The Popularity Parameter in Unstructured P2P File Sharing Networks



Similar documents
IDENTIFICATION OF THE DYNAMICS OF THE GOOGLE S RANKING ALGORITHM. A. Khaki Sedigh, Mehdi Roudaki

6.7 Network analysis Introduction. References - Network analysis. Topological analysis

A New Bayesian Network Method for Computing Bottom Event's Structural Importance Degree using Jointree

Green Master based on MapReduce Cluster

A Parallel Transmission Remote Backup System

Average Price Ratios

The Analysis of Development of Insurance Contract Premiums of General Liability Insurance in the Business Insurance Risk

How To Balance Load On A Weght-Based Metadata Server Cluster

RESEARCH ON PERFORMANCE MODELING OF TRANSACTIONAL CLOUD APPLICATIONS

Applications of Support Vector Machine Based on Boolean Kernel to Spam Filtering

SHAPIRO-WILK TEST FOR NORMALITY WITH KNOWN MEAN

APPENDIX III THE ENVELOPE PROPERTY

ECONOMIC CHOICE OF OPTIMUM FEEDER CABLE CONSIDERING RISK ANALYSIS. University of Brasilia (UnB) and The Brazilian Regulatory Agency (ANEEL), Brazil

Network dimensioning for elastic traffic based on flow-level QoS

An Approach to Evaluating the Computer Network Security with Hesitant Fuzzy Information

Cyber Journals: Multidisciplinary Journals in Science and Technology, Journal of Selected Areas in Telecommunications (JSAT), January Edition, 2011

IP Network Topology Link Prediction Based on Improved Local Information Similarity Algorithm

Efficient Traceback of DoS Attacks using Small Worlds in MANET

Simple Linear Regression

Software Reliability Index Reasonable Allocation Based on UML

T = 1/freq, T = 2/freq, T = i/freq, T = n (number of cash flows = freq n) are :

Low-Cost Side Channel Remote Traffic Analysis Attack in Packet Networks

The impact of service-oriented architecture on the scheduling algorithm in cloud computing

Dynamic Provisioning Modeling for Virtualized Multi-tier Applications in Cloud Data Center

Security Analysis of RAPP: An RFID Authentication Protocol based on Permutation

The Digital Signature Scheme MQQ-SIG

10.5 Future Value and Present Value of a General Annuity Due

Maintenance Scheduling of Distribution System with Optimal Economy and Reliability

Preprocess a planar map S. Given a query point p, report the face of S containing p. Goal: O(n)-size data structure that enables O(log n) query time.

Abraham Zaks. Technion I.I.T. Haifa ISRAEL. and. University of Haifa, Haifa ISRAEL. Abstract

A DISTRIBUTED REPUTATION BROKER FRAMEWORK FOR WEB SERVICE APPLICATIONS

Credibility Premium Calculation in Motor Third-Party Liability Insurance

ADAPTATION OF SHAPIRO-WILK TEST TO THE CASE OF KNOWN MEAN

Optimal Packetization Interval for VoIP Applications Over IEEE Networks

ANALYTICAL MODEL FOR TCP FILE TRANSFERS OVER UMTS. Janne Peisa Ericsson Research Jorvas, Finland. Michael Meyer Ericsson Research, Germany

Classic Problems at a Glance using the TVM Solver

RQM: A new rate-based active queue management algorithm

AnySee: Peer-to-Peer Live Streaming

The Gompertz-Makeham distribution. Fredrik Norström. Supervisor: Yuri Belyaev

Numerical Methods with MS Excel

Chapter 3. AMORTIZATION OF LOAN. SINKING FUNDS R =

Automated Event Registration System in Corporation

A system to extract social networks based on the processing of information obtained from Internet

Discrete-Event Simulation of Network Systems Using Distributed Object Computing

Fast, Secure Encryption for Indexing in a Column-Oriented DBMS

n. We know that the sum of squares of p independent standard normal variables has a chi square distribution with p degrees of freedom.

ANOVA Notes Page 1. Analysis of Variance for a One-Way Classification of Data

An Effectiveness of Integrated Portfolio in Bancassurance

AN ALGORITHM ABOUT PARTNER SELECTION PROBLEM ON CLOUD SERVICE PROVIDER BASED ON GENETIC

Banking (Early Repayment of Housing Loans) Order,

Fractal-Structured Karatsuba`s Algorithm for Binary Field Multiplication: FK

A Novel Method in Scam Detection and Prevention using Data Mining Approaches

An Evaluation of Naïve Bayesian Anti-Spam Filtering Techniques

Chapter = 3000 ( ( 1 ) Present Value of an Annuity. Section 4 Present Value of an Annuity; Amortization

Load Balancing Algorithm based Virtual Machine Dynamic Migration Scheme for Datacenter Application with Optical Networks

Modeling of Router-based Request Redirection for Content Distribution Network

Entropy-Based Link Analysis for Mining Web Informative Structures

Agent-based modeling and simulation of multiproject

On formula to compute primes and the n th prime

DECISION MAKING WITH THE OWA OPERATOR IN SPORT MANAGEMENT

STATISTICAL PROPERTIES OF LEAST SQUARES ESTIMATORS. x, where. = y - ˆ " 1

Report 52 Fixed Maturity EUR Industrial Bond Funds

Optimal multi-degree reduction of Bézier curves with constraints of endpoints continuity

How To Make A Supply Chain System Work

Mobile Agents in Telecommunications Networks A Simulative Approach to Load Balancing

Online Appendix: Measured Aggregate Gains from International Trade

Application of Grey Relational Analysis in Computer Communication

Dynamic Two-phase Truncated Rayleigh Model for Release Date Prediction of Software

A particle swarm optimization to vehicle routing problem with fuzzy demands

Projection model for Computer Network Security Evaluation with interval-valued intuitionistic fuzzy information. Qingxiang Li

Numerical Comparisons of Quality Control Charts for Variables

The Time Value of Money

A particle Swarm Optimization-based Framework for Agile Software Effort Estimation

Research on Cloud Computing and Its Application in Big Data Processing of Railway Passenger Flow

AP Statistics 2006 Free-Response Questions Form B

of the relationship between time and the value of money.

The analysis of annuities relies on the formula for geometric sums: r k = rn+1 1 r 1. (2.1) k=0

Impact of Mobility Prediction on the Temporal Stability of MANET Clustering Algorithms *

TESTING AND SECURITY IN DISTRIBUTED ECONOMETRIC APPLICATIONS REENGINEERING VIA SOFTWARE EVOLUTION

Chapter Eight. f : R R

Models for Selecting an ERP System with Intuitionistic Trapezoidal Fuzzy Information

Compressive Sensing over Strongly Connected Digraph and Its Application in Traffic Monitoring

VIDEO REPLICA PLACEMENT STRATEGY FOR STORAGE CLOUD-BASED CDN

Group Nearest Neighbor Queries

DYNAMIC FACTOR ANALYSIS OF FINANCIAL VIABILITY OF LATVIAN SERVICE SECTOR COMPANIES

A Framework of Business Intelligence-driven Data Mining for e-business

Using Phase Swapping to Solve Load Phase Balancing by ADSCHNN in LV Distribution Network

Impact of Interference on the GPRS Multislot Link Level Performance

Dynamic Service and Data Migration in the Clouds

A Smart Machine Vision System for PCB Inspection

Questions? Ask Prof. Herz, General Classification of adsorption

Web Service Composition Optimization Based on Improved Artificial Bee Colony Algorithm

On Error Detection with Block Codes

Optimization Model in Human Resource Management for Job Allocation in ICT Project

Capacitated Production Planning and Inventory Control when Demand is Unpredictable for Most Items: The No B/C Strategy

Proactive Detection of DDoS Attacks Utilizing k-nn Classifier in an Anti-DDos Framework

Statistical Pattern Recognition (CE-725) Department of Computer Engineering Sharif University of Technology

Fault Tree Analysis of Software Reliability Allocation

A COMPARATIVE STUDY BETWEEN POLYCLASS AND MULTICLASS LANGUAGE MODELS

Integrating Production Scheduling and Maintenance: Practical Implications

Transcription:

The Popularty Parameter Ustructured P2P Fle Sharg Networks JAIME LLORET, JUAN R. DIAZ, JOSE M. JIMÉNEZ, MANUEL ESTEVE Departmet of Commucatos Polytechc Uversty of Valeca Camo de Vera s/, 4622 Valeca SPAIN jlloret@dcom.upv.es, juadasa@doctor.upv.es, jojher@masters.upv.es, mesteve@dcom.upv.es Abstract: - Sce P2P became extremely popular betwee Iteret users, may researchers have tred to model those P2P etworks. Oe of the parameters, used these models, s the popularty of a fle. Some artcles demostrate that, f a fle s so popular, the probablty to fd ths fle sde the P2P fle sharg etwork s bgger. Ths artcle deals wth popularty parameter P2P fle sharg etworks. I order to do so, the ustructured publc doma Peer-to-Peer etworks Gutella,,,, Soulseek ad have bee measured. The authors have establshed a relatoshp betwee some flms, sogs, programs ad documets foud web search eges ad the same fles foud publc doma P2P fle sharg etworks. If all these aalyzed Peer-To-Peer fle-sharg etworks were tercoected, the probablty to fd a desred fle wll be cremeted. O the other had, those aalyzed P2P etworks seems to be specalzed dfferet type of fles as t s show the paper. Key-Words: - Peer to peer, Fle Popularty, Fle Search, Peer-To-Peer Itercoecto. Itroducto Sce Iteret became accessble to the world, oe of the frst users cocers s to fd the fle or the formato s lookg for. A measuremet study [] of the deep Web reveals that t cotas early 55 bllo of pages ad t s doublg each year. O the other had, the surface Web cotas a estmated 2.5 bllo documets, growg at a rate of 7.5 mllo documets per day, ad the deep Web s approxmately 5 tmes greater tha that vsble to covetoal search eges. Nowadays there are a lot of web search eges [2] ad a lot of them have bllos of textual documets dexed [3]. The Web search eges ca be classfed three types: - Crawler-Based Search Eges, such as Google, whch create ther lstgs automatcally. They "crawl" or "spder" documets by followg oe hypertext lk to aother, the people search through what they have foud. - Huma-Powered Drectores, such as Ope Drectory. It depeds o humas for ts lstgs. People have to submt a short descrpto to the drectory for ther etre ste. A search looks for matches oly the descrptos submtted. - Hybrd Search Eges, such as MSN search. It s mataed by a combato of prevous types ad preset both results. The Web search eges employ some kd of cetralzed algorthm. I order to have the best results, these algorthms use locato/frequecy method (search eges check to see f the search keywords appear ear the top of a web page ad how ofte keywords appear relato to other words a web page) ad the off-the-page factor (lke clckthrough measuremet). They are the major factor how search eges determe the popularty of a documet. Habtually, search results are sorted popularty order. Curretly there are a lot of P2P fle-sharg etworks exstece, ad may of them have mllos of o-le users ad mllos of data shared [4]. I ths type of etworks, what a user really wats s to fd the fle s lookg for to dowload t. The probablty to fd a desred fle, the etwork where a user s searchg, s assocated to the popularty of the fle. Some other parameters lke the type of fle t ca be shared, the avalablty of the fle ad ts replcato are also cosdered. I order to have real search measuremets about some flms, sogs, programs ad documets, we have selected some of the most popular publc doma P2P fle-sharg etworks. Those selected etworks are Gutella [5], [6], Opeap [7], Edokey [8], Soulseek [9] ad []. Although there are other etworks [], we have selected ths oes because they are so popular betwee Iteret users.

O the other had, we have selected two crawlerbased search eges, Google ad Altavsta, ad oe search drectory, Yahoo!, order to fd the same fles Web search eges. Later o, t s establshed a relatoshp betwee the results obtaed web search eges ad the results obtaed the peer-to-peer fle-sharg etworks aforemetoed. It wll gve us the popularty of those fles. Ths paper s structured as follows. Secto 2 dscusses the search techques used Peer-To- Peer fle-sharg etworks. I secto 3, t s descrbed the popularty parameter. Secto 4 shows the measuremets take the Peer-To-Peer flesharg etworks ad Web search eges selected. It s also show the relatoshp betwee them. I secto 5, t s dscussed how ca be creased the probablty to fd a desred fle Peer-To-Peer fle-sharg etworks. Fally, Secto 6, there are coclusos ad future works. 2 Search Techques Peer to Peer Fle Sharg Networks I order to fd a fle a P2P etwork, a search s eeded. The mplemeted search algorthm every etwork depeds o the type of the etwork (cetralzed P2P, decetralzed P2P ad partally cetralzed). There are several types of searchg algorthms ad they ca be classfed as follows: 2. Loosely cotrolled P2P search algorthms. They are used decetralzed Peer-To-Peer etworks. The data placemet s ot defed because the odes of the etwork decde what fles they wat to share. There are two kd of loosely cotrolled P2P search algorthms: 2.. Broadcast search techque The query search s set to all drectly coected eghbours ad they forward the query to all ther eghbours. The query s propagated to suffcet umber of odes to match the etry or utl a TTL value. If the eghbour has the cotet, t reples, otherwse f floods the query to ts eghbours. Ths type of search s used by Gutella etwork. 2..2 Selectve search techque The query search s set to some odes called superodes that act as a cetral odes. Ths superodes wll perform the search to other superodes order to fd the requested fle. The clets wth a hgher badwdth ad process capacty wll be cosdered automatcally superodes. Those clets wth less badwdth wll be superode clets. Ths type of system uses a flow cotrol algorthm for sedg queres ad reples. It also has a dagram of prortes used to dscard some messages. Ths type of search s used by ad Gutella 2 [2]. 2..3 Radomly search techque The query s set to k umber of radomly selected eghbours. Each of these eghbours forward the query to ay of ther radomly selected eghbours. The query s propagated to suffcet umber of odes to match the etry or utl a TTL value. Ths techque s descrbed [3]. 2..4 Probably search techque I ths case, the queres are set to specfc clets whch are cosdered to have the greater probablty fdg the request. Each ode matas a probablty value correspodg to each eghbour whch defes the chaces that a query wll be forwarded to that eghbour. A example of ths type of search s APS [4]. 2.2 Strogly cotrolled P2P search algorthms. I structured P2P etworks, data placemet ad topology wth the P2P fle-sharg etwork s tghtly cotrolled. These etworks are based Dstrbutes Hash Tables (DHT), ad the odes do ot decde what they store ad share wth other peers the etwork. The data placemet s defed by the algorthm. Whe a documet s publshed, t s routed to the clet whose ID s the most smlar to the documet s ID. I order to fd a fle, the queres are set to the clet whose ID s the most smlar to the documet s ID. The process s repeated utl a close match s foud. The ma search problem ths type of etworks s that they are ot very effcet for keyword based search. Ths type of search s used by Freeet [5], CAN [6], Chord [7], Pastry [8] ad Tapestry[9]. 2.3 Server-cetrally cotrolled P2P search algorthms. They are used peer-to-peer etworks where there s a server or a group of servers. Ths type of search s very smple ad has short query tme. There are two kd of server-cetrally cotrolled P2P search algorthms:

2.3. Sgle-Server search techque Itally, P2P clets coect to a cetral server where they publsh ther shared fles (the fles ames, ther szes, etc). Whe a search query s set to the server, t looks up ts dex database. If there s a matchg etry, the IP address of the ode that shares the fle s set to the oe that requested t, ad the, the drect coecto ad dowload takes place. Ths techque s used by the Soulseek etwork. 2.3.2 Farm-of-Servers search techque I ths type of P2P etworks, there s a group of avalable servers called brokers. P2P clets must be authetcated to oe of those cetral servers. Each broker has the dexes of the local clets ad some cases the dexes of some fles from eghbour brokers. Whe a clet performs a query to a broker, ths oe searches ts local database ad f t does t fd a match, t uses the local dex order to fd a eghbour broker that ca sed the request. The server dexes are ot statc ad t ca chage accordg to the fles the system. The etworks, ad use ths techque. 3 The Popularty Parameter There are dfferet ways to measure the popularty parameter, t depeds o where ths parameter s eeded. The followg are some ways to measure the popularty parameter: - I Web search eges, as such Google, t s gve a lot of mportace to the umber of webstes that lk to a webste, so the popularty parameter s measured by the umber of comg lks to the ste. Wth ths popularty parameter t s bult the PageRak. - The popularty of a fle ca be related to the umber of tmes the fle has bee retreved from the surveyed archve durg a certa perod of tme. It s used web servers. - The popularty of a fle ca be determed by the umber of users that have requested ts dowload. It s used to measure how may users use a certa software. - The popularty parameter moves ca be related wth the audece t have had cemas, the umber of DVD sold or the umber of reted moves by vdeoclubs. - The popularty parameter sogs ca be related wth rado or web top lsts. - I structured peer-to-peer etworks, the popularty of a fle or a servce s measured by the umber of tmes the fle s requested. It also affects to the probablty that t s replcated by other peers. The popularty of a fle govers how log t stays the etwork ad how ofte t s replcated. I peer-to-peer fle-sharg etworks, the popularty ca be mathematcally expressed as follows. Objects a peer-to-peer fle-sharg etworks do ot have the same popularty. Assumg that there are m fles of terest oe P2P etwork ad q represets ther ormalzed relatve popularty (umber of queres ssued for t), t s verfed: m = q = () All selected etworks are ustructured publc doma peer-to-peer etworks. So, there s o cotrol over etwork topology or data placemet ay of them. Some of those etworks are Zpf-lke dstrbutos, as such Gutella ad Napster [2][2]. m α = q = (2) α Where α s a Zpf coeffcet ad s the -th most popular fle. Ths Zpf dstrbuto ca be further used to determe the probablty for a query to be assocated wth the -th most requested fle q. Other studes demostrates that ad other P2P fle-sharg systems are o-zpf behavor [22][23]. Assumg that each fle s replcated o r odes, the total umber of terestg fles stored the etwork s R: R = m r = (3) We ca assume that the most popular fle s also the most replcated. Aalyzed P2P fle-sharg etworks have rgd assumptos o how replcatos of objects happe the system. Oly odes that request a fle makes copes of the fle. O the other had, some etworks, as such Gutella (decetralzed) ad (partally decetralzed), search cossts of radomly probg stes utl the desred fle s foud. Thus, the probablty to fd a fle Pr(k) o the k th probe s gve by [24]: k r = r Pr ( k ) (4) Where s the umber of odes the etwork.

Gutella SoulSeek Average umber of users 8 * 3.467.98 256.3.428.75 244.48 898 Average umber of shared fles 55.54 * 63.678.68 58.92.78 3.469.627 59.756.764 /t Average sze of total shared data,294 GB* 4.947.26 GB 5.49.326 GB /t 236.564 GB /t Max. Varato of users (%) 4,49 2,33 42,2 39,3 5,5,7 Max. Varato of shared fles (%) 26,35 8,63 53,65 36,76 5,47 /t Max. Varato of shared data (%) 349,49 5,72 34,58 /t 5,79 /t Table. Comparatve of the 6 archtectures measured (* o total etwork values, /t: measured ot take) 5% % % 83% 5% % % 2% % 37% 37% 59% % % 2% 83% 59% 2% Fgure : Moves percetage Fgure 2: Sogs percetage % % 8% 9% 9% 8% % % % 4% 95% 95% % 4% Fgure 3: Software percetage Fgure 4: Documets percetage 4 Search results We have measured the average umber ad the maxmum varato of peers, shared fles ad total amout of data shared of the sx selected etworks order to kow how may peers ad formato are sde the selected etworks. Those data are show Table. It has to be take to accout that Gutella,, Opeap, Edokey, ad Soulseek etworks permt to search every type of fle, but oly permts audo fles (mp3, ogg, wma, etc.). I order to establsh whch etwork s more probably to fd moves, sogs, software ad documets, we have measured the umber of peers that have fles wth keywords of 2 moves, 24 sogs, 2 software programs ad 8 documets. Ths measures have bee take every oe of the sx etworks. To avod wrog results our searches we have employed ext methodology: - To lmt the results to the type of fle we were lookg for, the type of fle was added (av, mpg, exe, pdf, doc, etc.) to they keywords of the search. - The umber of software versos were ot cluded the keyword of the search. - The ame of moves that have secod delveres (shreck 2, Spderma 2 ad so o) were ot cluded. Although there are peers tercoected to more tha oe peer-to-peer etwork, they are sgfcat compared wth the total umber of peers of every etwork. O the other had, the fles shared by those peers could ot be the same for all coected etworks. If all aalyzed etworks were tercoected, the type of fles shared by every etwork to the other are dfferet. Although all of them supports every type of etwork (except ), those etworks seems to be specalzed dfferet type of fles, as t s show Fgures to 4. The table 2 shows whch s the rakg for those sx etworks I order to kow the popularty of those moves, sogs, software ad documets, results have bee compared wth two search eges, Google ad Altavsta, ad oe search drectory, Yahoo!.

5 4 3 2 25 2 5 5 altavsta 3 25 2 altavsta 5 5 2 3 4 5 6 7 8 9 2 Selected Fles 3 5 7 9 3 5 7 9 2 23 Selected Fles Fgure 5: Number of results of Moves Fgure 6: Number of results of Sogs altavsta 4 2 8 6 4 2 2 3 4 5 6 7 8 9 2 Selected Fles altavsta 2 3 4 5 6 7 Selected Fles Fgure 7: Number of results of Software Fgure 8: Number of results of Documets P2P Network Moves Sogs Software Documets st 2 d st st 2 d st 2 d 3 rd 4 th 4 th 3 rd 2 d 3 rd 3 rd 4 th 4 th Soulseek 5 th 6 th 5 th 5 th X 5 th X X Table 2. Rakg moves, sogs, software ad documets To avod wrog results web searches ad to lmt the results to the type of fle we are lookg for, we have employed ext methodology: - For move searches we have added the word move to the search. - For sog searches, we have added ot oly the ame of the sog, but the ame of the group. - For documets searches we have added the type of the documet (pdf, doc, etc.). - For software searches we have searched oly specfc maufacturers software. I order to aalyze the P2P data collected ad compare t wth Web search eges we have scaled ad data results wth a x5 factor ad, ad wth x factor. These factors have bee used moves, sogs ad software. Fgures 5 to 8 shows the data collected. Ths measures have bee take oly for comparatve popularty purposes, t s preteded to kow f a popular fle Web search eges wll gve a popular fle peer-to-peer flesharg etworks. As t ca be see those fgures, there s o result of moves, software ad documets for due to ts oly-sogs mplemetato. 5 Icreasg the Probablty to Fd the Desred Fle The fles are ot always the peer-to-peer etwork where the user s searchg. Most of the etworks mplemeted owadays support ay fletype, but there are some that oly supports audo fles. What s eeded s a system whch wll allow to search every P2P etwork ad dowload from every peer of every etwork. The uo of Peer-To-Peer fle-sharg etworks, by the creato of a Peer-To-Peer fle-sharg etworks Itercoecto System, wll gve greater probablty to fd the desred fle. If there s Peer- To-Peer fle-sharg etworks, the total probablty wll be: P Eα = + P α P P ( ) Pα Pβ... Pη +... + ( ) PP 2... P η>... > β> α β β> + P P P +... α β γ γ> β> (5) As t ca be see, the total probablty to fd the desred fle wll be greater tha the probablty of oe them oly, but less tha the sum of all of them. 6 Cocluso Sx ustructured publc doma Peer-to-Peer etworks have bee measured order to kow what s the P2P fle sharg etwork wth most search results moves, sogs, software ad documets. As a result of our measuremets, those etworks seems to be specalzed dfferet type of fles.

The search results have bee compared wth those obtaed Web search eges. We have checked that f a fle s popular Web search eges, t s also popular P2P fle-sharg etworks. Future works wll try to fd the mathematcal relatoshp betwee the fle popularty Web searches eges ad the fle popularty P2P fle-sharg etworks. ackowledgemets Authors wat to ackowledge to Mr. Mguel A. Graados from Polytechc School of Gada for hs data collecto. Refereces: [] Mchael K. Bergma, The Deep Web: Surfacg Hdde Value, The Joural of Electroc Publshg, Volume 7, Issue. August, 2. [2] Day Sullva, Nelse NetRatgs Search Ege Ratgs, July 4, 24. Avalable at: http://searchegewatch.com/reports/artcle.ph p/25645 [3] Day Sullva, Search Ege Szes, September 2, 23. Avalable at: http://searchegewatch.com/reports/artcle.ph p/25648 [4] J. Lloret Maur, B. Mola Moreo, C. Palau Salvador y M. Esteve Domgo. Publc Peer- To-Peer Flesharg Networks Evaluato. The 2d Iasted Iteratoal Coferece O Commucato Ad Computer Networks. MIT Cambrdge, MA, USA. November 24. [5] Eyta Adar ad Berardo Huberma. Free rdg o gutella. Frst Moday, 5(), October 2. [6] Nathael Lebowtz, Mate Rpeau, ad Adam Werzbck. Decostructg the Kazaa Network, 3rd IEEE Workshop o Iteret Applcatos, Sa Jose, USA Jue 23. [7] http://opeap.sourceforge.et/ [8] Olver Heckma ad Axel Bock. The 2 Protocol. Techcal Report KOM-TR-8-22, Multm. Commucatos Lab, Darmstadt Uversty of Techology, December 22. [9] Soulseek http://www.slsk.org [] http://www.blubster.com/protocol.html [] Wkpeda http://www.wkpeda.org/wk/peer-to-peer [2] http://www.gutella2.com [3] Chrstos Gkatsds, Mlea Mhal, ad Am Saber, Radom Walks Peer-to-Peer Networks, The 23rd Coferece of the IEEE Commucatos Socety (Ifocom 24), Hog Kog, March 24 [4] D. Tsoumakos ad N. Roussopoulos: Adaptve Probablstc Search for Peer-to-Peer Networks. I Proceedgs of the 3rd IEEE Iteratoal Coferece o P2P Computg, Lkopg, Swede, September 23. [5] I. Clarke et al. Freeet: A dstrbuted aoymous formato storage ad retreval system, ICSI Workshop o Desg Issues Aoymty ad Uobservablty, It'l Computer Scece Ist., 2. [6] S. Ratasamy, P. Fracs, M. Hadley, R. Karp, S. Sheker, A Scalable Cotet- Adressable Network, ACM Sgcomm 2, Sa Dego, CA, USA, August 2, [7] I. Stoca, R. Morrs, D.Karger, F.Kaashoek, H. Balakrsha, Chord: A Scalable Peer-To-Peer Lookup Servce for Iteret Applcatos, ACM Sgcomm 2, Sa Dego, USA, August 2, [8] A. Rowstro ad P. Druschel, Pastry: Scalable, dstrbuted object locato ad routg for large-scale peer-to-peer systems, IFIP/ACM Iteratoal Coferece o Dstrbuted Systems Platforms (Mddleware), hedelberg, Germay, pages 329-35, November, 2 [9] B. Zhou, D.A. Joseph, J. Kubatowcz, Tapestry: a fault tolerat wde area etwork fraestructure, UC Berkeley techcal report UCB/CSD--4 [2] Kuwadee Srpadkulcha, the popularty of gutella queres ad ts mplcatos o scalablty. I O Relly s www.opep2p.com, February 2 [2] Zhu Ge, Dael R. Fgueredo, Sharad Jaswal, Jm Kurose, Do Towsley. Modelg Peer-Peer Fle Sharg Systems, Proceedgs IEEE INFOCOM 23, Sa Fracsco, March- Aprl 23. [22] Krsha P. Gummad, Rchard J. Du, Stefa Sarou, Steve D. Grbble, Hery M. Levy, Joh Zahorja, Measuremet, Modelg, ad Aalyss of a Peer-to-Peer Fle-Sharg Workload, Proceedgs of the eteeth ACM symposum o Operatg systems prcples, 23, p. 34-329. [23] J. Chu, K. Labote, ad B. N. Leve. Avalablty ad localty measuremets of peerto-peer fle-sharg systems. I Proceedgs of SPIE ITCom: Scalablty ad Traffc Cotrol IP Networks, volume 4868, July 22. [24] Q Lv, Pe Cao, Edth Cohe, Ka L, ad Scott Sheker, Search ad replcato ustructured peer-to- peer etworks, Proceedgs of the 6th teratoal coferece o Supercomputg, ACM Press, 22, p. 84 95.