P2P外文文献.doc

Peer-to-Peer ComputingDejan S. Milojicic, Vana Kalogeraki, Rajan Lukose,Kiran Nagaraja1, Jim Pruyne, Bruno Richard,Sami Rollins 2 , Zhichen XuHP Laboratories Palo AltoHPL-2002-57 (R.1)July 3rd , 2003*E-mail: dejan, vana, lukose, pruyne, zhichen , bruno_richard ,knagaraj cs.rutgers.edu, srollins cs.ucsb.edupeer-to-peer,decentralization,self-organization,anonymity, costof ownershipThe term “peer-to-peer” (P2P) refers to a class of systems andapplications that employ distributed resources to perform a functionin a decentralized manner. With the pervasive deployment ofcomputers, P2P is increasingly receiving attention in research,product development, and investment circles. Some of the benefitsof a P2P approach include: improving scalability by avoidingdependency on centralized points; eliminating the need for costlyinfrastructure by enabling direct communication among clients; andenabling resource aggregation.This survey reviews the field of P2P systems and applications bysummarizing the key concepts and giving an overview of the mostimportant systems. Design and implementation issues of P2Psystems are analyzed in general, and then revisited for eight casestudies. This survey will help people in the research community andindustry understand the potential benefits of P2P. For peopleunfamiliar with the field it provides a general overview, as well asdetailed case studies. Comparison of P2P solutions with alternativearchitectures is intended for users, developers, and systemadministrators (IT).* Internal Accession Date OnlyApproved for External Publication?12Rutgers University, NJ, 08901University of California at Santa Barbara, CA, 93106?© Copyright Hewlett-Packard Company 2002Peer-to-Peer ComputingDEJAN S. MILOJICIC1, VANA KALOGERAKI1, RAJAN LUKOSE1, KIRAN NAGARAJA2,JIM PRUYNE1, BRUNO RICHARD1, SAMI ROLLINS3, and ZHICHEN XU1dejan, vana, lukose, pruyne, zhichen, bruno_richard,knagarajcs.rutgers.edu, srollinscs.ucsb.edu1HP Labs, 2Rutgers University, 3University of California at Santa BarbaraAbstractThe term “peer-to-peer” (P2P) refers to a class of systems and applications that employ distributed resources toperform a function in a decentralized manner. With the pervasive deployment of computers, P2P is increasinglyreceiving attention in research, product development, and investment circles. Some of the benefits of a P2P ap-proach include: improving scalability by avoiding dependency on centralized points; eliminating the need forcostly infrastructure by enabling direct communication among clients; and enabling resource aggregation.This survey reviews the field of P2P systems and applications by summarizing the key concepts and giving anoverview of the most important systems. Design and implementation issues of P2P systems are analyzed in gen-eral, and then revisited for eight case studies. This survey will help people in the research community and industryunderstand the potential benefits of P2P. For people unfamiliar with the field it provides a general overview, aswell as detailed case studies. Comparison of P2P solutions with alternative architectures is intended for users, de-velopers, and system administrators (IT).Categories and Subject Descriptors: C.2.4 Computer-Communication Networks: Distributed Systems - network oper-ating systems; D.1.3 Programming Techniques: Concurrent Programming - distributed programming; D.4.7 OperatingSystems: Organization and Design - distributed systems; E.1 Data: Data Structures - distributed data structures; F.1.2Theory of Computation: Modes of Computation - parallelism and concurrency; H.3.4 Information Systems: Systemsand Software - Distributed systems.General Terms: design, experimentationAdditional Key Words and Phrases: peer-to-peer, decentralization, self-organization, anonymity, cost of ownership.1 INTRODUCTIONPeer-to-Peer (P2P) computing is a very controversialtopic. Many experts believe that there is not much newin P2P. There is a lot of confusion: what really consti-tutes P2P? For example, is distributed computing reallyP2P or not? We believe that P2P does warrant a thor-ough analysis. The goals of the paper are threefold: 1)to understand what P2P is and it is not, as well as whatis new, 2) to offer a thorough analysis of and examplesof P2P computing, and 3) to analyze the potential ofP2P computing.The term “peer-to-peer” refers to a class of systems andapplications that employ distributed resources to per-form a function in a decentralized manner. The resourc-es encompass computing power, data (storage andcontent), network bandwidth, and presence (comput-ers, human, and other resources). The critical functioncan be distributed computing, data/content sharing,communication and collaboration, or platform services.Decentralization may apply to algorithms, data, andmeta-data, or to all of them. This does not preclude re-taining centralization in some parts of the systems andapplications. Typical P2P systems reside on the edge ofthe Internet or in ad-hoc networks. P2P enables:· valuable externalities, by aggregating resourcesthrough low-cost interoperability, the whole is madegreater than the sum of its parts· lower cost of ownership and cost sharing, by usingexisting infrastructure and by eliminating or distribut-ing the maintenance costs· anonymity/privacy, by incorporating these require-ments in the design and algorithms of P2P systems andapplications, and by allowing peers a greater degree ofautonomous control over their data and resourcesHowever, P2P also raises some security concerns for us-ers and accountability concerns for IT. In general it is stilla technology in development where it is hard to distin-guish useful from hype and new from old. In the rest ofthe paper we evaluate these observations in general aswell as for specific P2P systems and applications.July 10, 2002 1:01 pm1. INTRODUCTIONPaper Organization and Intended Audience2. OVERVIEW2.1. Goals2.2. Terminology2.3. P2P Taxonomies3. COMPONENTS AND ALGORITHMS3.1. Infrastructure Components3.2. Algorithms4. CHARACTERISTICS4.1. Decentralization4.2. Scalability4.3. Anonymity4.4. Self-Organization4.5. Cost of Ownership4.6. Ad-Hoc Connectivity4.7. Performance4.8. Security4.9. Transparency and Usability4.10.Fault Resilience4.11.Interoperability4.12.Summary5. CATEGORIES OF P2P SYSTEMS5.1. Historical5.2. Distributed Computing5.3. File Sharing5.4. Collaboration5.5. Platforms6. CASE STUDIES6.1. Avaki6.2. SETIhome6.3. Groove6.4. Magi6.5. FreeNet6.6. Gnutella6.7. JXTA6.8. .NET My Services6.9. Summary7. LESSONS LEARNED7.1. Strengths and Weaknesses7.2. Non-Technical Challenges7.3. Implications for Users, Developers, and IT8. SUMMARY AND FUTURE WORK8.1. Final Thoughts on What P2P Is8.2. Why We Think P2P is Important8.3. P2P in the Future8.4. SummaryACKNOWLEDGMENTSREFERENCESAPPENDIX A P2P VS. ALTERNATIVESP2P gained visibility with Napsters support for musicsharing on the Web Napster 2001 and its lawsuit withthe music companies. However, it is increasingly becom-ing an important technique in various areas, such as dis-tributed and collaborative computing both on the Weband in ad-hoc networks. P2P has received the attention ofboth industry and academia. Some big industrial efforts2include the P2P Working Group, led by many industrialpartners such as Intel, HP, Sony, and a number of startupcompanies; and JXTA, an open-source effort led by Sun.There are already a number of books published Oram2000, Barkai 2001, Miller 2001, Moore and Hebeler2001, Fattah and Fattah 2002, and a number of thesesand projects in progress at universities, such as ChordStoica et al 2001, OceanStore Kubiatowicz et al.2000, PAST Druschel and Rowstron 2001, CAN Rat-nasamy 2001, and FreeNet Clark 1999.Here are several of the definitions of P2P that are beingused by the P2P community. The Intel P2P workinggroup defines P2P as “the sharing of computer resourcesand services by direct exchange between systems”p2pwg 2001. David Anderson calls SETIhome andsimilar P2P projects that do not involve communicationas “inverted client-server”, emphasizing that the comput-ers at the edge provide power and those in the middle ofthe network are there only to coordinate them Anderson2002. Alex Weytsel of Aberdeen defines P2P as “theuse of devices on the internet periphery in a non-clientcapacity” Veytsel 2001. Clay Shirky of OReilly andAssociate uses the following definition: “P2P is a classof applications that takes advantage of resources stor-age, cycles, content, human presence available at theedges of the Internet. Because accessing these decentral-ized resources means operating in an environment of un-stable connectivity and unpredictable IP addresses, P2Pnodes must operate outside the DNS system and havesignificant or total autonomy from central servers”Shirky 2001. Finally, Kindberg defines P2P systems asthose with independent lifetimes Kindberg 2002.In our view, P2P is about sharing: giving to and obtain-ing from a peer community. A peer gives some resourcesand obtains other resources in return. In the case of Nap-ster, it was about offering music to the rest of the com-munity and getting other music in return. It could bedonating resources for a good cause, such as searchingfor extraterrestrial life or combating cancer, where thebenefit is obtaining the satisfaction of helping others.P2P is also a way of implementing systems based on thenotion of increasing the decentralization of systems, ap-plications, or simply algorithms. It is based on the prin-ciples that the world will be connected and widelydistributed and that it will not be possible or desirable toleverage everything off of centralized, administrativelymanaged infrastructures. P2P is a way to leverage vastamounts of computing power, storage, and connectivityfrom personal computers distributed around the world.Assuming that “peer” is defined as “like each other,” aP2P system then is one in which autonomous peers de-pend on other autonomous peers. Peers are autonomouseServiceseBusinessWeb appsserverclientsFigure 1: Simplified, High-Level View of Peer-to-Peer ver-sus Centralized (Client-Server) Approach.when they are not wholly controlled by each other or bythe same authority, e.g., the same user. Peers depend oneach other for getting information, computing resources,distr. appsmanaged.self-organizedAeTIserver dependencies.independent lifetimeveDNS-based.custom namingRPC.asyncclient-server peer-to-peerforwarding requests, etc. which are essential for the func-clustersad-hoc NWtioning of the system as a whole and for the benefit of allpeers. As a result of the autonomy of peers, they cannotnecessarily trust each other and rely completely on thebehavior of other peers, so issues of scale and redundan-cy become much more important than in traditional cen-tralized or distributed systems.Conceptually, P2P computing is an alternative to thecentralized and client-server models of computing,where there is typically a single or small cluster of serv-ers and many clients (see Figure 1). In its purest form, theP2P model has no concept of server; rather all partici-pants are peers. This concept is not necessarily new.Many earlier distributed systems followed a similarmodel, such as UUCP Nowitz 1978 and switched net-works Tanenbaum 1981. The term P2P is also not new.In one of its simplest forms, it refers to the communica-tion among the peers. For example, in telephony userstalk to each other directly once the connection is estab-lished, in a computer networks the computers communi-cate P2P, and in games, such as Doom, players alsointeract directly. However, a comparison between client-server and P2P computing is significantly more complexand intertwined along many dimensions. Figure 2 is anattempt to compare some aspects of these two models.WANs gridsInternet intranetdistributed systemsFigure 2: Peer-to-Peer versus Client-Server. There is noclear border between a client-server and a P2P model. Bothmodels can be built on a spectrum of levels of characteristics(e.g., manageability, configurability), functionality (e.g., look-up versus discovery), organizations (e.g., hierarchy versusmesh), components (e.g., DNS), and protocols (e.g., IP), etc.Furthermore, one model can be built on top of the other orparts of the components can be realized in one or the othermodel. Finally, both models can execute on different types ofplatforms (Internet, intranet, etc.) and both can serve as an un-derlying base for traditional and new applications. Therefore,it should not be a surprise that there is so much confusion aboutwhat P2P is and what it is not. It is extremely intertwined withexisting technologies Morgan 2002.The following three lists, which are summarized inTable 1, are an attempt to define the nature of P2P, whatis and is not new in P2P.P2P is concerned with:· The historical evolution of computing in general andthe Internet in particular; computing at the edge of theInternet. (e.g., SETIhome and other distributed com-puting systems)The P2P model is quite broad and it could be evaluatedexamplemarkets/financialbiotech. communic.enterprise entertainmentfrom different perspectives. Figure 3 categorizes thescope of P2P development and deployment. In terms ofindustriessimulationmarketgenome instantsequence messagingprocessmgmtgamesdevelopment, platforms provide an infrastructure to sup-port P2P applications. Additionally, developers are be-ginning to explore the benefit of implementing variousexampleapplicationscalculationdemogr.analysisetc.proteinfoldingetc.whiteboardsetc.onlinestoragerich mediasharingfilesharingetc.horizontal technologies such as distributed computing,collaborative, and content sharin