DYNAMIC TOPOLOGY ALGORITHM FOR P2P NETWORKS

Size: px

Start display at page:

Download "DYNAMIC TOPOLOGY ALGORITHM FOR P2P NETWORKS"

Gabriel Chase
5 years ago
Views:

1 Joural of Theorecal ad Appled Iformao Techology 30 h November 01. Vol. 5 No JATIT & LLS. All rghs reserved. ISSN: E-ISSN: DYNAMIC TOPOLOGY ALGORITHM FOR PP NETWORKS 1 YUNXIA PEI, JUANJUAN LUO 1 Deparme of Mah ad Compuer Scece, Zhegzhou Uversy of lgh Idusry, Zhegzhou 5000, Hea, Cha Deparme of Ecoomc Trade Maageme, Hua Ralway Professoal Techology College, Zhuzhou 01, Hua, Cha ABSTRACT I order o mprove he Peer-o-Peer ewor resource allocao effcecy, ad o solae he malcous odes, hs paper has preseed a ew mehod amed dyamc opology algorhm (DTA), suable for usrucured PP ewor. As he eghbor ode ses varable me-o-lve (TTL) values, ad he eraco accordg o he eghbor ode forwardg o he TTL value s adused; records of he query pah drecly o a forwardg ode coeco; gve he forwardg formao placg mehod ad opologcal adusme algorhm. The smulao expermes ad aalyss show ha, he DTA opology cosruco mehod ca be malcous odes solaed o he ewor edge, mproves he effcecy of resource locag, effecvely mprove he qualy of servce PP ewor. Keywords: Peer-o-Peer, Usrucured, Topology, Resource Locag 1 INTRODUCTION I rece years, he applcao of Peer-o-Peer (PP) ewor creases rapdly, he PP sysem allows a large umber of cyber sources o be shared, such as fle, calculao ably ad so o. The PP sysem s esseally a d of dsrbued sysem, o ceral server. The PP sysem he ode has ercoeced o form a self-orgazg overlappg ewor [1]. PP s locaed he uderlyg physcal ewor opology ad form a vrual overlay layer, for all ypes of PP applcao suppor. The PP ewor opology ad PP sysem operao are closely relaed, he fluece of PP ewor servce qualy. Some research shows ha he PP ewor opology opmzao ca mprove he resource posog effcecy, reduce he expese of he sysem, so as o mprove he servce qualy PP ewor. Accordg o he opologcal srucure of dffere PP sysem s maly dvded o wo caegores: usrucured PP sysems ad srucured PP sysem. A prese, mos of he popular PP applcao s based o usrucured PP sysems. Usrucured PP sysems o radom opologcal srucure based o floodg ad query, resulg a large umber of reduda ewor query message, affecs he effcecy of resource locag []. I usrucured PP ewor, a "good" opology ca be more effecve for locag resources, cludg he possbly of fdg arge servce ode, replcao of he query message umber, solae he malcous odes ec. The leraure [3] a all agles of opology opmzao echology, bu dd o cosder ewor malcous ode problem. The leraure [] ode relably adused accordg o opology, bu o he eghbor ode capacy calculao mehod s oe-sded, loss of opology adusme faress. I order o mprove he effcecy of resource locag, ad solae he malcous odes, hs paper preses a suable for usrucured PP ewor dyamc opology cosruco mehod DTA, o he eghbor ode ses varable TTL value, ad he eraco of TTL value s adused, also recorded he query pah drecly o a forwardg ode coecos, ad are based o hese formao o adus he opology. The smulao expermes ad aalyss show ha he opology cosruco mehod DTA s effecve. I ca be malcous ode solao o he edge of he ewor, mprove he servce qualy. THE RELATED WORK Fas Trac[5] s based o he archecure of PP proocol. The proocol has seleced hgh performace of ode as he super ode o oher odes, as a commo ode. The roduco of he poro relave ably of he ode as he super ode bearg more compug wor ad order o accelerae he speed of query roug. The leraure

2 Joural of Theorecal ad Appled Iformao Techology 30 h November 01. Vol. 5 No JATIT & LLS. All rghs reserved. ISSN: E-ISSN: [] s based o ode all aspecs of he processg ably of opology opmzao mehod, respecvely, accordg o he ode degree, roug or mulple measurg sadard comprehesve o selec eghbor ode. The exsg PP ewor opology srucure ad he uderlyg physcal ewor opology ofe prese he opology msmachg problem, he leraure [7] for he opology msmach problem, pu forward a d of poso relaed opology machg echology ad srucural opology algorhm, whch selecs he physcal locao s closer o he ode as a eghbor, bu hs mehod requres all ode cloc s sychrozed wh he. Adapve Coeco Esablshme (ACE) for each ode o s eghbors by esablshg he mmum spag ree, opmzao of eghbor odes, ca dyamcally coec o a physcal locao closer o he ode. The mehod solves he opology msmach problem, ad he decrease of PP ewor raffc. The PP ewor odes he coe of eres ofe show a cera cocerao, so some research o peer eres srucure opology. The leraure [] s how o use he user's eres o form dffere clusers. Ths paper preses a dyamc adapve algorhm ad loo-up mehod. The leraure [9] he exsg ewor opology of he above, proposed oe d based o he ode of eres shorcu algorhm, odes ag he query, frs se o he shorcu search, ul f shorcus s o respodg o, wll coue he PP ewor opology seds a query, bu from he shorcu of respose s o ecessarly he mos relable respose. The leraure [] provdes a PP ewor wh a smlar eres he odes form a "cluser mehod, ca reduce he search rage query reques, a shor dsace o ge respose. The opology opmzao mehod does o cosder he problem of malcous odes. Based o a d of ode relably opology opmzao mehod based o he ode o provde he servce of cofdece degree adusme opology, odes are more lely o choose hgh rus value ode as eghbors, o provde relable servce qualy of malcous odes have smaller chace of beg seleced as he eghbor. The leraure [11] proposed credbly sesve opology evoluo algorhm, based o he ode rus value deermes he opology herarchy. APT for a class of hdde malcous odes proposed opology cosruco proocol, bu he query oly record he frs forward he query o he eghbor ode, ad accordg o he fal resuls reured for he frs eghbor ode s dsposed, s ufar, wll lead o he ormal ode. RC-ATP s also he record correspodg o he 9 respose of he frs forwardg of eghbor odes, dffere odes from her ow eracve experece o calculae he oher ode hrough muual recprocal capacy, ad ably o measure oher odes, ad o hs bass o selec eghbor odes, bu abou he eghbor ode forwardg capacy calculao mehod of oe-sded. Leraure [1] based o prof sharg model opology cosruco, adusme of each ode he measure of usg he servce relably ad credbly are weghed ad roug, ode he calculao of each subseque eghbor ode roug relably, egraed he ode follow-up eghbor ode from he ode hrough he forward all he query ge a share of he profs, compuaoal overhead s large, addo, deermg he source ode based o he feedbac credbly whe publc eracve ode algorhm, a large-scale ewor radg sparse cases are dffcul o deerme, hus gvg roug relably udgme brough dffculy. Pu forward a d of Guella sysem PP ewor fle sharg proocol Adapve Guella-le Proocol (AGP), hrough he ode for servce corbuo assessme of ode relably, accordg o he credbly ad share coe smlary of odes he ewor wll form a "cluser", collaborao of good ode ad credbly of he odes become opology core. 3 DTA RESOURCE LOCATOR PP ewors ca be expressed as a udreced graph ame G = ( P, E), where P s he se of odes, E s he edge se, (, ) sad ode ad ode coeco o,, P. PP ewor coeco s symmercal, ca be used o descrbe a eghbor ode se, he eghbor ode s bdrecoal, ad ode eghbor odes ca be expressed as N( ) = { (, ) E}. Frsly s roduced he DTA basc resource locao mehod. Nodes ca be smlar o Guella ewor mechasms for processg odes o ad query,.e. ode wh some currely a he odes of he ewor coeco s esablshed o ge coeced o he Iere, ad hrough he floodg way o delver he ewor query message, ad hrough TTL o corol query sze, TTL o express he message he ewor he rasmsso frequecy, each ode he query message o he eghbor ode rasfer before he TTL mus 1, whe TTL s zero, o loger passes he query message. I PP fle sharg ewor as a example, he resource locao mehod s descrbed as follows:

3 Joural of Theorecal ad Appled Iformao Techology 30 h November 01. Vol. 5 No JATIT & LLS. All rghs reserved. ISSN: E-ISSN: Fgure 1: Resource Locao Sech Map I Fgure 1, he ode P1 as he requesg ode, a query message o he eghbor ode, eghbor ode P, P3 ad P respecvely he query message o her respecve eghbor odes, ul TTL 0. Node P1 receved a seres of respose, ad from he respose ls a mechasm o choose he respose ode P ad seds reques message, wa for he ode P o provde servces, respose o he seleco mechasm deermes he ode prory wh whch ode o rade. The ode P1 resource locag process, query pah for P1-P3-P7-P, P3 ad P7 for forwardg ode, P3 for he requesg ode P1 eghbor ode, P7 ow as correspod o he servce ode P drec forwardg ode. Query pah geerally have mulple forwardg ode, bu compared o he oher ode, drec forwardg ode he forwardg pah plays a crucal role DTA, herefore, addo o record query pah o he eghbor ode forwardg resuls, also recorded he forwardg ode coeco, ad based o hese formao o adus he opology. FORWARDING INFORMATION POSITION I he DTA opology cosruco mehod PP ewor, each ode correspods o a eghbor ode ls ad a coecg facor ls, as show Table 1 ad Table shows. Neghbor ode ls recorded each eghbor odes ad he correspodg ID rused TTL value; coecg facor ls records correspodg o he servce ode forwardg ode coeco facor, represeg a drec forwardg ode coeco servce ode capacy, reflecg he drec forwardg ode coeco suao. Node eghbor ls as show Table 1, sad he eghbor ode correspodg o he ID; TTL for ode, he eghbor ode correspodg rused TTL value. Node coeco facor ls as show Table, sad ID f sad forwardg ode; drec forwardg ode coeco facor; Succ expressed hrough he ode drecly forward, he ode success respose mes; Fal expressed hrough he ode forwarded drecly, ode ge he falure respose mes. Where he coecg facor s defed as follows: Defo 1: For he ode, ode coeco facor ( Succ + Fal ) f + TraSa f = Succ + Fal +1 (1) I he formula (1), f sad ode o drec forwardg ode prevous coeco he comprehesve evaluao; TraSa forwardg ode for he curre eracve corbuo, he curre eraco s successful, for eracve sasfaco of 1/ or Trasm Peer ID 1 Table 1: Neghbor Ls Neghbor ID 1 Trused TTL TTL1 TTL TTL Table : Coeco Facor Ls Coeco facor f 1 f Success Tme Succ1 Succ Falure Tme Fal1 Fal f Succ 5 TOPOLOGY ADJUSTMENT ALGORITHM Fal Wh a group of frs opology cosruco mehod DTA wll use several parameer defo. Nb () : Node eghbor ode se; Nb : Node ca be coeced up o he eghbor umber; M (S) : Se S accumulaed rus he lowes ode; Max (S) : Se S he ode wh he hghes cumulave rus; 70

4 Joural of Theorecal ad Appled Iformao Techology 30 h November 01. Vol. 5 No JATIT & LLS. All rghs reserved. ISSN: E-ISSN: FMax (S) : Se S he ode wh he hghes servce ode has he rasmsso frequecy) o cumulave coeco facor; updae he query pah correspods o he eghbor Num (S) : Se S he umber of odes ; ode s he credbly of TTL = 1 Hops ; 3) If TTL TTL (, ) : Node eghbor ode, a rused TTL s 0, or a servce ode for he requesg ode eghbor ode, ode broe ad value for he eghbor ode se; ode coeco, ad he coeced wh a ew Local () : For he ode o provde he servce eghbor ode. odes colleco; ) If TTL s 1, dcag a reques o he Tra () : For he ode o provde query forwardg ad go drecly o he respose se of odes, ode record query pah correspodg o he servce ode drecly forwardg ode, amely he query o he servce ode ul afer he las ode; d : The servg ode drec forwardg ode; SumTrus (, ) : appears ode, ode good cumulave rus value; SumTra (, ) : appears ode, ode good cumulave coeco facor f *( Succ Fal ). Relaed prmves are as follows: Dscoec(, ): Node ad ode coeco; AddNeghbor(, ): Node ad ode o crease muual eghbor; RequesCoec(, ): Node reques ad as he eghbor ode, f he reques s acceped, he reur of rue, false oherwse; CoecToRadomPeers(): Node he ewor wh radom coecos a eghbor ode; The ode a each eraco has compleed, eed o updae he local servce ode formao, formao of odes ad coecg facor formao recommedao. Whe reachg he opology adusme me, accordg o he degree of sasfaco wh he servce ode, he mplemeao of dffere adusme procedure. 5.1 Usasfed Wh Servce Of Topologcal Adusme If ode for servce ode o sasfed wh he servces provded, eed o adus he eghbor ode ad s TTL value, adus he algorhm s as follows: 1) The requesg ode updae query pah drecly o a forwardg ode coeco facor; ) Accordg o he servce reques ode correspodg o he ode hops Hops ( Hops sad ode seds he query message arrves a he eghbor ode a rused TTL s 1, he he eghbor ode wll o loger have he forwardg fuco, oly as a respose o he ode, so he reques ode eed o deerme lef he eghbor ode s worh. If he eghbor odes' rus values ad cumulave rus value sasfes he codo, ad he eghbor ode TTL s equal o 1 of he eghbor ode umber suffce eghbor ode oal 1/, ca leave he eghbor ode; f o less ha half of he oal umber of eghbor odes, are avalable he eghborg odes o selec a replaceme for comparso, he eghbor ode's cumulave rus value ad he res of he TTL value of 1 eghbor ode he lowes cumulave rus value, f hgher ha he exsg eghbor mmum cumulave rus value s broe, he cumulave rus value lowes eghbor ode; oherwse brea he eghbor ode, ad he recoec a ew eghbor ode. Specfed DTA TTL 1 eghbor ode umber o more ha half of he oal umber of eghbor odes, s o esure ha he query prolferao ad respose umber. 5) Oher, he updaed TTL value a leas wo level forwardg, operao s o coduced, reur; Usrucured PP sysem based o floodg query dffuso hrough TTL o corol scale, accordg o he eghbor ode forwardg, for each eghbor ode assgs dffere rused TTL value, o oly o esure ha he query dffuso, bu also reduce he ewor query message flow. The coeco of a ew eghbor ode mehod s as follows: local records hsory formao servce ode selecs rused servce ode, ad he coecg facor o egave forwardg record, accordg o her cumulave rus value from hgh o low oe by oe he order reques. If he coeco requess are o acceped, he loo a he coecg facor ls ad o he local servce ode formao he ode, selec relable forwardg ode, accordg o he accumulaed coecg facor from hgh o low o sed he coeco reques. If sll o fd s a radomly chose ode s coeced o he coeco reques s acceped. 71

5 Joural of Theorecal ad Appled Iformao Techology 30 h November 01. Vol. 5 No JATIT & LLS. All rghs reserved. ISSN: E-ISSN: Servce Sasfaco Of Topologcal SIMULATION AND RESULTS ANALYSIS Adusme The requesg ode frs updae query pah drecly o a forwardg ode coecg facor, he chec he coeco facor ls servce ode coecg facor s wheher he egave forwardg record, f o egave forwardg record, s ryg o he servce ode for he eghbor. 1) If o hs eghbor, ad whe he umber of eghbor odes s less ha Nb : frs ry o pu he servce ode as eghbors, f he coeco reques s receved, he ode as eghbors; or ry o coec aga he forwardg ode d, he ode wll eed d o mee wo codos: o egave servce record; o egave rasm records. ) If s o a eghbor, ad whe he eghbor ode umber s o less ha Nb, he from he eghbor ls o fd a cumulave he rus value lowes eghbor ode M ( Nb( )) o prepare replaceme. If he eghbor ode M ( Nb( )) cumulave rus value less ha he sasfaco hreshold or less ha he cumulave rus value of he servce ode, ry o reques he coeco servce ode, f he coeco reques s acceped, he cu off a eghbor ode M ( Nb( )) ; f he coeco reques s o acceped, ry o coec drecly correspods o he respose forwardg ode, he ode d d also eed o mee he wo codos: o egave record of servce; o egave forwardg record. 5.3 Recevg Coeco Reques Receve coeco requess ode requres udgme, ad he decde wheher o accep he coeco reques. 1) Node local servce ode formao recordg shee ad coecg facor ls, f here are egave servce record or egave ode ae coeco requess forwardg record, do o accep he coeco reques of ode ; ) If here s o egave record ode Vew he umber of eghbors, eghbors umber o reached he mum Nb, he accep he coeco reques; eghbors reached he mum umber Nb, he exsg eghbors loog for he lowes rus value ode he replaceme, f he value s less ha he rus value of ode rus he eghbors, o accep he coeco reques. 7.1 Smulao Evrome I hs paper, a query cycle model smulao, ad cosruc a PP fle-sharg ewors. Each smulao cosss of a umber of smulao cycles, each smulao cycle, a ode he ewor ca ae he fle query ad respod o queres, he query reques broadcas maer smlar Guella's hrough he TTL corol he sze of he query. Node receves a query message o see f s respose, aed query ode was o receve a respose from he respose ls o selec he hghes rus value ode dowload ul he rasaco s successful or all resposes were usuccessful. I he smulao, he odes he ewor are dvded o wo caegores: ormal odes ad malcous odes; ormal ode o provde relable servce, malcous odes provde relable servce.. DTA Mehods To Assess I hs paper, we have used he characersc pah legh o assess DTA mehod. Node average characersc pah legh 1 cpl = shorespah(, ), for ode ad P \ P\ oher odes he ewor he shores pah bewee he average value, P / sad he ewor coas ode, oher odes colleco. Whe a ode ad oher odes bewee pahs does o exs, s he shores pah for he 15. Characersc Pah Legh Cycle1 Cycle Peer ID Fgure : Each Node To The Nodes Of The Average Pah Legh Fgure shows he frs query he cycle before ad 00h query cycle, each ode o all oher ormal cooperave ode average characersc pah legh. Every ode 1 o 0 s ormal ode, for he res of he malcous odes. As you ca see from Fgure, he frs query cycle opology adusme before, ode average characersc pah legh dfferece, aroud ad 3; 00h query cycle opology adusme, ca be see afer several eracve, ormal ode o he oher ode of he ormal average

6 Joural of Theorecal ad Appled Iformao Techology 30 h November 01. Vol. 5 No JATIT & LLS. All rghs reserved. ISSN: E-ISSN: characersc pah dd o chage sgfcaly, ad slghly lower, ad mos of he malcous odes o ormal ode average characersc pah legh s grealy creased. Characersc Pah Legh 1 Cycle1 Cycle00 Fgure 5 shows he ormal cooperave ode ad malcous ode average umber of eghbors wh query cycle chages. Fgure 5 shows ha he average umber of eghbors, he ormal ode a he begg of he pollg perod s creased couously, ad he maaed a ca have he mum umber of coecos ad malcous odes ear; average umber of eghbors wh he pollg perod crease decrease was cosa, hs s due o malcous odes o provde relable servce, more ad more ormal ode broe ad malcous ode coeco Peer ID Fgure 3: Normal Node To The Oher Malcous Node Average Characersc Pah Legh Fgure 3 shows he frs query ad a 00h query cycle perod, ormal cooperave ode o all oher malcous ode average characersc pah legh. Normal ode wh he query cycle creased, o oher malcous ode average characersc pah legh creasg, reducg he malcous odes o provde relable servce opporues. Characersc Pah Legh 1 1 Good Peer Malcous Peer Cycle Fgure : Node Average Characersc Pah Legh Wh The Perodc Varao Fgure shows he ormal cooperave odes ad he malcous ode o all oher ormal cooperave ode average characersc pah legh wh he query cycle chages. Fgure shows ha he ormal ode o oher odes, he average characersc pah legh s very sable, maaed a a relavely low value; ad he malcous odes o ormal ode average characersc pah legh wh he eracve cycle creased, he malcous odes gradually excluded o he edge of he ewor. Average Neghbor Number 0 Good Peer Malcous Peer Cycle Fgure 5: The Average Number Of Neghbors Wh Query Cycle Chages Wh smlar eress ad he qualy of servce odes of a cluser or eghbors, hese clusers wll be o he malcous ode ureachable. I he DTA proocol, he coeco bewee s based o prevous dowload resuls, herefore, he more smlar he eres bewee odes wll be more eracve coeco s esablshed, he greaer he probably ha he odes ca, hrough he shorer roue of access o servces, hs pheomeo s also called as "small world pheomeo". Node E cluserg coeffce s defed as C =, ( 1) he umber of eghbors for ode s, E for hese eghbor odes are coeced bewee sde. Ths paper assesses he each cycle he average cluser coeffce. Fgure shows ha he average cluserg coeffce, he ode wh he pollg perod crease whch creases cosaly, eres smlar odes form a cluser, erac wh each oher more closely. 73

7 Cluser Coeffce Joural of Theorecal ad Appled Iformao Techology 30 h November 01. Vol. 5 No JATIT & LLS. All rghs reserved. ISSN: E-ISSN: [5] L Xog ad Lg Lu, PeerTrus: Supporg 0. Repuao-Based Trus for Peer-o-Peer 0. Elecroc Commues, IEEE rasacos o owledge ad daa egeerg, Vol.1, No.7, , [] Sog SS, Hwag K, Zhou RF, Kwo YK, Trused PP rasacos wh fuzzy repuao 0.1 aggregao, IEEE Iere Compug, Vol.9, 0.1 No., 005, [7] Wag YF, Y ad Saura K, Characerzg ad Cycle repuao ecoomc ad socal properes of rus ad repuao sysems PP Fgure : Average Cluserg Coeffce Wh Perodc evrome, Joural of compuer scece ad Varao echology, Vol.3, No.1, 00, [] Cox D R, Hley D V, Theorecal Sascs, 7 CONCLUSIONS Chapma-Hall, Lodo, Brsh, 197. [9] Saurabh Gaerwal, Laura K Balzao, ad Ma Topology adusme DTA cosdered he B Srvasava, Repuao-based framewor for coeced facors ad he eghbor ode's TTL hgh egry sesor ewors, ACM value, ad APT opology algorhms, despe creasg he umber of sorage overhead, bu rasacos o sesor ewors, Vol., No.3, opologcal adusme cos reduco. The 00, smulao expermes ad aalyss show ha he [] Crosby G V, Pssou N, Cluser-based opology cosruco mehod DTA valdy, repuao ad rus for wreless sesor opology o he smaller overhead posog relable ewors, Proceedgs of he h Cosumer resources, ad he malcous ode solao o he Commucaos ad Neworg coferece, edges of he ewor, mprove he servce qualy. IEEE Coferece Publshg Servces, Jauary REFRENCES: 11-1, 007, pp [11] Audu Josag, The Bea repuao Sysem, [1] Mchal Feldma, Joh Chuag, Overcomg The 5h Bled elecroc Commerce Coferece free-rdg behavor Peer-o-Peer Sysems, e-realy: Cosrucg he e-ecoomy, IEEE ACM SIGecom Exchages, Vol.5, No., 005, Coferece Publshg Servces, Jue 17-19, pp , pp [] Kudarar A M, Umamaheswar S, Avodg [1] The PEERSIM webse. [Ole]. Avalable: whe washg PP ewors, Proceedgs of hp://peersm.sourceforge.e/ he 1s Ieraoal coferece o Commucao sysems ad ewors, IEEE Coferece Publshg Servces, March 3-5, 005, pp. 1-. [3] Gaesh Kumar M, Aru Ram K, Aaya A R, Corolg free rders Peer o Peer ewors by ellge mg, Proceedgs of he Ieraoal coferece o compuer egeerg ad echology, IEEE Coferece Publshg Servces, Jauary 3-5, 009, pp [] Kamvar S D, Schlosser M T, EgeRep: Repuao Maageme PP Newors, Proceedgs of he 1h Ieraoal world wde web coferece, IEEE Coferece Publshg Servces, May 0-, 003, pp

Linking factors for gross and seasonally adjusted series

Linking factors for gross and seasonally adjusted series Lkg facors for gross ad seasoally adjused seres The purpose of hs oe s o expla problems wh he curre lkg mehodologes used he EI ad propose chages o solve dscrepaces recely defed bewee gross ad seasoally