ebook img

Self-stabilizing systems : 5th International Workshop, WSS 2001, Lisbon, Portugal, October 1-2, 2001 : proceedings PDF

237 Pages·2001·2.8 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Self-stabilizing systems : 5th International Workshop, WSS 2001, Lisbon, Portugal, October 1-2, 2001 : proceedings

Lecture Notes in Computer Science 2194 EditedbyG.Goos,J.Hartmanis,andJ.vanLeeuwen 3 Berlin Heidelberg NewYork Barcelona HongKong London Milan Paris Tokyo Ajoy K. Datta Ted Herman (Eds.) Self-Stabilizing Systems 5th International Workshop, WSS 2001 Lisbon, Portugal, October 1-2, 2001 Proceedings 1 3 SeriesEditors GerhardGoos,KarlsruheUniversity,Germany JurisHartmanis,CornellUniversity,NY,USA JanvanLeeuwen,UtrechtUniversity,TheNetherlands VolumeEditor AjoyK.Datta UniversityofNevada,DepartmentofComputerScience Box454019,LasVegas,NV89154-4019,USA E-mail:[email protected] TedHerman UniversityofIowa,DepartmentofComputerScience IowaCity,IA52242,USA E-mail:[email protected] Cataloging-in-PublicationDataappliedfor DieDeutscheBibliothek-CIP-Einheitsaufnahme Self-stabilizingsystems:5thinternationalworkshop;proceedings/WSS 2001,Lisbon,Portugal,October1-2,2001.AjoyK.Datta;TedHerman (ed.).-Berlin;Heidelberg;NewYork;Barcelona;HongKong;London; Milan;Paris;Tokyo:Springer,2001 (Lecturenotesincomputerscience;Vol.2194) ISBN3-540-42653-1 CRSubjectClassification(1998):C.2.4,C.2,C.3,F.1,F.2.2,K.6 ISSN0302-9743 ISBN3-540-42653-1Springer-VerlagBerlinHeidelbergNewYork Thisworkissubjecttocopyright.Allrightsarereserved,whetherthewholeorpartofthematerialis concerned,specificallytherightsoftranslation,reprinting,re-useofillustrations,recitation,broadcasting, reproductiononmicrofilmsorinanyotherway,andstorageindatabanks.Duplicationofthispublication orpartsthereofispermittedonlyundertheprovisionsoftheGermanCopyrightLawofSeptember9,1965, initscurrentversion,andpermissionforusemustalwaysbeobtainedfromSpringer-Verlag.Violationsare liableforprosecutionundertheGermanCopyrightLaw. Springer-VerlagBerlinHeidelbergNewYork amemberofBertelsmannSpringerScience+BusinessMediaGmbH http://www.springer.de ©Springer-VerlagBerlinHeidelberg2001 PrintedinGermany Typesetting:Camera-readybyauthor,dataconversionbyChristianGrosche,Hamburg Printedonacid-freepaper SPIN10840630 06/3142 543210 Preface Physicalsystemswhichrightthemselvesafterbeingdisturbedevokeourcuriosity becausewewanttounderstandhowsuchsystemsareabletoreacttounexpected stimuli.Themechanismsareallthemorefascinatingwhensystemsarecomposed of small, simple units, and the ability of the system to self-stabilize emerges out of its components. Faithful computer simulations of such physical systems exhibit the self-stabilizing property, but in the realm of computing, particularly fordistributedsystems,wehavegreaterambition.Weimaginethatallmannerof software,rangingfrombasiccommunicationprotocolstohigh-levelapplications, could enjoy self-corrective properties. Self-stabilizingsoftwareoffersa unique,non-traditionalapproachtothe cru- cial problem of transient fault tolerance. Many successful instances of modern fault-tolerantnetworksarebasedonprinciples ofself-stabilization.Surprisingly, the most widely accepted technical definition of a self-stabilizing system does not refer to faults: it is the property that the system can be started in any ini- tial state, possibly an “illegal state,” and yet the system guarantees to behave properly in finite time. This, and similar definitions, break many traditional approaches to program design, in which the programmer by habit makes as- sumptions about initial conditions. The composition of self-stabilizing systems, initially seen as a daunting challenge, has been transformed into a manage- able task, thanks to an accumulation of discoveries by many investigators. Re- searchonvarioustopicsinself-stabilizationcontinuestosupplynewmethodsfor constructing self-stabilizing systems, determines limits and applicability of the paradigm of self-stabilization, and connects self-stabilization to related areas of fault tolerance and distributed computing. The Workshop on Self-Stabilizing Systems (WSS) is the main forum for re- search in the area of self-stabilization. The first workshop was held in Austin (1989), and since 1995, workshops have been held biennially: Las Vegas (1995), Santa Barbara (1997), Austin (1999), and Lisbon (2001). WSS 2001 was thus our first workshopheld outside North America, and reflected the strong growth and international participation in the area. We received 27 submitted papers for this workshop, which is a 50% increase from the previous workshops. The program committee selected 14 of the submitted papers, and Sukumar Ghosh presented our invited contribution. This volume covers many areas within the field and reflects current trends and new directions in self-stabilization. Important applications of distributed computing are topics in several papers (routing, group membership, publish- subscribe systems). Other papers strike a methodological tone, describing tools to construct self-stabilizing systems. Three papers have “agent” in their titles, which is a topic not mentioned in any previous workshop. Several papers in- vestigatenon-standarddefinitions(orweakenings)ofself-stabilization.Ourfield continues to grow and evolve. VI Preface WSS2001washeldOctober1-2inLisbonimmediatelyprecedingDISC2001, and we thank members of the DISC steering committee, who helped approve the joint WSS-DISC venture: Andr´e Schiper, Shmuel Zaks, and Michel Raynal. We are grateful to all the program committee members and reviewers for their hardand timely work.The referees thoroughlyreadeach submitted manuscript and provided extensive feedback to the authors. We also thank the following people from the University of Lisbon for local organization: Luis Rodrigues (Chair),PauloVer´ıssimo(Publicity),FilipeArau´jo(Treasurer),AlexandrePinto (Web), and Hugo Miranda (Registration). Finally we thank the following orga- nization for supporting the workshop: Fundac¸˜ao para a Ciˆencia e a Tecnologia, Minist´erio da Ciˆencia e da Tecnologia, Portugal. July 2001 Ajoy K. Datta Ted Herman Program Committee Anish Arora, The Ohio State University, USA Joffroy Beauquier, Universit´e de Paris Sud, France AjoyK.Datta,ProgramChair,UniversityofNevadaLasVegas,USA Shlomi Dolev, Ben-Gurion University of the Negev, Israel Mohamed Gouda, University of Texas at Austin, USA Ted Herman, Publications Chair, University of Iowa, USA Jaap-Henk Hoepman, University of Twente, The Netherlands Shing-Tsaan Huang, National Central University, Taiwan Sandeep Kulkarni, Michigan State University, USA Toshimitsu Masuzawa, Osaka University, Japan Franck Petit, Universit´e de Picardie, France S´ebastien Tixeuil, Publicity Chair, Universit´e de Paris Sud, France Vincent Villain, Universit´e de Picardie, France Referees A. Arora J. Beauquier P. Boldi S. Cantarell A. Ciuffoletti J. Cobb S. Dela¨et M. Demirbas S. Dolev F. Ga¨rtner C. Genolini S. Ghosh M. Gouda T. H´erault T. Herman J.-H. Hoepman R. Howell M. Gradinariu C.-T. Huang S.-T. Huang C. Johnen H. Kakugawa M. H. Karaata S. Kulkarni W. Leal T. Masuzawa N. Mittal M. Mizuno M. Nesterenko F. Petit O. Theel S. Tixeuil S. Vigna V. Villain H. Voelzer M. Yamashita Table of Contents Invited Paper: Cooperating Mobile Agents and Stabilization............. 1 Sukumar Ghosh Cross-OverComposition - Enforcement of Fairness under Unfair Adversary 19 Joffroy Beauquier, Maria Gradinariu, Colette Johnen Easy Stabilization with an Agent .................................... 35 Joffroy Beauquier, Thomas H´erault, Elad Schiller Stabilization of Routing in Directed Networks ......................... 51 Jorge A. Cobb, Mohamed G. Gouda Dijkstra’s Self-Stabilizing Algorithm in Unsupportive Environments ...... 67 Shlomi Dolev, Ted Herman Communication Adaptive Self-Stabilizing Group Membership Service (Extended Abstract) ............................................... 82 Shlomi Dolev, Elad Schiller (Im)Possibilities of Predicate Detection in Crash-Affected Systems ....... 98 Felix C. Ga¨rtner, Stefan Pleisch The Theory of Weak Stabilization.................................... 114 Mohamed G. Gouda On the Security and Vulnerability of PING............................ 124 Mohamed G. Gouda, Chin-Tser Huang, Anish Arora A New Efficient Tool for the Design of Self-Stabilizing (cid:2)-Exclusion Algorithms: The Controller.......................................... 136 Rachid Hadid, Vincent Villain Self-Stabilizing Agent Traversal...................................... 152 Ted Herman, Toshimitsu Masuzawa A Composite Stabilizing Data Structure .............................. 167 Ted Herman, Imran Pirwani Stabilizing Causal Deterministic Merge ............................... 183 Sandeep S. Kulkarni, Ravikant Fast Self-Stabilizing Depth-First Token Circulation..................... 200 Franck Petit On a Space-Optimal Distributed TraversalAlgorithm................... 216 S´ebastien Tixeuil Author Index.................................................... 229 Cooperating Mobile Agents and Stabilization Sukumar Ghosh(cid:1) The Universityof Iowa [email protected] Abstract. In the execution of distributed algorithms on a network of processes, the actions of the individual processes are scheduled by their local schedulers or demons. The schedulers communicate with their im- mediateneighborsusingsharedregistersormessagepassing.Thispaper examinesanalternativeapproachtothedesignofdistributedalgorithms, where mobile agents are allowed to traversea network, extract state in- formation,andmakeappropriatemodificationofthelocalstatestosteer the system towards a global goal. The primary emphasis of this paper is system stabilization. Both single-agent and multi-agent protocols are examined, and the advantages and disadvantages of agent-based stabi- lization are discussed. 1 Introduction Consider the execution of a program on a network of processes. The processes communicatewithoneanotherthroughsharedmemoryormessagepassing.Each processhasascheduler(alsocalledademon)thatcollectsinformationaboutthe states of the its neighboring processes, and schedules its local actions. Two of the well-known execution models rely on (i) central demon and (ii) distributed demons. In the central demon model, processes execute their actions serially, whereas in the distributed demon model, any subset of the set of enabled pro- cesses can execute their actions concurrently. This paper explores an alternative model for computation that uses mobile agents in the context of stabilization.A mobile agent[14] is a programthat can migrate from one node to another, perform various types of operations at these nodes, and take autonomous routing decisions. In contrast with messages that arepassive,anagentisanactiveentity,thatcanbecomparedwithamessenger. Conventional stabilizing distributed systems expect processes to run prede- fined programsthat have been carefully designedto guaranteerecoveryfromall possible bad configurations. However, in an open environment like the Internet where processes are not under the control of a single administration, expecting every process to modify its program for the sake of stabilization is unrealistic, and a more centralized mechanism for network administration is an viable al- ternative. This paper explores such an alternative mechanism for stabilizing a distributed system, in which processes are capable of accommodating visiting agents. (cid:1) ThisresearchwassupportedinpartbytheNationalScienceFoundationundergrant CCR-9901391. A.K.DattaandT.Herman(Eds.):WSS2001,LNCS2194,pp.1–18,2001. (cid:1)c Springer-VerlagBerlinHeidelberg2001 2 SukumarGhosh We assume thatthe underlyingsystemto be stabilizedusesmessagepassing forinterprocesscommunication.Notethatwecouldaswellconsiderinterprocess communication through shared memory, but the choice has been made only for the sakeofuniformity-agentspropagateacrosslinksexactlylikemessages.The state of the network consists of the states of each of the processes, as well as those the channels. Thus, an action by a process to update its own state, or to sendorreceivemessagesalsomodifiesthe stateofthenetwork.Thesetofstates ofthenetworkcanbeclassifiedintothecategorieslegalandillegal.Astabilizing system guarantees that regardless of the starting state, the network eventually reachesa legalstate,and remainsin the legalstate thereafter.Convergenceand closure are the two cornerstones of stabilizing systems [3]. We further assume that in addition to the ongoing activities in the above networkofprocesses,oneormoreprocessescansendoutmobileagentsthatcan migratefromoneprocesstoanother,readthelocalstatesofthevisitedprocesses, andupdatetheselocalstateswhenevernecessary.Whiletheindividualprocesses maintain the closure of legal configurations,agents take up the responsibility of steering the system to a legalconfigurationwhen they detect a configurationto beillegal.Thedetectioninvolvestakingatotalorpartialsnapshotofthesystem state. Corrective actions involve the modification of the states of one or more processes visited by the agents. In the past, Kutten, Korach, and Moran [12] used tokens with identities to solvetheleaderelectionproblem.Eachprocessinitiatesanetworktraversalwith a token labeled with its identity. In the paper Distributed Reset [4], processes first elect a leader and then construct a spanning tree with the leader as the root. Subsequently, the sending of the reset waves can be viewed as the send- ing out reset agents by the leader down the network. In the area of electronic commerce, the use of mobile agents has been steadily increasing. In network management, primitive agents mimicking biological entities like ants (see the article on Swarms [17]) traversing a network have been used in solving various problems like shortest-path computation and congestion control. In such sys- tems,the individualagentsdonothaveexplicitproblem-solvingknowledge,but intelligent action emerges from the collective action by ants. These papers pro- vide the generalmotivation behind exploring how mobile agents can be utilized to stabilize distributed systems. In our systems, an agent is a sequential pro- gram with no obvious limit on its intelligence - problems can be solved either by a single agent, or by a group of agents. Furthermore, there is the additional challenge that the initial system configuration can be arbitrary, and the agents themselves may be corrupted in transit. This paper is about using agents as a tool for stabilization, and not about a formal computational model using agents, which appears in [1]. The paper has sixsections.Section2providesageneraldescriptionoftheagentmodel.Section 3 illustrates the construction of a spanning tree using a single reliable agent. Section4explainshowtoimplementareliableagent.Section5presentstwodif- ferentsolutionstothespanningtreeconstructionproblemusingmultipleagents. Finally Section6 containssome concluding remarksaboutagent-stabilizingsys- tems.

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.