A paging design for mobile cellular internet enhanced by locality in user-behavior

Transcription

1 Scientific Research and Essay Vo. (), pp , October 8 Avaiabe onine at ISSN Academic Journas Fu Length Research Paper A paging design for mobie ceuar internet enhanced by ocaity in user-behavior E. N. Onuka Department of Eectrica and omputer Engineering, Federa University of Technoogy Minna, Nigeria. E-mai: onukaiz@yahoo.com. Accepted 9 September, 8 Mobie ireess communication is groing in scope and importance a around the ord. The IP netork is fast becoming the goba netork of choice for mutimedia communications. Mobie IP netork is aso fast making its ay to the consumer market. Hoever, efficient and scaabe mobiity management is a key enaber in the mobie IP netorks. Secondy, mobie terminas are portabe devices that often depend on battery poer; they therefore require efficient poer management in order to proong battery ife. Paging is an important mobiity management function that, not ony heps to reduce ocation update cost, but aso heps to conserve mobie termina battery poer. Therefore, efficient paging design is necessary for reiabe IP mobiity management. Severa paging agorithms have been proposed in the iterature. These proposas, hie generay achieving the overa aim of termina poer conservation, differ in the baance beteen the amount of deay incurred to ocate a given mobie node and the bandidth cost incurred. This paper provides a paging design for ceuar mobie Internet based on the architecture of HIMA (a ompetey Hierarchica IPv6-based micro mobiity Management Architecture). The paging agorithm took advantage of the ocaity in user behavior to achieve a fast and sequentia mobie node tracking, hich consideraby reduced bandidth and deay costs. Key ords: Mobie communications, mobie IP, paging, resource management. INTRODUTION Mobie ireess communication is groing in scope and importance a around the ord, and IP netork is fast becoming the goba netork of choice for mutimedia communications. Mobie IP netork is aso fast making its ay to the consumer market. Hoever, ithout efficient and scaabe mobiity management and mobiity management functions such as efficient and scaabe paging procedure, the netork i fai to deiver a satisfactory performance. Secondy, mobie terminas are portabe devices that are battery poered and therefore, require efficient poer management. Paging is an important mobiity management function that pays the dua roe of heping to reduce ocation update signaing cost, hich improves netork scaabiity, and aso heping to conserve mobie termina battery poer, hich increases the convenience eve of mobie devices and enhances their usefuness. Wireess mobie communication introduces a considerabe amount of signaing traffic in the netork. This is because the netork uses signaing to keep track of mobie nodes (MN) ocations in the netork. This mobie node tracking is very necessary in order for the netork to be abe to timey ocate, and deiver information to MNs. The tracking process requires a MN to register its ocation ith the netork nodes, and to reguary update its ocation information ith the netork. This process hoever, consumes both the MN s and netork resources. Frequent registration and ocation information updating consumes MN s poer. It consumes netork bandidth in the air interface and on the ired portion of the netork, it aso consumes computationa and storage resources on netork nodes. In order to reduce the consumption of netork resources due to MN tracking, mobie netorks impement paging. Paging enabes the MN to be in either active or passive/standby mode. The MN is said to be in active mode hen it is activey communicating ith the netork. When in this mode, the MN s poer consumption is high. When the MN is not activey communicating ith the netork, it goes into standby state and thus turns its

2 Onuka 46 poer consumption o. The MN sti updates the netork ith its ocation information but ess frequenty. Thus, this mode heps to proong the MN s battery ife. The netork has ony a vague idea of the MN s ocation. When information comes for the MN in standby mode, the netork performs a search for the MN in order to find its exact ocation. This search is caed paging in mobie netorks. This is done by poing a or seected groups of ces in the netork. The netork node in the ce here the MN is residing then repies to the po. Paging process conserves MN s battery; hoever, it aso introduces some cost in the netork in terms of bandidth consumed for poing, both in the air interface and the ired portion of the netork, and in terms of the deay incurred in ocating a MN before information is deivered to it. If paging is performed at sma granuarity, say one ce at a time, this minimizes the average cost on netork bandidth but i maximize the average deay incurred before a given MN is ocated. On the other hand, if paging is performed on arge granuarity, say poing a the ces at the same time, this i minimize the deay incurred before a MN is found but i maximize bandidth consumption. Both bandidth and time are important netork resources. To resove this to conficting issues, researchers have taken to deveoping agorithms that tend to reduce the bandidth consumption (Ved Kafe et a., ; Kyoungae et a., 4) and at the same time reduce the deay (Goodman, 996; Rose and Yates 9) in ocating a MN. Such agorithms are usuay compex and therefore introduce another cost function in terms of difficuty in impementation, and high cost of computationa and storage resource in the netork nodes. In mobie teephone netorks, paging agorithm is designed ith respect to a given ocation update agorithm. Some of the ocation update agorithms in iterature incude the time-based, the movement-based, distance-based, state-based agorithms (Bar-Noy et a., 9; Naor and Levy, 999) etc. In mobie IP netorks hoever, paging agorithm design and performance depends, among other things, on the netork architecture. This means that a given netork architecture may better support cost-effective paging agorithm than another. There are aso to categories of mobie IP netork architecture: the macro-mobiity management architecture such as the MIP (Perkins, ) and the MIPv6 (Johnson and Perkins, ), and the micro-mobiity architecture such as the ceuar IP (ampbe et a., ), Haaii (Ramjee et a., 999) HMIPv6 (Soiman et a., ) etc. Paging agorithms have been designed based on some of these architectures (asteucia, 999; Ramjee et a., ). Hoever, none of these paging agorithms made use of MN ocation probabiity to reduce MN-search cost and compexity. The cost of ocating a MN in standby mode i be greaty reduced if the netork has some knoedge of here the MN coud be at a given time. This paper took advantage of the ocaity in mobie user behavior (Kirby, 9) to design a simpe and cost-effective paging agor- ithm. The author is currenty competing a survey on the mobiity pattern of mobie users in Nigeria. The resut is confirming the observations made in simiar surveys in other environments ike US and UK, hich shos that mobie users move about most often in their region of residence or vicinity of their orking environments. It has been aso observed that users often do not ive very far from their offices/orking paces. A these impy that in ceuar mobie netorks, users are most often ocated in their home ces or neighboring ces. The knoedge of this user behavior is very usefu in the design of costeffective and ess compex paging agorithm. The rest of this paper is organized as foos: Section presents the paging protoco; Section, the paging agorithm; Section 4, the performance anaysis; Section, resuts and discussion, hie the paper is concuded in section 6. PAGING PROTOOL Our paging design is based on HIMA (ompetey hierarchica IPv6-based micro Mobiity Architecture) (Onuka and Niu, 4). HIMA is a domain-based micro mobiity protoco hich, besides being competey hierarchica, aso utiizes a hybrid of prefix and hostbased routing techniques to achieve a scaabe mobie ireess Internet soution. The basic design of HIMA makes provision for a simpe, scaabe, and cost-effective paging service. This is because, unike other Internet micro mobiity architectures in iterature, HIMA pushed compex netork functions to the edge, thus making it easy to add ne systems and services ithout degrading performance Figure. The paging function is impemented in the SDR. A mobie node that is at home ce, and that has not transmitted packets for a period of time, say, s, goes into a seep mode. When this happens, a Rest packet is transmitted to the SDR via the home Access Router (AR). On reception, the SDR i enter the MN into a paging ist. The home AR i activate the REST fied on the MN s record, and forard the Rest packet to the SDR. When a packet originates for a MN in its home ce, the home AR checks the status of the MN, if it is AWAY or at REST, the home AR i route the packet upards to the SDR. If the MN is in the paging ist of the SDR, the SDR i initiate the paging process to ocate the MN. Each SDR is responsibe for paging a dormant MNs in its coverage area, this means that paging oad is distributed and ocaized, and aso, the ARs, hich act as home agents (HA), are not further burdened ith paging functions. This arrangement ensures oad distribution in the access netork, and therefore increases the efficiency. When a mobie node is visiting in a subdomain other than its home subdomain, and goes into seep-mode, it transmits the Rest packet to the serving SDR via the serving AR. A MN in the seep-mode is required to periodicay refresh the SDR ith the seep-mode status

3 46 Sci. Res. Essays Internet Domain No. :::?:?:?/ Legend Routing tabe ::8:?:?:?:?/7 DR :::?:?:?:?/7 DR = Gateay Router SDR = Sub-Domain Router AR = Access Router SD No. ::8:?:?:?:?/7 Routing tabe SDR SDR SDR :::?:?:?:?/64 ::8:?:?:?:?/64 Routing tabe AR AR AR AR AR AR AR AR AR ::8:?:?:?: Mobie termina Mobie termina ::8:?:?:?:D Figure. Basic architecture of HIMA. ith a period quite onger than the active MN refresh period. When packet arrives for him in this foreign subdomain, the serving SDR i initiate the paging process. The steps taken to page a MN is described under the paging agorithm in the next section. PAGING ALGORITHM There are three basic categories of paging agorithms in the iterature; they are the banket poing, the fixed paging area concept, and the dynamic paging area concept. Each of these categories introduces some costs either in the form of bandidth consumption, deay, or compexity. Usuay, the authors assume a given mobiity pattern for the users. Banket poing method requires that to ocate a MN in seep-mode, the hoe netork ces are poed simutaneousy. This contributes the minimum deay in ocating a MN, and the east ocation updating cost. But it is prohibitivey costy in bandidth consumption. The fixed paging concept requires that the netork area be divided into paging areas (PA). When a MN in seep-mode crosses a paging area boundary, it shoud update the netork ith ocation information. When packet arrives for this MN, the netork pos the PAs sequentiay in search of the MN, hen the MN is ocated, the search stops. There are various configurations proposed in iterature for poing the PAs. This method has the advantage of reducing the bandidth consumption due to poing, but it introduces increased update cost. The update cost depends on the granuarity of the PAs. Large PAs i contribute ess update cost but i increase bandidth consumption. Sma PAs i contri-bute high update cost but ess bandidth cost. The amount of deay introduced by this method depends primariy on the agorithm used in seecting hich PA to po first and hich one to po next. The research soutions here aim at getting optimum size of PAs and at deveoping the agorithm that ocates the MN faster. The dynamic paging area concept argues that the fixed PA methods are not sufficienty efficient, because PA sizes are determined based on such individua activities as packet arriva rate, mobiity rate, mobiity pattern, and ca arriva rates. But these activities are not the same for a MNs. Therefore, the dynamic paging area category deveops agorithms that configure a PA for each MN based on its mobiity pattern, packet arriva rate and mobiity rate (Wan and Lin, 997). This category of paging agorithms utiizes netork resources more efficienty, but has the disadvantage of increased compexity hich may impact on impementation cost. It may aso be difficut to impement dynamic PA schemes in ceuar netorks here netork nodes are designed to perform netork function ith respect to given netork zones or coverage areas. To improve the paging efficiency ith a ess compex agorithm that convenienty fits into the reguar ceuar system, our agorithm uses a hybrid of fixed paging area

4 Onuka 46 Figure. e arrangement for a fixed PA. (FPA) concept and the dynamic paging area (DPA) concept. The method empoyed in searching a MN in seepmode is simpe and efficient and it is as foos: The netork ide concept of paging area is fixed. Each subdomain constitutes a paging area, and contains a variabe number of ces, arranged in a circuar custer as shon in Figure (Akyidiz and Wang, ). The SDR maintains the paging ist and pages any MN in its paging area. Whenever, a MN in seep-mode crosses a FPA boundary (in this case a subdomin boundary), it is required to send a ocation update to the SDR in the ne subdomain. Each FPA is further divided into conceptua DPAs for each MN. These DPAs consist of the home ce of each MN and the neighboring ces, hich are the rings of ces surrounding it in the subdomain. The DPA for each MN is derived from the concept of ocaity in mobie user behavior. According to the ocaity concept, ith a high probabiity, MNs move about ithin their home ces or its neighborhood (Kirby, 9). Based on this, hen packets arrive for a MN in seep-mode, the SDR pages seected groups of ces in the FPA sequentiay, starting from the ce here the MN is most ikey to be found, and progresses in a decreasing order of the probabiity of here the MN coud be ocated. It starts by paging the home ce of the MN, say, ce in Figure. If the MN is not found, the next neighbor, hich is the ring of ces numbered, are paged. Then the next ring, numbered is paged. After the second ring of ces is paged ithout success, the rest of the ces in the FPA are paged. Anaysis shos that the MN i most ikey be found after the first to pos. This agorithm is simpe and costeffective as e are going to demonstrate in this paper. PERFORMANE EVALUATION In this section e carry out an evauation of our hybrid paging agorithm. Our aim is to demonstrate the effect of paging according to the ocaity probabiity. We do this by estimating the bandidth cost of ocation update and paging both in the ired and ireess inks of the access netork. So the tota cost is given by = p + u Where p is paging cost ad u is update cost. To estimates these costs, e define the fooing variabe: = bandidth cost on ireess ink = bandidth cost on the ired ink λ = rate of session arriva α = No of times a MN goes to seep-mode per unit time ρ = MN density per unit area We aso make the fooing assumptions for the ocaity probabiity p: The probabiity that a MN is in its home ce, at D = is p =. The probabiity that the MN is in its nearest neighborhood, i.e., ithin the ring D =, is p =. The probabiity that the MN is in its neighborhood ith the ring D =, is p =. The probabiity that the MN is far from home, i.e., outside the ring of D =, is p 4 =. Estimation of the update cost According to this agorithm, update packets are sent at the fooing times: When a MN sitches to seep-mode, it sends a status packet to the SDR to notify it of this status. This packet is propagated to the SDR hich paces this MN on paging ist. It i receive an acknoedgement. That yieds a cost of α () ( + ) When a MN receives a page, it responds through the serving AR by updating its current ocation. This occurs at session arriva rate, λ. That yieds λ () ( + ) Aso hen a MN crosses a FPA boundary, it updates the serving SDR. This occurs at the rate of crossing the FPA boundary, hich in turn depends on the perimeter of the FPA. The perimeter of a FPA of size D is given by L ( D) = ( D + 6) R () Where D is the number of concentric rings of ces numbered,,, Figure, and R is the radius of each ce.

5 464 Sci. Res. Essays Bandidth cost update at.6km/hr paging ithout ocaity update at km/hr Figure. Individua contributions of paging and update ithout ocaity probabiities (ce radius m). Tabe. Parameter vaues used for computation. λ R v.6.4/h 8, m..6 km/h, km/h We use the fuid fo mobiity mode to estimate the number of MNs that cross the FPA boundary per unit time. This is given by vl(d) ρ = = Φρ π R c (4) Where v is the average veocity of the MNs. This means the cost of crossing the boundary is ( + )Φρ () Putting (), (), and () together e get tota update cost as u = α ( = [(α + λ + Φ)( + ) ρ + λ( + )] ρ Estimate of the paging cost + ) ρ + ( + ) Φρ Paging cost depends on the number of ces that are paged before a MN is found. The number of ces ithin a given ring, D, is D(D+)+. Thus, the average bandidth cost of paging a singe MN is estimated as: [ D( D + ) + ]( + p (7) ) Where p is ocation probabiity. Therefore the tota cost of update and paging for MNs due to our agorithm is (6) = [(α + λ + Φ)( + )] ρ + [D( D + ) + ]( + ) pρ (8) = {[(α + λ + Φ)( + )] + p[d( D + ) + ]( + )} ρ NUMERIAL RESULTS AND DISUSSION The agorithm presupposes that the SDR knos the position of the ces in its coverage area and therefore the neighbors that constitute the DPA. We demonstrate the effect of ocaity probabiity on the average paging and update costs ith the MNs hose home ce is the ce at D = in Figure. The parameter vaues used to estimate the costs is shon in Tabe. The agorithm is designed for micro ceuar systems hich are more suitabe for urban areas. In micro ces, about 8 ~ 9% of mobie users are pedestrians or vehices moving at o speeds. We investigate the individua contributions of paging and update to the overa cost. It as discovered that the system bandidth consumed for paging is much greater than that consumed for update hen the ocaity probabiities ere not used. This resut is shon in Figure. Notice that even at higher average MN speed, hich incurs more update cost, the cost of paging ithout consideration for the ocaity in mobie behavior is sti higher. Hoever, hen the ocaity factor in mobie user behavior as considered, paging cost drasticay came don, and contributes much ess than the update in the overa cost as shon in Figure 4. Figure shos the overa cost of paging and update. Observe that the tota cost due to paging and update is reduced by over 4% hen ocaity in user movement is considered at a speed of km/h, but at oer speeds such as.6/h, hen MNs do not cross ce or PA bounda-

6 Onuka 4 Bandidth cost update at.6km/hr paging ith ocaity Figure 4. Individua contributions of paging and update costs ith ocaity probabiity (ce radius m). Bandidth cost ith ocaity, km/hr ithout ocaity, km/hr ith ocaity at.6km/hr Figure. Overa cost of paging and update, ith and ithout ocaity probabiity, at km/h and.6 km/h (ce radius m). boundaries so fast, the overa cost is reduced by over 6%. Notice that, unike other dynamic paging area agorithms that suffer from compexity, this cost reduction is achieved ith a simpe agorithm that i be easy to impement. Aso notice that the cost estimation is for the orst case condition, here a the DPAs i be searched before a MN is found. In paces ike Africa, here mobie communication is the major means of communication, this agorithm i prove more effective and economica. It shoud be noted that this ork aimed at demonstrating that consideration of user ife-stye in designing mobie systems i hep to remove unnecessary system cost, hich aso improves overa performance. The idea presented here has never, to the best of the author s knoedge, been expored in paging cost reduction in mobie communications systems; so, there is room for further improvements. This ork has some short comings in the sense that the cost estimation is fairy rough; a better computation method coud be evoved to bring out the benefits of the technique even better. Again, a MN resident in the center ce (ce numbered in Figure ) as used for this cost estimation. When every singe ce is considered, the boundary effect, ith respect to the FPA, i be taken into consideration. This is because some ces have neighbors hich beong to another FPA (subdomain). This impies that the overa cost in a given FPA i be reduced. This eve of computation i give a basis for fair comparison ith other methods. Anticipated difficuty in impementation coud arise from the fact that, ithin a FPA, each MN i be identified ith respect to its home ce and its eves of neighbors in order for the paging machine to compute the ces to be

7 466 Sci. Res. Essays Bandidth cost ith ocaity ithout ocaity Figure 6. Overa cost of paging and update, ith and ithout ocaity probabiity, at km/h and (ce radius 8 m). paged. This may impact on system memory. oncusion A paging design for a ireess ceuar Internet is considered. It as observed that considering user behaviora parameters in the design of mobiity management function such as paging heps to obtain a more accurate and ess expensive soutions. When the concept of ocaity in mobie user behavior, hich means that users often stay ithin their home ces or its cosest neighborhood, is used in constructing a paging agorithm, 4~6% reduction in paging cost coud be achieved. The agorithm is simpe and economica, hich makes it suitabe for African environments here mobie communications is the major means of communications, and here a great majority of the popuace is poor. The author observed that using parameters derived from hu-man statistics and behaviora patterns for designing systems that serve human beings is better than using parameters derived from ony ordinary mathematica ogic. The cost estimation used in this ork is fairy rough. There is therefore room for further improvements. This method may aso impact on system memory due to individua consideration of MNs. Hoever, in rura areas here MN density is ess, this probem is not envisaged. REFERENES Goodman D, Krishnan P, Suga B (996). Minimizing Queuing Deays and Number of Messages in Mobie Phone Locations. Mobie Netorks and Appications, (): Johnson DB, Perkins (). Mobiity support in IPv6, Internet Draft, draft-ietf-mobieip-ipv6-.txt, Juy; ork in progress. Kirby G (9). Locating the user, ommunications Internationa, p. 49. Kyoungae K, Sangheon P, Yanghee (4). On Reducing Paging ost in IP-based Wireess/Mobie Netorks Lecture Notes in omputer Science, Springer Berin/Heideberg, 9/4: 4-6. Naor Z, Levy H (999). e Identification odes for Tracking Mobie Users, in Proc. IEEE INFOOM 99, Ne York, NY, March. Onuka EN, Niu Z (4). HIMA: A Hybrid Prefix/Host-Based Routing Approach for Scaabe Micro Mobiity Management in a euar Mobie Internet, IEIE Trans. On omm. E87-B(9): - 8. Perkins (). IP Mobiity Support, IETF RF 44. Ramjee R, La Porta T, Thue S, Varadhan K, Sagarei L (999). IP Micro-mobiity support using HAWAII, Internet Draft, draft-ietfmobieip-haaii-.txt, June ork in progress. Ramjee R, Li L, Porta TL, Kasera S (). IP Paging Service for Mobie Hosts, Wireess netorks, 8(): Rose, Yates R (9). Minimizing the Average ost of Paging Under Deay onstraints, AM Wireess Netorks, pp. -9. Soiman H, asteuccia, E Maki K, Beier L (). Hierarchica MIPv6 mobiity management (HMIPv6), Internet Draft, draft-ietfmobieip-hmipv6-.txt, Juy; ork in progress. Ved K, Sangheon P, Yanghee (). Performance Evauation of IP Paging ith Poer Save Mechanism, Proceedings of the 8th Annua IEEE Internationa onference on Loca omputer Netorks (LN ). Wan G, Lin E (997). A Dynamic Paging Scheme for Wireess ommunications, Proceedings of the rd annua AM/IEEE internationa conference on Mobie computing and netorking, Budapest, Hungary, pp.. Akyidiz IF, Wang W (). A dynamic Location Management Scheme for next-generation Mutitier PS Systems, Proc. IEEE Trans on ire ess omm. (): Bar-Noy A, Kesser I, Sidi M (9). Mobie Users: To Update or Not to Update? AM/Batzer J. Wireess Netorks, ():-. ampbe AT, Gomez J, Kim S, Vako AG, hieh-yih Wan, Turanyi ZR (). Design, Impementation, and Evauation of euar IP, IEEE Pers. omm., Aug. pp asteucia (999). Extending Mobi IP ith adaptive Individua Paging: A Performance Anaysis, Proc. AM Mobie omputing and ommunications Revie. (): -6.