipant p1 is the coordinator and p2, p3, p4, and p5 are the other participants. The coordinator sends a request to all par- Compute D = Commit/Abort ticipants to get their votes. Suppose p4 and p5 answer Yes. Ac- Vote request cording to the set of answers p1 receives, it determines the de- P1 Yes/No Crash cision value D (Commit or Abort) and sends it to p2 and p3, P 2 Crash but crashes before sending it to p4 and p5. Moreover, p2 and Yes/No p crash just after receiving D. So, participants p and p are P 3 4 5 3 Crash blocked: they cannot know the decision value because they Yes ? P4 voted Yes and because p1, p2, and p3 have crashed (p4 and p5 Yes ? cannot communicate with one of them to get D). Participants P5 p4 and p5 cannot force a decision value because the forced value could be different from D. So, p4 and p5 are blocked un- til one of the crashed participants recovers. That is why the ba- Figure A. A failure scenario for the two-phase commit sic two-phase commit protocol is blocking: situations exist protocol. that nonfailed participants must wait for the tually detected. Synchronous systems provide recovery of failed participants to terminate such a guarantee, thanks to the upper bounds their commit procedure. For instance, a com- on scheduling and message transfer delays mitment protocol is blocking if it admits ex- (protocols use time-outs to safely detect fail- ecutions in which nonfailed participants can- ures in a bounded time). Asynchronous sys- not decide. When such a situation occurs, tems do not have such bounded delays, but nonfailed participants cannot release re- we can compensate for this by equipping sources they acquired for exclusive use on the them with unreliable failure detectors, as transaction’s behalf (see the sidebar “The we’ll describe later in this article. (For more Two-Phase Commit Protocol Can Block”). on the difference between synchronous and This not only prevents the concerned trans- asynchronous distributed systems, see the re- action from terminating but also prevents lated sidebar.) So, time-outs in synchronous other transactions from accessing locked systems and unreliable failure detectors in data. So, it is highly desirable to devise asynchronous systems constitute building NBAC protocols that ensure transactions blocks offering the liveness guarantee. With will terminate (by committing or aborting) these blocks we can design solutions to the despite any failure scenario. NBAC agreement problem. Several NBAC protocols (called 3PC pro- tocols) have been designed and implemented. Distributed systems and failures Basically, they add handling of failure scenar- A distributed system is composed of a fi- ios to a 2PC-like protocol and use complex nite set of sites interconnected through a subprotocols that make them difficult to un- communication network. Each site has a lo- derstand, program, prove, and test. More- cal memory and stable storage and executes over, these protocols assume the underlying one or more processes. To simplify, we as- distributed system is synchronous (that is, the sume that each site has only one process. process-scheduling delays and message trans- Processes communicate and synchronize by fer delays are upper bounded, and the proto- exchanging messages through the underly- cols know and use these bounds). ing network’s channels. A synchronous distributed system is Liveness guarantees characterized by upper bounds on message Nonblocking protocols necessitate that transfer delays, process-scheduling delays, the underlying system offers a liveness guar- and message-processing time. δ denotes an antee—a guarantee that failures will be even- upper time bound for “message transfer de- July/August 2001 IEEE SOFTWARE 41 Authorized licensed use limited to: University of Minnesota. Downloaded on January 28, 2010 at 14:57 from IEEE Xplore. Restrictions apply. A Fundamental Difference between Synchronous and Asynchronous Distributed Systems mous “Generals Paradox” shows that dis- We consider a synchronous distributed system in which a channel con- tributed systems with unreliable communi- nects each pair of processes and the only possible failure is the crash of cation do not admit solutions to the NBAC processes. To simplify, assume that only communication takes time and that problem.2 a process that receives an inquiry message responds to it immediately (in A site or transaction participant can fail ∆ zero time). Moreover, let be the upper bound of the round-trip communi- by crashing. Its state is correct until it cation delay. In this context, a process pi can easily determine the crash of crashes. A participant that does not crash ∆ another process pj. Process pi sets a timer’s time-out period to and sends during the transaction execution is also said an inquiry message to pj. If it receives an answer before the timer expires, it to be correct; otherwise, it is faulty. Because can safely conclude that pj had not crashed before receiving the inquiry we are interested only in commitment and message (see Figure B1). If pi has not received an answer from pj when the not in recovery, we assume that a faulty par- timer expires, it can safely conclude that pj ticipant does not recover. Moreover, what- ∆ has crashed (see Figure B2). ever the number of faulty participants and P In an asynchronous system, processes i the network configuration, we assume that can use timers but cannot rely on them, ir- Inquiry I am here each pair of correct participants can always P respective of the time-out period. This is i communicate. In synchronous systems, because message transfer delays in asyn- (1) crashes can be easily and safely detected by chronous distributed systems have no up- ∆ using time-out mechanisms. This is not the per bound. If pi uses a timer (with some P case in asynchronous distributed systems. ∆ i time-out period ) to detect the crash of pj, the two previous scenarios (Figures B1 and P The consensus problem B2) can occur. However, a third scenario i Crash (2) Consensus6 is a fundamental problem of can also occur (see Figure B3): the timer distributed systems (see the sidebar “Why ∆ expires while the answer is on its way to Consensus Is Important in Distributed Sys- ∆ P pi. This is because is not a correct upper i tems”). Consider a set of processes that can bound for a round-trip delay. fail by crashing. Each process has an input P Only the first and second scenarios can i value that it proposes to the other processes. occur in a synchronous distributed system.
Load more

Recommended publications
  • ACM SIGACT News Distributed Computing Column 28
    ACM SIGACT News Distributed Computing Column 28 Idit Keidar Dept. of Electrical Engineering, Technion Haifa, 32000, Israel [email protected] Sergio Rajsbaum, who edited this column for seven years and established it as a relevant and popular venue, is stepping down. This issue is my first step in the big shoes he vacated. I would like to take this opportunity to thank Sergio for providing us with seven years’ worth of interesting columns. In producing these columns, Sergio has enjoyed the support of the community at-large and obtained material from many authors, who greatly contributed to the column’s success. I hope to enjoy a similar level of support; I warmly welcome your feedback and suggestions for material to include in this column! The main two conferences in the area of principles of distributed computing, PODC and DISC, took place this summer. This issue is centered around these conferences, and more broadly, distributed computing research as reflected therein, ranging from reviews of this year’s instantiations, through influential papers in past instantiations, to examining PODC’s place within the realm of computer science. I begin with a short review of PODC’07, and highlight some “hot” trends that have taken root in PODC, as reflected in this year’s program. Some of the forthcoming columns will be dedicated to these up-and- coming research topics. This is followed by a review of this year’s DISC, by Edward (Eddie) Bortnikov. For some perspective on long-running trends in the field, I next include the announcement of this year’s Edsger W.
    [Show full text]
  • Edsger Dijkstra: the Man Who Carried Computer Science on His Shoulders
    INFERENCE / Vol. 5, No. 3 Edsger Dijkstra The Man Who Carried Computer Science on His Shoulders Krzysztof Apt s it turned out, the train I had taken from dsger dijkstra was born in Rotterdam in 1930. Nijmegen to Eindhoven arrived late. To make He described his father, at one time the president matters worse, I was then unable to find the right of the Dutch Chemical Society, as “an excellent Aoffice in the university building. When I eventually arrived Echemist,” and his mother as “a brilliant mathematician for my appointment, I was more than half an hour behind who had no job.”1 In 1948, Dijkstra achieved remarkable schedule. The professor completely ignored my profuse results when he completed secondary school at the famous apologies and proceeded to take a full hour for the meet- Erasmiaans Gymnasium in Rotterdam. His school diploma ing. It was the first time I met Edsger Wybe Dijkstra. shows that he earned the highest possible grade in no less At the time of our meeting in 1975, Dijkstra was 45 than six out of thirteen subjects. He then enrolled at the years old. The most prestigious award in computer sci- University of Leiden to study physics. ence, the ACM Turing Award, had been conferred on In September 1951, Dijkstra’s father suggested he attend him three years earlier. Almost twenty years his junior, I a three-week course on programming in Cambridge. It knew very little about the field—I had only learned what turned out to be an idea with far-reaching consequences. a flowchart was a couple of weeks earlier.
    [Show full text]
  • Mastering Concurrent Computing Through Sequential Thinking
    review articles DOI:10.1145/3363823 we do not have good tools to build ef- A 50-year history of concurrency. ficient, scalable, and reliable concur- rent systems. BY SERGIO RAJSBAUM AND MICHEL RAYNAL Concurrency was once a specialized discipline for experts, but today the chal- lenge is for the entire information tech- nology community because of two dis- ruptive phenomena: the development of Mastering networking communications, and the end of the ability to increase processors speed at an exponential rate. Increases in performance come through concur- Concurrent rency, as in multicore architectures. Concurrency is also critical to achieve fault-tolerant, distributed services, as in global databases, cloud computing, and Computing blockchain applications. Concurrent computing through sequen- tial thinking. Right from the start in the 1960s, the main way of dealing with con- through currency has been by reduction to se- quential reasoning. Transforming problems in the concurrent domain into simpler problems in the sequential Sequential domain, yields benefits for specifying, implementing, and verifying concur- rent programs. It is a two-sided strategy, together with a bridge connecting the Thinking two sides. First, a sequential specificationof an object (or service) that can be ac- key insights ˽ A main way of dealing with the enormous challenges of building concurrent systems is by reduction to sequential I must appeal to the patience of the wondering readers, thinking. Over more than 50 years, more sophisticated techniques have been suffering as I am from the sequential nature of human developed to build complex systems in communication. this way. 12 ˽ The strategy starts by designing —E.W.
    [Show full text]
  • Lecture Notes in Computer Science 2584 Edited by G
    Lecture Notes in Computer Science 2584 Edited by G. Goos, J. Hartmanis, and J. van Leeuwen 3 Berlin Heidelberg New York Barcelona Hong Kong London Milan Paris Tokyo André Schiper Alex A. Shvartsman Hakim Weatherspoon Ben Y. Zhao (Eds.) Future Directions in Distributed Computing Research and Position Papers 13 Series Editors Gerhard Goos, Karlsruhe University, Germany Juris Hartmanis, Cornell University, NY, USA Jan van Leeuwen, Utrecht University, The Netherlands Volume Editors André Schiper École Polytechnique Fédérale de Lausanne, Faculté Informatique et Communication IN-Ecublens, 1015 Lausanne, Switzerland E-mail: andre.schiper@epfl.ch Alex A. Shvartsman University of Connecticut, Computer Science and Engineering Unit 3155, Storrs, CT 06269, USA E-mail: [email protected] and [email protected] Hakim Weatherspoon Ben Y. Zhao University of California at Berkeley, Computer Science Division 447/443 Soda Hall, Berkeley, CA 94704-1776, USA E-mail: {hweather, ravenben}@cs.berkeley.edu Cataloging-in-Publication Data applied for A catalog record for this book is available from the Library of Congress Bibliographic information published by Die Deutsche Bibliothek Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliographie; detailed bibliographic data is available in the Internet at <http://dnb.ddb.de>. CR Subject Classification (1998): C.2.4, D.1.3, D.2.12, D.4.3-4, F.1.2 ISSN 0302-9743 ISBN 3-540-00912-4 Springer-Verlag Berlin Heidelberg New York This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting, reproduction on microfilms or in any other way, and storage in data banks.
    [Show full text]
  • 31 International Symposium on Distributed Computing Andréa W
    31 International Symposium on Distributed Computing DISC 2017, October 16–20, Vienna, Austria Edited by Andréa W. Richa LIPIcs – Vol. 91 – DISC2017 www.dagstuhl.de/lipics Editor Andréa W. Richa Computer Science and Engineering School of Computing, Informatics and Decision Systems Engineering (CIDSE) Arizona State University Tempe, AZ, USA [email protected] ACM Classification 1998 C.2 Computer-Communication Networks, C.2.4 Distributed Systems, D.1.3 Concurrent Programming, E.1 Data Structures, F Theory of Computation, F.1.1 Models of Computation, F.1.2 Modes of Computation ISBN 978-3-95977-053-8 Published online and open access by Schloss Dagstuhl – Leibniz-Zentrum für Informatik GmbH, Dagstuhl Publishing, Saarbrücken/Wadern, Germany. Online available at http://www.dagstuhl.de/dagpub/978-3-95977-053-8. Publication date October, 2017 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. License This work is licensed under a Creative Commons Attribution 3.0 Unported license (CC-BY 3.0): http://creativecommons.org/licenses/by/3.0/legalcode. In brief, this license authorizes each and everybody to share (to copy, distribute and transmit) the work under the following conditions, without impairing or restricting the authors’ moral rights: Attribution: The work must be attributed to its authors. The copyright is retained by the corresponding authors. Digital Object Identifier: 10.4230/LIPIcs.DISC.2017.0 ISBN 978-3-95977-053-8 ISSN 1868-8969 http://www.dagstuhl.de/lipics 0:iii LIPIcs – Leibniz International Proceedings in Informatics LIPIcs is a series of high-quality conference proceedings across all fields in informatics.
    [Show full text]
  • Byzantine Consensus in Asynchronous Message-Passing Systems: a Survey
    Int. J. Critical Computer-Based Systems, Vol. x, No. x, xxxx 1 Byzantine Consensus in Asynchronous Message-Passing Systems: a Survey Miguel Correia*, Giuliana Santos Veronese, Nuno Ferreira Neves and Paulo Verissimo Faculdade de Ciˆenciasda Universidade de Lisboa Campo Grande, 1749-016 Lisboa, Portugal E-mail: [email protected] E-mail: [email protected] E-mail: [email protected] E-mail: [email protected] *Corresponding author Abstract Consensus is a classical distributed systems problem with both theo- retical and practical interest. Asynchronous Byzantine consensus is cur- rently at the core of some solutions for the implementation of highly- resilient computing services. This paper surveys Byzantine consensus in message-passing distributed systems, by presenting the main theo- retical results in the area, the main classes of algorithms and by dis- cussing important issues like the performance and resilience of these algorithms. Keywords: distributed algorithms; distributed systems; consensus; Byzantine faults; arbitrary faults; asynchronous model; message pass- ing; Byzantine consensus Reference to this paper should be made as follows: Correia, M., Veronese, G. S., Neves, N. F. and Verissimo, P. (xxxx) ‘Byzantine Con- sensus in Asynchronous Message-Passing Systems: a Survey’, Int. J. Critical Computer-Based Systems, Vol. x, No. x, pp.xxx–xxx. Biographical notes: Miguel Correia is an assistant professor in the Department of Informatics at the Faculty of Sciences at the University of Lisboa, and an adjunct faculty member of the Carnegie Mellon In- formation Networking Institute. More information about his research can be found at http://www.di.fc.ul.pt/∼mpc.
    [Show full text]
  • Money Transfer Made Simple Alex Auvolat, Davide Frey, Michel Raynal, François Taïani
    Money Transfer Made Simple Alex Auvolat, Davide Frey, Michel Raynal, François Taïani To cite this version: Alex Auvolat, Davide Frey, Michel Raynal, François Taïani. Money Transfer Made Simple. 2020. hal-02861511v2 HAL Id: hal-02861511 https://hal.archives-ouvertes.fr/hal-02861511v2 Preprint submitted on 17 Jun 2020 (v2), last revised 16 Feb 2021 (v3) HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Money Transfer Made Simple Alex Auvolat École Normale Supérieure, Paris, France Univ Rennes, Inria, CNRS, IRISA, Rennes, France [email protected] Davide Frey Univ Rennes, Inria, CNRS, IRISA, Rennes, France [email protected] Michel Raynal Univ Rennes, Inria, CNRS, IRISA, Rennes, France Department of Computing, Polytechnic University, Hong Kong [email protected] François Taïani Univ Rennes, Inria, CNRS, IRISA, Rennes, France [email protected] Abstract It has recently been shown (PODC 2019) that, contrarily to a common belief, money transfer in the presence of faulty (Byzantine) processes does not require strong agreement such as consensus. This article goes one step further: namely, it shows that money transfers do not need to explicitly capture the causality relation that links individual transfers.
    [Show full text]
  • Distributed Computing Column 36 Distributed Computing: 2009 Edition
    Distributed Computing Column 36 Distributed Computing: 2009 Edition Idit Keidar Dept. of Electrical Engineering, Technion Haifa, 32000, Israel [email protected] It’s now the season for colorful reviews of 2009. While you can read elsewhere about the year in sports, top-40 pop hit charts, people of the year, and so on, I throw my share into the mix with a (biased) review of this year’s major distributed computing events. Awards First, let’s look at awards. This year we learned that two women were recognized with ACM and IEEE prestigious awards for their achievements in, (among other things), distributed computing. Barbara Liskov won the 2008 ACM Turing Award for a range of contributions to practical and theoretical foundations of programming language and system design, some of which are in distributed computing. The award citation mentions her impact on distributed programming, decentralized information flow, replicated storage, and modular upgrading of distributed systems. I include below a short reflection, by Rodrigo Rodrigues, on Barbara Liskov’s Turing Award and legacy. Looking ahead to 2010, it was recently announced that Nancy Lynch is the recipient of the prestigious 2010 IEEE Emanuel R. Piore Award1 for her contributions to foundations of distributed and concurrent computing. The award is given for outstanding contributions in the field of information processing in relation to computer science. Leslie Lamport has won the same award in 2004 for his seminal contributions to the theory and practice of concurrent programming and fault-tolerant computing. The 2009 Edsger W. Dijkstra Prize in Distributed Computing was awarded to Joseph Halpern and Yoram Moses for “Knowledge and Common Knowledge in a Distributed Environment”.
    [Show full text]
  • Distributed Eventual Leader Election in the Crash-Recovery and General Omission Failure Models
    University of the Basque Country UPV/EHU Department of Computer Architecture and Technology Distributed Eventual Leader Election in the Crash-Recovery and General Omission Failure Models Dissertation for the degree of Doctor of Philosophy Candidate Christian Fernández-Campusano Supervisors Roberto Cortiñas Mikel Larrea 2019 (c)2020 CHRISTIAN FERNANDEZ CAMPUSANO Abstract Distributed applications are present in many aspects of everyday life. Banking, healthcare, or transportation are some examples of such applications. These are built on top of distributed systems. Roughly speaking, a distributed system is composed of a set of processes which collaborate among them to achieve a common goal. When building such systems, designers have to cope with several issues, such as dierent synchrony assumptions and failure occurrences. Distributed systems must ensure that the delivered service is trustworthy. Agreement problems compose a fundamental class of problems in distributed systems. Agree- ment problems follow a similar pattern: all processes must agree on some common decision. Interestingly, most of the agreement problems can be considered as a particular instance of a problem called consensus, and, as a consequence, they can be solved by reduction to consensus. However, there exists a fundamental impossibility result, namely (FLP), which states that in an asynchronous distributed system, it is impossible to achieve consensus deterministically when at least one process may fail. A way to circumvent this obstacle is by using unreliable failure detectors. A failure detector is an abstraction that allows encapsulating synchrony assumptions of the system and provides (possibly incorrect) information about process failures. A particular failure detector, called Omega, has been shown to be the weakest failure detector for solving consensus with a majority of correct processes.
    [Show full text]
  • Characterizing Consensus in the Heard-Of Model
    1 Characterizing consensus in the Heard-Of model 2 A. R. Balasubramanian 3 Technical University of Munich, Germany 4 Igor Walukiewicz 5 CNRS, LaBRI, University of Bordeaux, France 6 Abstract 7 The Heard-Of model is a simple and relatively expressive model of distributed computation. Because 8 of this, it has gained a considerable attention of the verification community. We give a characterization 9 of all algorithms solving consensus in a fragment of this model. The fragment is big enough to cover 10 many prominent consensus algorithms. The characterization is purely syntactic: it is expressed in 11 terms of some conditions on the text of the algorithm. 12 2012 ACM Subject Classification Theory of computation → Distributed computing models; Software 13 and its engineering → Software verification; Theory of computation → Logic and verification 14 Keywords and phrases consensus problem, Heard-Of model, verification. 15 Digital Object Identifier 10.4230/LIPIcs.CONCUR.2020.9 16 Related Version A full version of this paper is available at https://arxiv.org/abs/2004.09621. 17 Funding A. R. Balasubramanian: UMI ReLaX, ERC Advanced Grant 787367 (PaVeS) 18 Igor Walukiewicz: ANR FREDDA ANR-17-CE40-0013 19 1 Introduction 20 Most distributed algorithms solving problems like consensus, leader election, set agreement, 21 or renaming are essentially one iterated loop. Yet, their behavior is difficult to understand due 22 to unbounded number of processes, asynchrony, failures, and other aspects of the execution 23 model. The general context of this work is to be able to say what happens when we change 24 some of the parameters: modify an algorithm or the execution model.
    [Show full text]
  • Michel Raynal an Algorithmic Approach
    Michel Raynal Fault-Tolerant Message-Passing Distributed Systems An Algorithmic Approach Fault-Tolerant Message-Passing Distributed Systems [email protected] Michel Raynal Fault-Tolerant Message-Passing Distributed Systems An Algorithmic Approach [email protected] Michel Raynal IRISA-ISTIC Université de Rennes 1 Institut Universitaire de France Rennes, France Parts of this work are based on the books “Fault-Tolerant Agreement in Synchronous Message- Passing Systems” and “Communication and Agreement Abstractions for Fault-Tolerant Asynchro- nous Distributed Systems”, author Michel Raynal, © 2010 Morgan & Claypool Publishers (www. morganclaypool.com). Used with permission. ISBN 978-3-319-94140-0 ISBN 978-3-319-94141-7 (eBook) https://doi.org/10.1007/978-3-319-94141-7 Library of Congress Control Number: 2018953101 © Springer Nature Switzerland AG 2018 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication.
    [Show full text]
  • Scalable Consensus in a Hybrid Communication Model Michel Raynal, Jiannong Cao
    One for All and All for One: Scalable Consensus in a Hybrid Communication Model Michel Raynal, Jiannong Cao To cite this version: Michel Raynal, Jiannong Cao. One for All and All for One: Scalable Consensus in a Hybrid Commu- nication Model. ICDCS’19 - 39th IEEE International Conference on Distributed Computing Systems, Jul 2019, Dallas, United States. pp.464-471, 10.1109/ICDCS.2019.00053. hal-02394259 HAL Id: hal-02394259 https://hal.archives-ouvertes.fr/hal-02394259 Submitted on 5 Dec 2019 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. One for All and All for One: Scalable Consensus in a Hybrid Communication Model Michel Raynal⋆,†, Jiannong Cao† ⋆Univ Rennes IRISA, 35042 Rennes, France †Department of Computing, Polytechnic University, Hong Kong Abstract—This paper addresses consensus in an asyn- shown that enriching the read/write system with synchro- chronous model where the processes are partitioned into clus- nization operations such as compare&swap(), fetch&add(), ters. Inside each cluster, processes can communicate through or LL/SC allows consensus to be solved for any number of a shared memory, which favors efficiency. Moreover, any pair of processes can also communicate through a message- processes, despite asynchrony and the crash of all except passing communication system, which favors scalability.
    [Show full text]