Enabling Technologies for Wireless E-Business phần 5 pps

7 Data Management for Mo bil e A d -Hoc Networ ks 1 5 1 or of the im p lementation details of e ach requested transaction. A hoppin g propert y i s added to model the mobility of the transactions. Each subtransaction represents t he unit of executi o n at one base station and is called a joey transaction (JT). The authors define a Pouch to be the sequence of global and local transactions, which are execute d un d er a g i ven KT. Eac h KT h as a un i que id ent i f i cat i on num b er con - s i st i ng of t h e b ase stat i on num b er an d un i que sequence num b er w i t hi n t h e b ase stat i on. Wh en a mo bil e un i t moves fr o m one ce ll to anot h er , t h e contro l of t h e KT chan g es to a new DAA at another base station. The DAA at the new base station creates a new JT as result of the hand-off p rocess. JTs have se q uenced identifica - t ion numbers consistin g of both t h e KT identification number and an increasin g number. The mobility of the transaction m o del is captured by the use of split t ransactions. The old JT is committed ind e p endent of the new JT. If a failure o f any JT occurs, w hi c h i n tur n may resu l t i n un d o i ng t h e ent i re KT, a compensat i on f or any prev i ous l y comp l ete d JTs must b e assure d . T h erefore , a KT cou ld b e i n a sp li t mo d e or i n a compensat i ng mo d e. A sp li t transact i on di v id es an ongo i ng t ran s a c ti o n int o se rializ ed sub tran s actions. Earlier created subtransaction ma y be committed and the remainin g ones can continue in its execution. However, the dec i s i o n o n a s t o a bo rt o r co mmit a cu rr e n t l y executin g subtransa c ti o n i s l e ft t o th e main DBMS. Previous JTs may not be compe n sated so that neither splitting mode nor compensating mode guarantees serializability of KTs. Alt h ough compensating mode assures atomicity, isolation may be violated because locks are obtained an d re l ease d at t h e l oca l transact i on l eve l . Wi t h t h e compensat i ng mo d e, j oey su b- transact i ons are ser i a li za bl e. T h e MTM k eeps a transact i on status ta bl e on t h e b ase s tat i on DAA to ma i nta i n t h e status of t ho s e transact i ons. It a l so k eeps a l oca l l og i nt o which t he M T M wr i tes t h e recor d s nee d e d for recover y purposes. Most recor d s i n t h e l o g ar e r el at ed t o KT s tat us an d so m e compensat i n g i nformat i on. A pp roaches f or Data D i ssem i nat i on and Re p l i cat i on Thi s sect i on presents re l ate d wor k on d ata di sse m i nat i on an d rep li cat i on w i t hi n w ir e l ess n e t wo rk s. Th e wo rk o n d ata d i sse minati o n a ssu m es that se r ve r s ha ve a relativel y hi g h bandwidth broadcast capacit y w hil e c li e nt s c ann o t tran s mit o r c an do so onl y over a lower bandwidth li n k . Th e d ata d i sse minati o n m ode l s ar e co n- cerne d w i t h rea d -on l y transact i ons, w h ere mo bil e c li ents usua ll y i ss u e a query t o l ocate part i cu l ar i nformat i on or a serv i ce b ase d on t h e current l ocat i on of t h e device. An o t he r m odel f o r d ata disse m i nat i on can b e app li e d w h en a g roup of c li- e nts s h ares t h e same servers an d t h e y can, i n g enera l , a l so b enef i t from accept i n g responses a dd resse d to ot h er c li ents i n t h e i r g roup. Reference [1] p resents a b r o a dc a s t- b a sed m ec hani s m f or disseminating infor- r mation in a wireless environment. To improve performance for nonuniforml y accessed data, and to efficientl y utilize the available bandwidth, the central idea is t h at servers repeate dl y b roa d cast d ata to m u l t i p l e c li ents at v ar i ous frequenc i es. Th e aut h ors super i mpose m u l t i p l e di s k s of di fferen t si zes an d spee d s to create an ar bi trar il y f i ne-gra i ne d memory hi erarc h y, an d stu d y c li ent cac h e management p o li c i es to max i m i ze performance. T h e aut h ors ar g ue t h a t i n a wi r eless m obile networ k , servers ma y h ave a re l at i ve ly hi gh b an d w id t h b roa d cast capac i t y w hil e clie nt s c ann o t tran s m i t o r c an do so o n ly over a l ower b an d w id t h li n k . Suc h 1 5 2 sy stems have been proposed for man y application domains, includin g hospital i nformation systems, traffic information systems, and wireless classrooms. Tradi- tional client–server information systems employ a pu l l-based algorithm, where clients initiate data transfers by sending requests to a server. The broadcast disks on t h e ot h er h an d exp l o i t t h e a d vantage i n b an d w id t h b y b roa d cast i ng d ata to mu l - tiple clients at the same, and thus employ a push-based approach. In this approach, y a server cont i nuous l y an d repeate dl y b roa d casts d ata to t h e c li ents, w hi c h effec - tivel y causes a creation of a disk from which clients can retrieve data as it g oes b y . T he authors then model and stud y performance of various cache techniques at the client side and broadcast p atterns at the server side within their architecture. The i nherent limitations of this a pp roach, howe v er, restrict the clients to employ read - only transactions. In additio n , it re q uires the client to w ait for incoming data until i t appears on t h e b roa d cast di s k , even t h ou gh t h e c li ent may momentar il y h ave a near-perfect w i re l ess connect i v i ty to a part i cu l ar server. Reference [75] presents an i nte lli gent h oar d i ng approac h for c ac hi ng f il es on t h e c li e nt s i de f o r m ob il e n etworks. The authors consider the case of a voluntar y , c li e nt-initiat ed d i sco n n ection as o pp osed to involu n t ar y disconnection that was u nder the scrutin y of man y approaches described earlier. Therefore, the authors a ttempt to present a solution for intelligently caching important data at the clien t side, in their case files, once the client has informed the system about its planned disconnection. This is known as the ho a rding problem, wherein hoarding tries to a eli m i nate cac h e m i sses ent i re l y d ur i ng t h e per i o d of c li ent di sconnect i on. T h e a ut h ors f i rst d escr ib e ot h er approac h es con s i st i ng of d o i ng not hi ng, ut ili z i ng exp li c- i t l y user-prov id e d i nformat i on, l ogg i ng user’s p ast act i v i ty, an d b y ut ili z i ng some semant i c i nformat i on. T h e i r approac h i s b ase d on t h e concept of prefetc hi n g , a n d can b e referre d to as transparen t analytical spying. The algorithm relies on the t not i on of wor ki n g sets. It automat i ca lly d e t ects t h ese wor ki n g s ets for a user’s ap- plications and data. It then provides generalized delimiters for periods of activity, which is used to se p arate time p eriods for wh i c h a different collection of files is re q uired. Infostat i ons [35] i s a system c o ncept propose d to suppor t m any time, man y w h ere w i re l ess d ata serv i ces, i nc l u di ng vo i ce ma il . It a ll ows mo bil e term i na l s to co mm u n ic at e t o Inf os tat io n s wi t h v ar i a ble d ata tran s m issio n rat e t o ob ta i n t he opt i m i ze d t h rou gh put. T h e ma i n id ea i s to use eff i c i ent cac hi n g tec h n i ques to h oar d as d ata as poss ibl e w h en connecte d t o serv i ces w i t hi n an i s l an d of high b andwidth covera g e, and use the cache d inf o rmati o n w h e n u na bl e t o co nta c t th e s ervices directl y . This idea is ver y similar to the previousl y described work b y [ 75 ] . Reference [38] di scusses an opt i m i st i ca ll y rep li cate d f il e system d es i gne d fo r use i n mo bil e computers. T h e f il e system, ca ll e d Rumor, uses a peer mo d e l t h at a ll ows opportun i st i c up d ate propagat i on among any s i tes rep li cat i ng f il es. T hi s w or k d escr ib es t h e d es ig n an d i mp l ementat i on of t h e Rumor f il e s y stem, an d t h e f eas ibili t y of us i n g peer opt i m i st i c rep li cat i on to support mo bil e comput i n g . T h e aut h ors di scuss t h e var i ous rep li cat i on d es ig n a l ternat i ves an d j ust i f y t h e i r c h o i ce of a p eer-to- p eer based o p timistic re p l ication. Replication s y stems can usefull y be classified alon g several dimensions base d on update t y pe, device classification, F. P e r ich e t a l. 7 Data Mana g ement for Mo bil e A d -Hoc Networ ks 153 and propa g ation methods. Conservative update repli c ation s y stems prevent all con - current updates, causing mobile users who store re p licas of data items to have thei r u pdates frequently rejected, particularly w h en connectivity is poor or nonexistent. O p timistic re p lication on the other hand a l lows any device storing a replica to per - f orm a l oca l up d ate, rat h er t h an requ i r i ng t h e mac hi ne to acqu i re l oc k s or votes f rom ot h er rep li cas. Opt i m i st i c rep li cat i on m i n i m i zes t h e b an d w id t h an d connec- ti v i ty requ i rements for perform i ng up d ates. A t t h e same t i me , o pt i m i st i c rep li ca- t ion s y stems allow conflictin g updat es t o occu r . Th e dev i ces c an be c la ss ifi ed e ither into client and servers to as p ee r s. In th e c li e nt– s e rver re p lication, all u p- dates must be first propa g ated to a server device that further propa g ates them to all clients. Peer-to-peer syste m s, on the other hand, allow any replica to propagate u pdates to any other replica. Although t he client–server a pp roach sim p lifies the system d es i gn an d ma i ntenance, t h e peer-to-peer system can propagate up d ates f aster b y ma ki ng t h e use of any a v ail a bl e connect i v i ty. Last l y, t h e l ast di mens i o n di fferent i ates b etween an i mme di a te propagat i on versus a per i o di c reconc ili at i on. In the first case, an update must be propa g ated to all replicas as soon as it is (locall y ) committed, while in the latter case a batch method can be emplo y ed to conserve t he constrained resources, such as band w idth and batter y . The authors, therefore, decided to design Rumor as an op timistic, p eer-to- p eer, re c onciliation-based re p li- cated file system. Rumor operates on file sets known as volumes. A volume is a continuous portion of the file system tree, larger than a directory but smaller than a fil e system. Reconc ili at i on t h en operates at t h e vo l ume granu l ar i ty, w hi c h i nc r e ases th e poss ibili ty of con fli ct i ng up d ates an d l arge memory an d d ata requ i rement fo r storage an d sync h ron i zat i on. At t h e s ame t i me, t hi s approac h d oes not i ntro d uce a high ma i ntenance over h ea d . A ddi t i ona lly , t h e Rumor s y stem emp l o y s a se l ect i ve rep li cat i on met h o d an d a per-f il e reconc ili at i on mec h an i sm to l ower t h e unnecessar y cos t . Reference [41] has investigated an epide m ic update protocol that guarantees consistency and serializa b i lity in spite of a write-anywh e re capability and conduct s imulation experiments to evaluate this protocol. The authors argue that the tradi - t i ona l rep li ca management approac h es suffe r from s i gn i f i cant p erformance pena l - t i es. T hi s i s d ue to t h e requ i rement of a s y nc h ronous execut i on o f eac h i n di v id ua l rea d -an d -wr i te operat i on b efore a transact i on can comm i t. An a l ternat i ve approac h i s a l oca l execut i on of operat i ons w i t h o ut s y nc h ron i zat i on w i t h ot h er s i tes. In t h e ir approac h , c h an g es are propa g ate d t h rou gh out t h e networ k us i n g an ep id em i c approach, where updates are pi ggy -backed on messa g es. This ensures that eventu- all y all updates are propa g ated throu g hout the entire s y stem. The authors advocate that the e p idemic a pp roach works well f or sin g le item updates or updates tha t commute ; h owever , w h en use d for mu l t i operat i on transact i ons, t h ese tec h n i ques d o not ensure ser i a li za bili ty. To reso l ve t h ese i ss u e s, t h e aut h ors h ave d eve l ope d a h y b r id approac h w h ere a transact i on executes l oca ll y an d uses ep id em i c commu- n i cat i on to propa g ate a ll i ts up d ates t o a ll rep li cas b efore actua lly comm i tt i n g . T ransact i on i s on ly comm i tte d , once a s i te i s ensure d t h at up d ates h ave b een i ncor - p orate d at a ll cop i es t h rou gh out t h e s y s tem. T h e y present exper i menta l resu l ts s upportin g this approach as an alternative to ea g er update protoc o l s f o r a d i s tri b - uted database environment where serializabilit y is needed. The e p idemic p rotocol 1 5 4 relieves some of the limitations of the traditional approach b y eliminatin g g lobal deadlocks and by reducing delays caused by blocking. A d ditionally, the authors claim that the e p idemic communication tec h ni q ue is more flexible than the reliable, synchronous communication req uired by the traditional ap q q proach, and justify this b y present i ng resu l ts of t h e i r performance eva l uat i ons. T h ese resu l ts i n di cate t h a t f or mo d erate l eve l s of rep li cat i on, ep id em i c rep li cat i on i s an accepta bl e so l ut i on w hil e s i gn i f i cant l y re d uc i ng t h e transm i ss i on cost. 7.3.5 Transaction Management Layer T his subla y er deals with the m ana g in g transactions initiated by devices in mobile a d h oc n e t wo rk s. F o r a tran s a c ti o n t o succe e d, a device must b e a b l e t o co mmit it s updates at the appropriate data manager that can be located in the wired network t o r on some of the device’s p eers in the c u r rent vicinity. Additionally, when dat a a re modified at the primary side, all mobile devices should receive corresponding u p d ates for t h e i r rep li cas. Reference [57] d ef i nes transact i on as a b as i c un i t of cons i stent an d re li a bl e c omput i ng, cons i st i ng of a sequence of d ata b ase operat i ons execute d as an atom i c a ct i on. T hi s d ef i n i t i on encom p asses t h e fou r i m p ortant p ro p ert i es of a transact i on: a tom i c i t y , cons i stenc y , i so l at i on, an d d ura bili t y (i .e., ACID propert i es ) . A tomicit y r e f e r s t o t he fa c t t h at a tran s a c t i on i s treate d as a un i t of o p erat i on. C onsistenc y refers to a transaction being a correct t r ansformation function from mapping one c onsistent state of a database o nto another consistent state. I solation re q uires tha t t he data changes trigg e red by a transaction are hidden from others until the trans- a ct i on comm i ts. Last l y, duration of a transact i on i mp li es t h at an outcome of com - m i tte d transact i on i s permanent an d can n o t b e su b sequent l y remove d . Anot h er i mportant feature of a transact i on i s t h a t i t a l ways term i nates, b y e i t h er comm i tt i ng th e c h an g es or by a b ort i n g a ll i ts up d ates. Th e transact i on pro bl ems i n m obile e n vi r o nm e nt s ar ise due t o t he tra di t io na l concurrency control technique. The cont rol technique often relies on locking, t where a client wishin g to modif y data o n the server database must first ac q uire a valid lock. For exam p le, in the two- ph a se commit p rotocol (2PC) [32, 57] each p artici p ant and coordinator enter a state, where the y are waitin g on a messa g e from one another. The only other escape from the idle state is only triggered by an m e xp i re d t i mer. S i nce mo bil e d ev i ces may b ecome i nvo l untar il y di sconnecte d , t hi s t ec h n i que ra i ses ser i ous pro bl ems. If a l oc k i s esta bli s h e d on a mo bil e d ev i ce , w hi c h b ecomes di sconnecte d, t h e l oc k m a y b e act i ve for a l on g t i me, t h us bl oc k - i n g t h e term i nat i on of a transact i on. On t h e ot h er h an d, w h en a lock is es ta blished on a wired device b y a mobile device, w hich since becomes disconnected, the data availabilit y is reduced. These problems h a v e s p urred numerous s o l utions [ 14, 27, 3 0, 54, 67]. These approaches are usuall y based on modelin g a novel breed o f mobile transactions by proposing different transaction-processing t tec h n i ques, suc h as [30], an d/ or b y re l a x i ng t h e ACID propert i es a s for examp l e i n [80]. Hav i ng re l axe d t h e ACID propert i es, one c an no l onger guarantee t h at a ll rep li- cas are synchronized. Consequently, the da ta management layer must address this a a F . P e r ich e t a l. 7 Data Mana g ement for Mo bil e A d -Hoc Networ ks 155 i ssue. Traditional replica control protocols, based on votin g or lock principles [ 31], assume that all replica holders are always reachable. This is often invalid i n mobile environments an d may limit the ability t o synchronize the re p lica locate d on mobile devices. Approaches addressing this issue include data division into vo l ume groups an d t h e use of vers i ons fo r pess i m i st i c [27] or opt i m i st i c up d ates [ 38, 50]. Pess i m i st i c approac h es requ i re ep id em i c or vot i ng protoco l s t h at f i rs t mo di fy t h e pr i mary copy b efore ot h er r e pli cas can b e up d ate d an d t h e i r h o ld ers can o p erate on them. On the other hand, o p timistic re p lication allows devices to operate on their replicas immediatel y , which ma y result in a conflict that will re q uire a reconciliation mechanism [4 1 ]. Alternativel y , the co nfli c t m us t be avoided by calculating a voting quorum [48] for distributed data objects. Each rep- l ica can obtain a quorum by gathering w eighted votes from other replicas in the system an d b y prov idi ng i ts vote to ot h ers. Once a rep li ca o b ta i ns a vot i ng quo- rum, i t i s assure d t h at a ma j or i ty of t h e rep li cas agree w i t h t h e c h anges. Conse - quently, the replica can com mit its proposed updates. m m 7.3.6 Security and Privacy Plane Th e i ssues re l ate d to secur i t y an d pr i vac y are v er y i mportant i n mo bil e a d h oc networ k s. T h e t h ree ma i n reasons for t hi s are t h e l ac k of an y not i on of secur i t y on th e transm i ss i on me di um, t h e l ac k of g uarantee d in te g r i t y of d ata store d on mo bil e devices in the environment, and the real possibility of theft of a user’s mobile device. Des p ite the increased need for security and privacy in mobile environments, the i n h erent constra i nts on mo bil e d ev i ces h ave prevente d l arge-sca l e researc h an d d eve l opment of secure protoco l s. L i g h twe i gh t vers i ons of Internet secur i ty proto - c o l s are lik e l y to fa il b ecause t h ey i gnore or m i n i m i ze certa i n cruc i a l aspects of t h e l atter, i n or d er to save computat i on a n d/ or memor y . T h e trava il s of t h e w i re d e qu i va l ent pr i vac y ( WEP ) protoco l d es ig ne d for t h e IEEE 802.11 b are we ll k nown [ 81]. T h e IEEE 802.11 b wor ki n g g roup h as now re l ease d WEP2 for t h e ent i re c lass of 802.1x p rotocols. Bluetooth also provides a link la y er securit y protocol t hat co n s i s t s o f a pairing procedure, which accepts a user-supplied passkey to g g enerate an initialization ke y . The initializati o n ke y is used to calculate a link ke y , w hi c h i s f i na ll y use d i n a ch a ll enge–response sequence, after b e i ng exc h ange d . Th e current B l uetoot h secur i ty protoco l uses proce d ures t h at h ave l ow computa - ti on comp l ex i t y , so t h e y are s uscept ibl e to attac k s. To sec u re d ata at t h e rout i n g l a y er i n c li ent–server an d c li ent–prox y –ser v e r arc hi tectures, IPSec [ 49 ] i s use d i n c on j unct i on w i t h Mo bil e IP. Researc h i n s e c ur i n g rout i n g protoco l s for networ k s u sin g peer-to-peer architectures has resu l ted in interestin g protocols such as Ariadne [86] and securit y -aware Ad hoc routin g [85]. The wireless transport la y e r securit y (WTLS) protocol is the onl y known p rotocol for securin g transport la y er d ata i n mo bil e networ k s. T hi s protoco l i s part of t h e WAP stac k . WTLS i s a c l ose re l at i ve of t h e secure soc k ets l aye r protoco l t h at i s d e j ure i n secur i ng d ata i n t h e Internet. Transact i on an d app li cat i o n l ayer secur i ty i mp l ementat i ons are a l so b ase d o n SS L. 1 56 7.3.7 Sy stem Mana g ement Plane T he system management plane provides interfaces so that any layer of the stack in Fig. 7.1 can access system level informat i on. System level information includes d ata suc h as current memor y l eve l , b att e ry power, an d t h e var i ous d ev i ce c h arac - ter i st i cs. For examp l e, t h e rout i n g l a y er m ight need t o de t e rm i n e whe t he r t he cu rr e n t li n k l a y er i n use i s IEEE 802.11 b or B l uetoot h to d ec id e pac k et s i zes. Transact i on mana g ers will use memor y information to deci d e whether to respond to incomin g transaction re q uests or to p revent the use r from sendin g out an y more transaction requests. The application lo g ic will a c q uire device characteristics from the s y ste m management plane to inform the other end ( s e rver, proxy, or peer) of the device’s s creen resolution, size, and other related information. The service discovery laye r might use system level information to decide whether to use semantic matching or si mp l e matc hi ng i n di scover i ng serv i ces. 7.4 Peer-to-Peer Data Management Model Thi s sect i on presents t h e MoGATU mo d e l i ntr o d uce d by [58–63], w hi c h attempts to answer mobile data mana g ement c h allen g es raised b y traditional mobile com- p utin g environments and those challen g es specific to mobile ad hoc networks. The g oal of the model is to allow mobile devices present in the environment to utilize e fficiently their current resource-rich vici n i ty while pursuing their individual and collective tasks. The model makes three p ro p ositions: 1. Postulate 1: All d evices in mobile a d hoc networks are p eers. T he w idespread adoption of short-range ad h oc networking technologies allows mobile devices to interact with othe r d evices in their current vicinity w ithout the need of a back-end wired infrastructure. As a result, a mobile dev i ce c an be bo th an inf o rmati o n co n su m e r, i.e., a client in the traditional mobile model, or an informati o n p rovider, i.e., a server in the traditional mobile model. Consequentl y , there a re no ex p licit clie n ts an d se r ve r s in t hi s para di gm anymore. Instea d , t h ey b ecome peers t h at can b ot h consume an d prov id e di fferent serv i ces an d d ata. 2 . P ostu l ate 2: A ll devices in mobile ad hoc networks are semiautonomous, self-describing, highly i nteractive, and adaptive. T h e c h aracter i st i cs of mo bil e a d h oc networ k s i mp l y t h at a d ev i ce’s v i c i n i ty i s hi g hl y vo l at il e. S i nce a ll d ev i ces i n t h e v i c i n i t y ma y b e mo bil e, t h ere i s no g uarantee a b out t h e d urat i on of a connect i on amon g an y pa i r of mo bil e d ev i ces. Conse- quent ly , mo bil e d ev i ces must b e auto n omous i n or d er to operate correct ly w hile their vicinity changes consta n t ly. Additionally, as mobile devices move and as new data may a rrive at any moment, there is no guarantee about the type of information available at any given time and space. M o bil e d ev i ces must b e a d apt i ve to t h i s nature of t h e env i ronment i n t h at t h ey must b e a bl e to c h ange t h e i r funct i ona li ty an d nee d s b ase d on w h at F . P e ri c h e t al . 7 Data Mana g ement for Mo bil e A d -Hoc Networ ks 15 7 i s currentl y available to them. Mob i le devices must also be self-describin g . T hey must be able to articulate their n e eds, which together with adaptivity will allow them to better utilize their vi c inity. Finally, mobile devices mus t be highly interactive by offering data and services y to their peers and b y query i ng t h e i nformat i on ava il a bl e on t h ose peers. 3 . P ostu l ate 3 : A ll devices in mobile ad hoc networks require crosslayer inter - action between their data management and communication layer s . It i s i n su ffi c i e nt f o r m ob il e dev i ces t o emplo y mobile data mana g ement solu - t i o n s that do n o t co n s i de r th e u n de r l y in g network characteristics. At the same time, it is simpl y not enou g h t o attem p t to solve the underl y in g net - working problems, including device d iscovery and routing of traffic between devices, independently from the data management aspect. Suc h so l ut i ons wou ld waste t h e li m i te d b an d w id t h an d ot h er resources. T h ey wou ld a l so fa il d ue to t h e i na bili ty to a ll ow mo bil e d ev i ces to comp l ete l y sat i sfy t h e i r i n di v id ua l an d co ll ect i ve tas k s. As argue d i n Sect. 7.3, i t i s imperative that all mobile devices emplo y a m ode l that co n s i de r s bo th th e networkin g and the data mana g ement aspects of the environments. Fi g ure 7.2 illustrates the correspondin g r e p resentation of mobile devices from t he MoGATU model’s perspective. Applying definitions from [84], the model can be classified as chained a r chitectu r e with a random re p lication and local inc r e- m ental polic y . Fi g. 7.2 . Dev i ce a b stract i on i n t h e Mo G AT U mo d e l Th e mo d e l i s a c h a i ne d arc hi tecture b ecause eac h d ata source , or consumer , registers with a local information manager only. Remote information managers r p resent on other devices in the s y stem are unaffected. When a quer y is placed, firs t t he local information mana g er attempts to answer it. If it is unable to answer the query, only then the information manager, forwards the query to some remote Manager to which it is curren t ly connected, i.e., to which it is chained . d Th e mo d e l emp l oys a ran d om rep li cat i on po li cy b ecause any mo bil e d ev i ce can o b ta i n an d cac h e a spec i f i c d ata o bj ects. T h ere i s no pr i or k now l e d ge t h at can d e- t erm i ne t h e l ocat i on of a ll cop i es of a spec i f i c d ata o bj ect w i t h respect to t i me an d space . 1 58 T he model also emplo y s a local incremental update polic y because there is no g uarantee how long two devices may be able to communicate with each other. In order to overcome short session durations and network bandwidth limitations , i nformation managers do not attempt to load all information their peers have d ava il a bl e. Instea d , an i nformat i on mana ge r on ly l earns i ncreme n ta lly t h e capa bili- t i es of i ts peers as i t quer i es t h em or t h rou gh rece i v i n g remote a d vert i sements. Spec i f i ca lly , t h e MoGATU mo d e l a dd resses t h e d ata mana g ement c h a ll en g es f rom Sect. 3.4 as follows: • A utonom y . As described earlier, all devices are treated as inde p endent enti- t ies acting autonomously from others. • M o b i l it y . T h e mo d e l d oes not p l ace any restr i ct i on on t h e mo bili ty patterns of d ev i ces. • H etero g eneit y . Mo bil e a d h oc networ k s are highly h etero g eneous i n terms of d ev i ces, d ata resources, an d networ ki n g tec h no l o gi es. T h e mo d e l a dd resses thi s i ssue by h av i n g eac h d ev i ce i mp l ement an i nformat i on mana g er. Eac h s t o r ed inf o rmati o n an d se r v i ce that i s a ble to g enerate additional informa- t ion is further abstracted b y information providers. Lastl y , networkin g t echnolo g ies are abstracted in terms of communication interfaces that allow d ev i ces to i nteract regar dl ess o f t h e un d er l y i ng networ k s. • D istri b utio n . Mo bil e d ev i ces may h ave mu l t i p l e i nformat i on prov id ers, eac h h o ldi ng a di str ib ute d su b set of t h e g l o b a l d ata repos i tory. T h e mo d e l a ll ows d ev i ces to a d vert i se, so li c i t, exc h an g e, an d mo di f y suc h d ata w i t h t h e i r peers. • L ack o f a g lobal catalo g an d schema . The model does not require a g lobal catalog or schema. Instead, the model em ploys ontologies based on a seman- m m ticall y rich lan g ua g e – a set of common vocabularies. These ontolo g ies ena bl e d ev i ces to d escr ib e i nformat i on pr o v id e d b y any i nformat i on pro- v id er. T h ese onto l og i es are a l so use d to a d vert i se, di scover, an d query suc h i nformat i on among d ev i ces. • N o g uarantee o f reconnection . To reme dy t h e effects of reconnect i on, t h e mo d e l i s a b est-effort on ly an d re li es on proact i ve ly cac h ed i nf o rmat io n . A ddi t i ona lly , a d ata- b ase d rout i n g a lg or i t h m i s i ntro d uce d, w hi c h a ll ows closer devices to provide answers to queries placed b y their peers when- ever data are more important than its ori g in. • S p atiotem p oral varia t ion of d ata an d d ata source availabilit y . The model encourages every d ev i ce to gat h er i nformat i on proact i ve l y w i t h out h uman i nteract i on. A user prof il e i s use d for r epresent i ng t h e necessary i nforma - t io n i n o r de r t o a llow devices t o act i n d epen d ent ly . T h e prof il e i s a l so annotate d i n a semant i ca lly r i c h l an g ua g e an d i s use d by d ev i ces for a d apt- i n g t h e i r cac hi n g an d quer yi n g b e h av i or. T he model abstracts each p eer device in terms of information p roviders, infor - mation consumers, and information mana g ers. Additionall y , the model defines abstract communication interfaces for supportin g multiple networkin g technolo - gi es. T hi s i s ill ustrate d i n F i g. 7.2. F . P e r ich e t a l. 7 Data Management for Mo bil e A d -Hoc Networ ks 1 5 9 Information p roviders, described in later, re p resent the available data sources. E very information provide r h olds a p artial distributed set, a f ragmen t , of heteroge - neous data available in the whole mobile ad hoc network. The data model, describe d i n Sect. 7.4.1, is a set of ontologies with data instances ex p resse d in a semantic l anguage. Eac h i nformat i on prov id er stores i ts d ata i n t h e d ata’s b ase form accor d - i ng to t h e onto l ogy d ef i n i t i on. Data i nvo l v i ng one onto l ogy are a l rea d y expresse d i n b ase form an d store d i n t h e format i n w hi c h t h ey were o b ta i ne d . For d ata i nvo l - vin g multiple ontolo g ies, a provider dec o m p oses the data into their base forms and maintains a view linkin g to the base forms. Usin g this approach a view is repre - s ented as a list of pointers to the respective base fra g ments. It ma y be impossible t o maintain global consistency a m ong all information providers because the mobile ad hoc network fre q u ently remains partitioned. As a result, mobile nodes attempt to b e v i c i n i ty-cons i stent on l y. Informat i on consumers, d escr ib e d l ater, rep r esent ent i t i es t h at query an d up d ate d ata ava il a bl e i n t h e env i ronment. Informat i on consumers can represent not on l y human users but also proactive a g ents that activel y prefetch context-sensitive i nf o rmati o n fr o m o th e r dev i ces in th e e n v ir o nm e nt . L astl y , an instance of an information mana g er, described later, must exist on e very mobile device. Information managers a re res p onsible for n etwork communi- cation and for most of the data management functions. Each information manage r i s res p onsible for maintai n ing information about p eers in its vicinity. This infor - mat i on i nc l u d es t h e types of d ev i ces an d i n f ormat i on t h ey prov id e. An i nformat i on manager a l so ma i nta i ns a d ata cac h e for stor i ng i nformat i on gat h ere d from ot h e r mo bil e d ev i ces an d for cac hi ng i nformat i on generate d b y i ts l oca l prov id ers. N ot ill ustrate d i n F ig . 7.2 i s t h e fact t h at eac h i nformat i on mana g er a l so i nc l u d es a user’s prof il e ref l ect i n g some of t h e user ’ s b e li efs, d es i res, an d i ntent i ons ( BDI ) . Th e BDI mo d e l h as b een exp l ore d i n m u l t i a g ent i nteract i ons [11]. For prof il es, it s ignificantly extends [21], which ex p licitly enumerates data and its utility. In con- trast, by using the BDI concept, profiles adapt to the environment by varying both data and their utility over time and present situations. Therefore, a profile enables p roact i ve d ev i ce b e h av i or b ecause mo bil e d ev i ces can a d apt t h e i r operat i on an d f unct i ona li ty d ynam i ca ll y b ase d o n t h e current contex t an d user’s nee d s w i t h out wa i t i n g exp li c i t ly for a user’s i nput. T h e i nformat i on mana g er uses t h e prof il e fo r a d apt i n g i ts cac hi n g strate gi es an d for i n i t i at i n g co ll a b orat i on w i t h peers i n or d e r t o ob ta i n desi r ed i nf o rmat io n . 7.4.1 Data Representation Model E ver y mobile device holds a subset of g loball y available hetero g eneous d ata . Since the mobile ad hoc networks are, b y d efinition, open s y stems there are no restric - tions or rules specifying the type and format of available data. In order to support t h eterogeneous de vi ces b ut at t h e same t i me a ll ow t h ese d ev i ces to i nteract , i t i s i mportant t h at t h ese d ev i ces s pea k using a common language. k Th e efforts of t h e sema n t i c We b commun i ty attemp t to a dd ress s i m il ar i ssues by d ef i n i n g a semant i ca lly r i c h l an g ua g e – t h e We b onto l o gy l an g ua g e ( OWL ) 1 60 [ 82]. This semanticall y rich lan g ua g e all o w s the s p ecification o f numerous t y pes of data in terms of classes and their pr o p erties, and also defines relationshi p among the classes and the properties. It is advantageous to employ their propose d s olution. In fact , as advocated in [60], the use of ontologies in these environments is vi ta l. B y a dh er i n g to an a l rea dy ex i st i n g l an g ua g e, t h e s y ntax an d r ules do n o t h a ve t o b e re i nvente d by d ef i n i n g new forma l l an g ua g e. Secon d , by ut ili z i n g a l an g ua g e used b y the semantic Web communit y , mobile devices will be able to use the resources available in their current vicinit y as well as the vast resources available on the Internet. Therefore, the model assumes that information instances, p rofiles, and other data objects are rep r esented using the OWL. By using ontologies, the model, however, imposes a requirement that each d ev i ce i s a bl e to parse OWL-annotate d i nformat i on. T hi s i s not to sa y t h at a ll d ev i ces w ill, or must , un d erstan d a ll onto l o gi es. Rat h er, t h e mo d e l ant i c i pates a s cenar i o w h ere eac h d ev i ce h as some k now l e dg e over a set of onto l o gi es. For ne w o r u nkn ow n on t olo g ies alread y present in the environment, a device can at leas t detect some metadata information b y a ppl y in g default OWL r u l es. Thi s w ill all ow a device to match q ueries with informat i o n providers’ advertisements without an y knowledge about the particular data. For example, this may allow a device that understands ontolog y A to use data annotated in an unknown ontology B , i f B is a subClassOf A . For example, a device may not be able to deduce that Joy Luck is a C hi nese restaurant , b u t i t w ill at l east k now t h a t Joy Luc k i s a restaurant. S i nce eac h d ev i ce i s requ i re d to on l y parse OWL-annotate d i nformat i on, t h e i ntro d uct i on of onto l og i es i n t h e system do e s not requ i re muc h more process i ng resources than alread y available. 7.4.3 Application Layer T he application la y er defines the specifi c lo g ic emplo y ed b y mobile devices. The l o g ic specifies the interface for allowin g users to operate over the devices. It also defines logic for devices to initiate actions and interact with other mobile devices. This logic can be abstracted in two types. One type represents information con- d sumers, i.e., applications that search for information, while the other type of logic represents i nformat i on pro d ucers. Informat i on pro d ucers are t h ose app li cat i ons th at can store or pro d uce i nformat i on req u e ste d b y ot h er app li cat i ons, w hi c h ca n r eside o n t he s am e o r r e m o t e devices. The model can be represented using the layered approach illustrated in in Fig. 6.1. d C ommun i cat i on i nterfaces are respons ibl e for t h e funct i ona li t y of t h e commun i ca - t ions layer. Information providers and infor m ation consumers are s p ecific instances defining the application logic a t t he application layer. Lastly, the information manager combines the tasks of data and transaction management layers and thei r di scovery an d l ocat i on su bl ayers. F i gure 7.3 ill ustrates t hi s poss ibl e reor d er i ng o f th e var i ous components of t h e MoGATU mo d e l . F . P e r ich e t a l. [...]... Processing, pp 427–434, 2000 174 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 F Perich et al T Imielinski and B.R Badrinath Mobile wireless computing: challenges in m data management Communications of the ACM, 37(10):18–28, 1994 T Imielinski, S Viswanathan, and B.R Badrinath Data on air: Organization and access IEEE Transactions on Knowledge and Data Engineering, pp 352 –372, May/June 1997 J Ioannidis,... representation for the underlying networks is twofold: First, the addition of a new networking technology does not require changes to the information manager component An information manager is not burdened by different packet formats and message sizes for different networking technologies Moreover, the abstraction allows an information manager to route data across multiple network technologies at once An information... Moreover, each information manager can intercept all messages it receives or routes to provide shortcuts for cached routes Here, each information manager maintains a route entry for one-hop peers The information manager also maintains a route entry for peers 7 Data Management for Mobile Ad-Hoc Networks 169 more than one hop away if those peers are used in on-going interactions or the information manager... limited to one-hop neighbors only 7 Data Management for Mobile Ad-Hoc Networks 1 65 Query Processing Component While advertising and soliciting for Information providers is an important functionality, the key objective of an information manager is to provide querying capabilities The querying is initiated by an information consumer, which sends the information manager a query annotated in OWL, as defined... renewed its registration Additionally, information manager may advertise the provider to other devices in the vicinity if the provider is willing to process queries for remote devices The advertisement frequency is a tunable parameter for each information manager Information Consumers Information consumers represent entities that can query, consume, and update data Information consumers represent not only... layer for data management in mobile ad hoc networks It details the information manager, which is responsible for proactive profile-driven discovering, processing, combining, and storing of data available in the environment The information manager is also responsible for evaluating the integrity of peer devices and the accuracy of peer provided information, to provide the best results to its local information... manager also maintains information about peers in its vicinity This information includes the identity of devices – a unique identification number similar to an Internet protocol address, and types of information they can provide, i.e., provider advertisements Lastly, information manager maintains a data cache for storing information obtained from other mobile devices as well as the information provided by... connectivity in mobile information systems In Database and Expert Systems Applications, pp 51 0 52 0, 1996 Pullela, L Xu, D Chakraborty, and A Joshi A component based Architecture for Mobile Information Access In Workshop in conjunction with International Conference on Parallel Processing, pp 65 72, August 2000 J Rekesh UPnP, Jini and Salutation – a look at some popular coordination frameworks for future network... queries 2 Information manager caches remote advertisements only for the lifetime specified in the message or until replaced by another entry 3 Information manager caches both advertisements and answers 4 Information manager caches all advertisements and answers, and makes them available to other peers This type of an information manager can effectively serve as a temporary partial catalog for all peers... process model that provides the information, and the input, i.e., query, restrictions, and requirements Moreover, the language supports efficient discovery and matching approaches required for locating information providers, cached answers, or for answering queries [18] 162 F Perich et al Upon start-up, each information provider registers itself with the local information manager by sending a registration . b r o a dc a s t- b a sed m ec hani s m f or disseminating infor- r mation in a wireless environment. To improve performance for nonuniforml y accessed data, and to efficientl y utilize the available. b an d w id t h li n k . Suc h 1 5 2 sy stems have been proposed for man y application domains, includin g hospital i nformation systems, traffic information systems, and wireless classrooms. Tradi- tional. prov id e. An i nformat i on manager a l so ma i nta i ns a d ata cac h e for stor i ng i nformat i on gat h ere d from ot h e r mo bil e d ev i ces an d for cac hi ng i nformat i on generate d