RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM (artículo arbitrado)

Handoff Between a Wireless Local Area Network (WLAN) and a Wide Area Network (UMTS)

Traspaso entre una red inalámbrica de área local (WLAN) y una red de cobertura amplia (UMTS)

J. Sánchez-García Departamento de Electrónica y Telecomunicaciones, CICESE, México. E-mail: [email protected]

Luis A. Villaseñor-González Departamento de Electrónica y Telecomunicaciones, CICESE, México. E-mail: [email protected]

Mario E. Vaquera-Flores Departamento de Electrónica y Telecomunicaciones, CICESE, México. E-mail: [email protected]

Raúl Aquino-Santos Facultad de Telemática, Universidad de Colima, México. E-mail: [email protected]

(Recibido: octubre de 2007; aceptado: diciembre de 2007)

Abstract With the appearance of wireless data networks with variable cove rage, band width and handoff strate gies, in addition to the growing need of mobile nodes to freely roam among these networks, the support of an intero pe rable handoff strategy for hybrid wireless data networks is a require ment that needs to be addressed. The current trend in wireless data networks is to offer multi media access to mobile users by emplo ying the wireless local area network (WLAN) standard IEEE802.11 while the user is located indoors; on the other hand, 3rd gene ra tion wire less networks (WAN) are being deployed to provide coverage while the user is located outdoors. As a result, the mobile node will require a handoff mechanism to allow the user to roam between WLAN and WAN environ ments; up to this date several strate gies have been proposed (Sattari et al., 2004 and HyoJin, 2007) in the lite rature, however, none of these have been stan dardized to date. To support this intero pe rabi lity, the mobile node must be equipped with configu rable wireless inter faces to support the handoff between the WLAN and the WAN networks. In this work a new algo rithm is proposed to allow a mobile node to roam between a wireless local area network (IEEE802.11) and a WAN base station (UMTS), while emplo ying IP mobi lity support. The algo rithm is implemented in simulation, using the Network Simulator 2.

Keywords: Handoff, WLAN, UMTS, Mobile IP (MIP).

Resumen Hoy en día exis te una gran varie dad de re des ina lámbri cas de da tos que propor cio nan co - bert uras varia b les, dife r entes anchos de banda que imp lemen t an dife r entes estra t egias de traspa so; adicio nal m ente se presen ta una gran deman da por part e de los usuari os móv iles por uti li zar li bre men te la co ber tu ra que pro por cio nan es te ti po de re des. Por lo an te rior, Handoff Between a Wire less Local Area Network (WLAN) and a Wide Area Network (UMTS)

exist e una nece si dad de la imp lemen ta ci ón de meca ni smos de traspa so que permi tan la inte r o- pera b ili d ad en los traspa sos que se presen tan en redes inalám bri cas híbr idas. La tenden cia actual , es la de propor cio nar acce so para aplica cio nes multi m edia util izan do redes inalám bri - cas de área local (WLAN), utili zan do el están dar IEEE 802.11 para inte rio r es; por otro lado, se están inst alan do redes inalám bri cas de 3a gene ra c ión para propor cio nar cober tu r a en exte - rior es. Como resul ta do, los nodos móvi les requie r en un meca nis mo de traspa so para permi tir la comu n ica ció n en ambien tes WLAN y WAN; a la fecha, se han propues t o diver sas estra t egias en la lite r atu r a (Satta ri et al., 2004 y HyoJin, 2007); sin embar go, ningu na de és tas ha si do es tan da ri za da. Pa ra ha bi li tar la in te ro pe ra bi li dad, el no do mó vil de be ser au to-con fi gu ra ble para per mitir la mo vili dad entr e las re des WLAN y WAN. En este traba jo se propo ne un algo - ritmo para per mitir el tras paso de un no do móvil dentro de una red ina lámbri ca de área local (IEEE 802.11) y una es tación base WAN (UMTS), utili zan do el sopor te de movi li dad IP. El algo rit mo se imple m entó en simu la ción utili zan do la her ramien ta Network Simulator 2.

Descripto res: Traspa so, re des inalám bri cas de área local (WLAN), UMTS, IP Móvil (MIP).

I. Intro duc tion charge an ad ditiona l fee to provide data services to their cus tom ers. With the large deplo y ment of WLAN access points There are two HOs types (Stemm et al., 1998): the (APs) in public places, high speed wireless data services hori zon tal HO that is defined like a HO between BSs are becom ing in creasingl y popu lar, but they are not (or APs) that are using the same wireless network tech- availabl e every where. Simi l arly, cellu lar comp anies are nol ogy, and the verti cal HO which takes place between migrat ing from second gener a tion (2G) to 2.5 or 3G base stations that are us ing differ ent wireless network - technol ogies, thus allow ing them to offer higher data ing technol o gies; this work consid ers the verti cal HO rate wide area network (WAN) connec tions that enable strategy, as well as, hori zon tal HOs be tween WLANs. faster data transf ers. Futu re wireless multi me d ia termi - The verti cal HO is divided in two cat e gories: an as- nals will be able to con nect to both WLANs and WANs, cendan t verti cal HO defined as a HO from a WLAN to- and switch between them without in terrupt ing on go- wards a wireless network with larger cell size whose ing com mu ni ca tion ses sions. cover age in cludes the WLAN, and a de scendan t verti cal It is highly ad vanta geous to make use of WLAN net- HO defined as a HO towards a wireless network with works, when ever they are available, and be able to smaller cell size. To illus trate this pro cedure, figure 1 switch to a wide area cel lular network when the WLAN shows a per son walking with a mo bile termi nal while is no longer available. This is because WLANs (like connected to the Internet. As the termi nal moves in side IEEE 802.11) offer high data rates at the expense of a re- the hybrid net work, several HOs must take place: start- duced cover age area (gener all y free), while WANs –like ing at point a), the termi nal passes from UMTS to a the Gen eral Packet Ra dio Service (GPRS) or Univer sal WLAN, this corre sponds to a de scendan t verti cal HO; Mo bile Tele com mu ni ca tions Sys tem (UMTS)– of fer point b) indi cates the HO between two WLANs, called great geograph ic cover age, but with reduced data rates. hori zon tal HO; point c) illu strates an ascen d ant verti - The handoff (HO) proce dure is the mecha nism by cal HO. which a data con nection in progress between a mo bile ter mi nal and a cor re spond ing ter mi nal is trans ferred from an access point (AP) to an other one, or to a cel lu-

lar network base station (BS). The HO from a WLAN c) WLAN UMTS AP to a UMTS base station should be made with very b) low prior ity, while the HO from a UMTS BS to an AP

should be made whenever a WLAN is avail able, as a re- a) WLAN sult of the increased bandwidth availabil ity in WLANs (Pahlavan, et al., 2000); in addi tion, the uti liza tion cost pa ram e ter must be con sid ered too, as cel lu lar op er a tors Figure 1. HO types: a) Descendant vertical HO, b) Horizontal HO, c) Ascendant vertical HO

168 RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM J. Sánchez-García, Luis A. Villaseñor-Go nzález, Mario E. Vaquera-Flores and Raúl Aquino-Santos

A HO is triggered based on the availabil ity of a network Liu and Zhou (2005) pro pose an internetworking archi - and the prior ity pol icy used to select a network. One of tecture between WLAN and UMTS, based on IPv6, the param e ters used to decide that a HO must take Mobile IP and the 802.16 air inter face (called Hybrid place corre spon ds to the Received Signal Strength In- Couplin g with Ra dio Access System). They propose tensity (RSSI), but this is not enough to give prior ity to two processes, one for verti cal downward Handoff (a a net work dur ing the HO pro ce dure; pa ram e ters like mobile node starts in a UMTS network and goes to a utili za tion cost and avail able bandwidth must also be WLAN) and another for verti cal upward Handoff. They consid ered. There are other methods that can be used compare their archi tec ture with both a tight couplin g to make the HO deci sion, like the one described in and a loose couplin g, showing that their proposal has a (Matusz et al., 2003), where a HO deci sion is based on better perfor man ce (in terms of transmis sion and pro- the throughput and a time hyster esis (in the order of cess ing cost). second s) is consid ered to avoid the ping-pong effect 1; Sasikala et al (2006) pro pose a Handoff techniq ue however, the time they consider to wait for a handoff is from WLAN to UMTS, based on the loose couplin g in- rela tively high and they do not specify the radio chan- te gra tion (in de pend ent ar chi tec ture), aim ing at the re - nel envi ro nmen t empl oyed in their simu l ation s. duction of delay and packet loss when the mobile In this work a metr ic is used to trigger the HO proce - moves from a WLAN bound ary to a UMTS network. dure; this metr ic is based on four param e ters from each The deci sion algo rithm is based on the received signal network: cost, band width, dis tance and RSSI (each of strength, where the stations have to detect a weak re- these param e ters is assigned a dif ferent weight). The ceived signal or failed frame transmis sion. They assume ver ti cal and hor i zon tal HO mech a nism is eval u ated, a hot spot WLAN is served by vari ous Ac cess Points while veri fy ing that packet loss is mini mized and the (APs), where the bound ary APs trans mit a bit indi cat - connec tion is main tained during the HO process. ing that it is possi ble to exit the hot spot cover age. The remain der of this paper is orga nized as fol lows. Their simu la tion results show a re duction in latency A review of recently proposed handoff proce dures is for the op timized algo rithm, compared with the seam - presented in Section II. Section III describes possi ble less techniq ue. ways to in ter connect UMTS and WLAN. The mobil ity Kim and Song (2007) pro pose a ver ti cal Handoff support in IPv4 is pre sented in Section IV. The pro- scheme that uses Mobile IP, aiming at the reduc tion of posed HO al gorithm, in herent problems and the em - the packet loss due to the ping pong ef fect. They as- ployed metrics are described in Section V. Section VI sume that the WLAN is lo cated as a hot spot inside the presents a descrip tion of the sim ula tion model and the UMTS cover age area. The pro tocol is based on the re- simu l ation scenar i os empl oyed. The simu l ation results ceived signal strength of both the WLAN AP and the are presented in Section VII. Fi nally, the con clusions of UMTS base station; before making the Handoff deci - this work are presented in section VIII. sion, it veri fies if the user is a pedes trian or if it is travel - ing on a motor ized vehi cle. To decrease the ping pong II. Related work effect, the sig nal strength thresholds can be adjusted dy nam i cally. Sattari et al. (2004) pro pose a network archi tec ture for a The proposal presented in this work differs from the seamless handover between WLAN and UMTS, based previ ous ap proaches in the way the threshold variable on two al ready ex ist ing mo bil ity pro to cols, Ses sion Ini- (met ric) is calcu lated ; the proposed algo rithm takes into tia tion Pro tocol (SIP) and Mo bile IP. They define a ac count four pa ram e ters: Hand over En tity, which is a set of functions located within the UMTS network, as the respon sible for - Cost of the service Hand over Man age ment. They pro pose three so lu tions: - Available bandwidth Mobile IP based, SIP based and a Hybrid based solu tion. - Distance to the AP (or BS) Although they ana lyze the problems posed by each so- - Signal strength lution, they do not provide any simu la tion results. III. Inter con nec ting a WLAN and an UMTS Network

1 Several unne ces sary HOs are trig gered when the thres hold is The transmit power is one of the main param e ters that crossed several times during a short time. are world wide regu lated , and the transmit ter power

RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM 169 Handoff Between a Wire less Local Area Network (WLAN) and a Wide Area Network (UMTS)

varies among differ ent region s. Many of the IEEE (SGSN) and the Gateway GPRS Support Node (GGSN). 802.11 wire less man u fac tur ers have se lected 100 mW The SGSN mon itors the user lo cation and provides secu - as the trans mit power level by defaul t. In Eu rope 100 rity and access control functions. The GGSN pro vides mW is the highest power level, while in North America routing infor ma tion of the con nected users to the Packet the highest power level is 1000 mW (O´Hara et al., Switched (PS) net work and it provides inter connec tion 1999). The transmit power val ues used in base stations with exter nal PS networks. (BS node in UMTS), range from 30 dBm to 21 dBm ac- The tele com mu ni ca tions tech nol ogy trend is to- cording to the 3GPP Rec ommen da tion (3GPP TR 25. wards an IP world, for this purpose the 3GPP release 5 942 V3. 1.0, 1999). proposes an archi tec ture as shown in figure 2. The for- UMTS is a third gen era tion wireless system (3G) eign agent (FA) (or home agent, HA) function al ity will de vel oped in Eu rope and con tem plated in IMT-2000 be placed in side of the GGSN; the GGSN will send (In ter na tional Mo bile Tele com mu ni ca tions 2000), this is a Agent Adver tise ments when it receives a Packet Data 3G stan dard that in cludes sev eral tech nol o gies, from Proto col (PDP) request from a mobile equipment. An- which the dominants in the market are UMTS and other proposal is that the SGSN and GGSN can be com- cdma2000 (devel oped in the USA). bined in a node called Internet GPRS Support Node This work is based on UMTS TDD, as this was the (IGSN); the IGSN will act like a FA and will provide a model available in the simu la tion tool NS-2. One of the tunnel to transfer packets from, and to, the user HA advan tages of TDD is the fact that the transmis sion (3GPP TR 25.923 V3.0.0, 2000). rates can be dynam icall y adjusted in the uplink and As can be appre ci ated from the 3GPP recom men da - downlink, something that can not be done in FDD tions, the ba sis for the next gener ation net work with IP mode because the links are symmet ric. If the TDD conver gence have been estab lished ; the 3GPP release 6 downlink needs more slots and the uplink has avail able speci f ies the general proce d ures for the partic u la r case slots, the avail able slots may be transferred from the of the HO between WLAN and UMTS. In the present uplink to the downlink. work, the IGSN and Uni ver sal Terres trial Radio Access (UTRA) were consid ered as a unique entity (figure 3). a) UMTS Network architecture

GPRS is the packet core network of UMTS. It includ es two network nodes: the Serving GPRS Sup port Node

HSS Uu P I

Iu e R n NC o N

N b S S Nod k G e B G c G MN S a B Package Core Network UTRA CN Figure 2. UMTS architec tur e release 5

TS UM

Figure 3. UTRAN and IGSN as a unit

170 RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM J. Sánchez-García, Luis A. Villaseñor-Go nzález, Mario E. Vaquera-Flores and Raúl Aquino-Santos

b) UMTS and WLAN interconnection points Ra dio Net work Con trol ler (RNC) em u la tor be tween the WLAN and SGSN. In the sec ond inter connec tion One of the main objec tives in the WLAN and UMTS point (known as loose coupling ) the WLAN is connected inter connec tion is to avoid, as much as pos sible, as a rout ing area through the Gi inter face; where the changes at the physi cal and link layers. This will in sure IWU is based in a SGSN emu la tor between the WLAN that the cur rent networks will keep working, without and GGSN. In both cases (1 and 2, in figure 4), UMTS re quiring the user to make changes at these lower lay- will be the master network and WLAN the slave net- ers; the approac h proposed in this work is based on work, this means that UMTS man ages the mobil ity modi fi ca tions at the network and upp er layers. and secu rity issues (Prasad et al., 2003). All the traffic There are two config u ra tion scenar ios for the inter - will first reach the SGSN, or GGSN, before it ar rives to connec tion of UMTS and WLANs; the first case is its desti na tion, even though the final desti na tion is in- when the cellu lar oper a tor is the adm inis tra tor and side the WLAN, this is an incon ve nien t, because it may owner of the WLAN, the second case is when the Wire- become a bottle neck in the UMTS network. less Internet Service Provider (WISP) is the owner. In In the third inter con nec tion method, the archi tec - the first scenario , the cellu lar op er ator has billing and ture is based in the Virtual Access Point (VAP) which authen ti ca tion mecha n isms that all ow him the man - reverses the function played by the compo nents in the agement and to tal control of the WLAN. In the case last two cases. Here, the WLAN is the master network where the WLAN is op er ated by WISP, both op er ators and the UMTS is the slave. The mobil ity is man aged will have their own billing and authen ti ca tion mecha - accord ing to the WLAN and the recent draft Interaccess nism (Salkintzis et al., 2002). Point Pro to col (IAPP) which has been speci fied to stan- Figure 4 shows the five pos sible op tions for the in- dard ize the com mu ni ca tion be tween APs over the ter connec tion points between a UMTS and a WLAN wired in ter face (Pahlavan et al., 2000), but the archi tec - (Prasad et al., 2003). The inter con nec tions (1) and (2) ture is not scalable because the VAP must always be as- will always have inter ac tion between the AP (via the so ci ated with the Ex tended Ser vice Set (ESS) of the Interworking Unit, IWU) and the UMTS Core Net- WLAN. The tran sit technol ogy (when roam ing) that the work (CN of PS UMTS); the in tercon nec tion is possi - WLAN sees is between differ ent APs in the ESS and the ble trough the SGSN and GGSN respec tively. VAP is consid ered as another WLAN AP. From the In the first inter con nec tion point (known as tight WLAN point of view, UMTS appears as a BSS or cou pling) the WLAN is connected as a UMTS cell, like picocell asso ci ated with another AP (VAP, in this case). any other Radio Access Network (RAN), through the In the UMTS air inter face, ev ery packet will have dou - Iu in ter face; where the Interworking Unit (IWU) is a ble UDP/TCP and IP/PPP headers. Moreover, there will

WLAN IEEE 802.11 Server Router

Servidor Enrutador AP 5 MG VAP 4 Host IWU 3 1 2 Iu Gi 5 AP G M 4 VAP Host MN IWU 3

2 1

Iu Gi Iu Gi RN N No C N S do B S G SG G UTRA CN of PS UMTS Red UMTS

Figure 4. Intercon nec tion architec tur es between WLAN IEEE 802.11 and UMTS

RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM 171 Handoff Between a Wire less Local Area Network (WLAN) and a Wide Area Network (UMTS)

be a MAC 802.11 header in all the head ers re lated with - Cor re spon dent Node (CN). A node that commu ni - UMTS. This in ter con nec tion is con sid ered in ef fi cient, cates with a mobile node (MN). It can be mobile or due to the exces sive use of headers and it is not clear station ary. how will the VAP oper ate with the 802.11 AP (Prasad et - En cap su late. Con sist in adding a header to the origi - al., 2003). nal IP packet (i.e. packets are encap su lated ). The fourth in ter con nec tion method in tro duces a - Tun nel. A de livery path between two nodes where mob ility gatew ay (MG) between the UMTS and WLAN the packets payload are origi nal pack ets. networks. The MG is a proxy imple mented in the - Home Address . An IP address that is assigned for an UMTS or WLAN side. Its purpose is to manage the mo- extended period of time to a MN. bility and routing problems. But the real reduc tion of - Care-of-ad dress (COA). An ad dress as so ci ated packet headers (over head) by employ ing this archi tec - with a MN while it vis its a foreign link. ture is not signif i cant, due to the need for pro prietary - Home Agent (HA). A router of the MN where it contro l proto c ols in add ition to transit between diff er - has regis tered its cur rent COA. ent technol o gies. - For eign Agent (FA). A router at the mobile node’s Another proxy incon ve nien ce is the poor perfor - visited network. mance due to increased latency in the client server com- muni ca tion tra jectory and the fact that it is not easy to b) Mobile IP Function a lity know the num ber of proxies needed to achieve an opti - mum per for mance. The dif ferent tasks exe cuted by the mobile nodes dur- The fifth inter connec tion method is based on mo- ing the HO process are illus trated in figure 5. bile IP to handle the mobil ity man agement problems (known as no cou pling). Here, UMTS and WLAN are - The mobile agents ad vertise its presence via Agent peer to peer networks and the same problems exist as in Ad ver tise ment mes sages. An im pa tient MN may op- the pre vi ous in ter con nec tion method (4); fur ther more, tion ally so licit an Agent Ad ver tise ment message. the refer ence point that makes it possi ble to con nect - A MN receives the Agent Adver tise ments and deter - UMTS, via the GGSN, to the IP network op era tor or mines whether it is located at its home net work or at Internet is also the Gi in terface (Salkintzis et al., 2002). a foreign net work. Since IP is in the core of actual and future wideband - Packets addressed to the mobile node arrive to the networks, this work employs this type of inter con nec - home net work via standard IP routing (step 1). tion method (i.e. num ber 5). - When a MN moves away from its home network, it gets a COA at the foreign net work, either by re quest IV. Mobi lity support in IPv4 or by listen ing to an Agent Adver tise ment; by con - tactin g a Dynam ic Host Conf igu ra tion Proto col The mobil ity support at the network layer has the ad - (DHCP); or by means of a Peer to Peer Proto col (PPP). vantage of being in depend ent from the link layer. Mo - While the MN is away from the home network, the bile IP 2 im ple ments the re quired func tion al ity to pre- MN regis ters every new COA with its HA. vent the discon nec tion of ongo ing comm uni c ation ses - - While the MN is at a foreign net work, the packets sions during the handoff proce dure of a mobile node are inter cepted by the HA, en capsu lated and tun- (MN); the mobile node is always iden tified by its home neled to the MN us ing the COA (step 2). address, in depend ently of its current point of attach - - The encap su lated pack ets arrive to the FA, the exter - ment to the Internet. While the mobile node is outside nal IP header is removed and the origi nal pack ets are of its home IP sub-network, it can be asso ci ated to a forwarded to the MN (step 3). care-of-address (COA); the COA indi cates the current - The data packets gener ated by the MN are routed by MN’s lo cal iza tion. The fol low ing sub sec tions ex plain means of stan dard IP routing; this is the model em- with more detail the compo nents and function al ity of ployed in this work. Mo bile IPv4.

a) Mo bile IPv4 Ter mi nol ogy 2 IP and IPv4 are used inter chan geably throug hout the docu ment. 3 The terms, and most of the func tio na lity of the proce dures that This section pres ents a descrip tion of the termi nol ogy are presented in this section, are consis tent with the termi no - employed in Mobile IPv4 (MIP)3. logy employed in the early paper by Perkins (2002).

172 RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM J. Sánchez-García, Luis A. Villaseñor-Go nzález, Mario E. Vaquera-Flores and Raúl Aquino-Santos

- In the reverse direc tion, pack ets sent by the MN are proce dure, as the MN gets tem porall y discon nected gen er ally de liv ered to its des ti na tion, not nec es sar ily during the period when the Agent Adver tise ment mes- passing through the home agent (step 4). This is sages stop being received, to the time when the MN known as tri an gu la tion. gets con nected to another router; it is also possi ble that the MN receives messages from several agents and may c) MIP message flow during a HO pro cedure get con nected to the less appro pri ate. As a result, the MN can improve the HO proce dure by em ploying addi - The sequen ce of messages in terchan ged between the tional infor ma tion when ever this may be avail able at MN and the HA (via the Foreign Agent) are shown in the MN, for exam ple, by using the infor ma tion pro- fig ure 6. vided from the physi cal layer (Perkins et al., 1996; 2003). d) Route op ti mi za tion This work consid ers that the MN has an Appli ca - tion Pro gramma ble In terface (API) to know its po sition To avoid the trian gu la tion, it is possi ble to employ a and deter mine the distance to its current agent (by route op timi za tion mecha nism ; the idea is that the cor - means of a Global Posi tion ing System, GPS), and also to respon dent node (CN) will be able to directly send en- sense the power asso ci ated with the pack ets received capsu lated pack ets to the mobile node by employ ing from the adver tis ing agents. It is also consid ered that the MN’s COA; this is achieved by means of a binding the MN can receive packets from differ ent base sta tion update message sent by the MN to the CN. The route with the same inter face. Further more, it is assumed op ti mi za tion ap proach re quires changes in the pro to col that Agent Ad vertise ment messages pro vide the cost stack of the CN. The route op timi za tion mecha nism is in for ma tion, num ber of us ers and net work type. not em ployed in the actual MIP model used in this Each time a MN receives a mobil ity Agent Adver - work. tisement message, it veri fies if the agent is already in - cluded in the List of known mo bility agents, if it is not, it is e) Mobil ity detec tion at layer 3 added to the List. Then, the param e ters (i.e. RSSI, Dis- tance, Network Type, Network cost, Network users) The mobil ity agent (e.g. the HA or FA) peri od i call y are up dated and stored. The consid er ation of these pa- broad casts Agent Ad ver tise ment mes sages, con se- rame ters (except Distan ce) in the deci sio n for calcu lat - quently the MN has the oppor tu nity to verify if it is lo- ing a met ric and choosing the preferred agent is one of cated at its home network or if it has moved to a for- the main con tribu tions of this work. The Agent Ad ver- eign network. The HO mecha nism is triggered by this tisement messages are gener ated ev ery second by the type of message. agents, and whenever the agents receive an Agent Re- When the MN detects that it has moved to a differ - quest message. When the MN detects that it returns to ent network (i.e. a foreign network) it gets a new COA its home network (by receiv ing an Agent Adver tise - by using the infor ma tion provided in the Agent Adver - ment from its HA), the MN informs to its HA that it tisement messages sent by the mobile agents (e.g. the must stop inter cept ing packets addressed to it; conse - FA); once the MN has acquired a COA, it pro ceeds to quently data pack ets sent to the MN by a CN will get regis ter the new COA with the HA. It should be noted routed directly to the MN within the home network. that several data packets may be lost during the HO From this point on, the MN uses its home address as

Mobile node Foreign agent Home agent

Agent request H ome Foreign agent agent Agent advertisement Mobile Node Register request Register request Corresponding node Answer register Answer register

Figure 5. MIP Packet Flow Diagram Figure 6. MIP packet flow during a HO proce dure

RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM 173 Handoff Between a Wire less Local Area Network (WLAN) and a Wide Area Network (UMTS)

the lo cal iza tion iden ti fier in stead of the COA. With the infor ma tion to make a deci sion: to change, or not, to help of the MIP param e ters described in this section, the UMTS network. The connec tion proce dure is the MN can get a list of pos sible mobile agents to use, started to acti vate the UMTS wireless interface card. and with the aid of a HO al gorithm deter mine which is 3)The HO algo rithm in the MN de cides to disas so ci ate more con venien t to use. The next section expla ins from the WLAN and asso ci ate with the UMTS some HO algo rithms, includ ing the HO al gorithm pro- network. posed for this work. 4)The MN gets in touch with the foreign agent (FA) and ob tains a COA. V. Handoff algo rithm 5)The HA in the WLAN is informed about the new COA and begins the inter ception of datagrams ad- a) Handoff Proce dure between WLAN and UMTS dressed to the MN, it en capsu lates them and sends them to his FA. The FA then deencapsulates the The Handoff (HO) pro cedure em ployed by a MN to datagrams and deliv ers them to the MN. switch between WLAN and UMTS net works is il lus- 6)If the MN moves into the WLAN cover age, it begins trated in figure 7 (Prasad et al., 2003). a reverse HO. There are four possi ble actions that may be taken by the MN as a re sult of the HO al gorithm: While op er at- b) Handoff and the Ping-Pong ef fect ing in the WLAN, the HO al gorithm will select op tion A given that it re ceives a good signal strength from its Handoff is the mecha nism by which a data con nection current asso ci ated AP; option B in dicates that the MN in progress between a mobile termi nal and a corre - is begin nin g to receive a poor signal strength from the sponding termi nal is transferred from a connec tion AP and should pre pare to start a HO; option C is se - point to an other: for exam ple, as the MN moves away lected when a handoff to the UMTS network is re- from a base station, the signal level gets weak and a quired and there is not another WLAN AP to be se- handoff proce dure must be exe cuted to switch to an - lected; option D is se lected to trigger a HO proce dure to other base sta tion. The signal level degra da tion is a ran - se lect a new WLAN AP. dom process and a simple deci sion mecha nism which is The fol lowing scenario depicts a MN moving away only based on the signal power mea surement may re- from WLAN cover age and moving into UMTS cov er age: sult in the so called ping-pong effect (Pahlavan et al., 2000). The ping-pong effect refers to several con secu - 1)Initiall y, the MN receives a strong sig nal from the tive back ward and forward HOs; this hap pens when AP, and the MN is connected to the WLAN network. the signal power is varying around the threshold 2)As the MN moves away from the AP, the received value used within the HO algo rithm to de cide when signal strength becomes weaker. The MN is contin u - to make a HO. This is an in conve nien t for the user, ally ana lyz ing the re ceived sig nals, looking for a due to the switching cost and the induced delay in better AP or BS. The HO algo rithm uses the avail able each handoff.

Working in WLAN Working in UMTS IEEE 802.11

HO algorithm Other Is near No better No A B C D WLAN IEEE UMTS? Prepare HO other 802.11? WLAN Yes Yes Ready for UMTS? HO algorithm Yes Poor Good Good UMTS No Wait Prepare HO other Prepare UMTS HO to HO to 802.11 UMTS

Figure 7. HO Procedure between WLAN and UMTS

174 RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM J. Sánchez-García, Luis A. Villaseñor-Go nzález, Mario E. Vaquera-Flores and Raúl Aquino-Santos

c) Problems during the Handoff - dur ing a pre de ter mined re ten tion time (Dwell-timer) the new agent metric (metr ic_NA) minus an hyster - The de ci sion mech a nism or HO con trol can be located esis margin is greater than the metric of the actual or in a network entity (like a cellu lar base station) or at older agent (metr ic_VA) (Eq. 2) and, the mobile node; these scenar ios are called HO con - - the actual agent metric (metr ic_VA) is smaller than a trolled by network and HO controlled by mobile, re- pre de fined thresh old (Um) value. spectively. In GPRS, the infor ma tion sent by the MN can be em ployed by the network entity to make the The agent metrics are calcu lated by taking into ac - HO deci sion; this is called HO assisted by the mo bile. count several para me ters, accord ing to the foll owing In this work, the deci sion mecha nism for HO is im ple- equa tion: mented at the MN. This work em ploys the signal power, distance, net- met ric=wight_Cost(1–Cost) work cost and avail able bandwidth param e ters; fur- +weight_Band Width(BW_user) (1) thermore, it uses a retain time counter and a hyster - + weight_Dis tance(1–Dis tance) esis margin (Hm), which is not defined as a received + weight_RSSI(RSII) signal power margin, but as a metric mar gin result ing from all the pre vious param e ters. It also uses a met ric A new agent (NA) is chosen if at the end of Dwell-timer threshold (Um). the fol low ing in equal ity holds true:

d) Handoff al go rithm met ric_NA–Hm>met ric_VA

Almost all known al gorithms are based in the received Um5 is a threshold which is used to deter min e if the signal power (Pollini, 1996) and were designed for HO metr ic_VA is poor. between the same types of net works (mostly for cellu - In equation , the variables denoted with the prefix lar net works). Other tech niques have recently been “weight” have a value between 0 and 1, and added are proposed , like pattern recog ni tion based on neuronal equal to one. Cost is a value that repre sents the network networks or diffuse logic systems, however these algo - charge (from 0 to 1). BW_user is the bandwidth that rithms in crease the compu ta tiona l cost in the mobile corre spond s to the MN, which is calcu lated by divid ing node; in addi tion, these techniq ues still emplo y the the total bandwidth used in the network by the num- RSSI para me ter. ber of MNs in that network. The variable Dista nce is This work follows a prior ity strategy for the HO be- the normal ized distance that exists between the MN tween WLAN and UMTS, and differ ent pri ori ties, cost and the BS or AP. Final ly the RSSI variable is the values and num bers of mobile nodes were pre-estab - strength in dica tor from the received signal . To avoid lished durin g simu l ation 4. The al go rithm im ple mented choosing an agent which is received with very low in this model consid ers that UMTS and WLAN have power, a RSSI_Threshold is defined , such that the agent dif fer ent trans mis sion sig nal strength lev els, then it met ric (i.e. metr ic_NA or met ric_VA) is set to zero when does not make sense to use only the RSSI param e ter to its strength is below the thresh old: decide among them (UMTS BSs gener ally transmit with a higher power strength than WLAN APs). If only if RSSI£RSSI_Thresh old then metr ic=0 (3) this param e ter is used, the MN will always be con- nected to UMTS, even if the MN was inside the WLAN During a HO pro cedure the MN follows the next se- cover age area. The use of a sin gle RSSI param e ter is ap- quence of steps: prop riate for hori zon tal HO. In the proposed HO al gorithm, a new BS or AP will 1)The MN Link-Timer is restarted every 0.5 second s. be chosen (called new agent, NA) if and only if the fol - 2)The HO-trig ger func tion ver i fies if Dwell-Timer is less lowing two condi tions are met: or equal to zero. If so, Monitor-timer is reset

4 The WLAN AP can report how many users are currently 5 During simu lations it was observed that Um delayed the HO connected to it, in order for the MN to roughly esti mate the without any benefit; then a value of 100 was used to speed up avai lable band width; the same criteria can be used in the UMTS. the HO proce dure.

RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM 175 Handoff Between a Wire less Local Area Network (WLAN) and a Wide Area Network (UMTS)

andreturns a HO-Trigger Autho r iza tion . Oth er wise, a) Autoconfigurable mobile node the function returns a HO-Trig ger Deny. 3)If the MN is not asso ci ated to any network and the To simu late a mobile node, a termi nal model with a HO-Trig ger is autho rized, the MN sends a HO-Re- common layer 3 and two differ ent lower layers (one for quest to both, UMTS and WLAN net works. UMTS and another for 802.11) was consid ered. The 4)The base sta tion receives the Agent Request and an- mod ule that performs the inter face between the com - swers with an Agent Adver tise ment. mon layer 3 proto col and the two dif ferent lower proto - 5)If the third step con dition is not fulfill ed, the col stacks (MAC and PHY sublayers) is called RAMON Link-Timer is restarted. (Re con fig ur able Ac cess Mod ule for Mo bile Com put ing Ap pli ca tions). VI. Im ple men ta tion of the sim u la tion model RAMON is a pro ject devel oped at several Italian uni- versi ties, among them the Palermo Univer sity; it con - The UMTS TDD exten sion for the Network Sim ula tor sists of several in terfaces added to the origi nal ns-2 MN (NS-2) was devel oped at the Uni versity of Roma model, as can be seen in fig ure 8. The primi tives sent (Vacirca et al., 2003) using the ns-2.1b9a distri bu tion, between layer 3 and layers 2 are han dled transpar ently which had to be ported to the ns-2.1b7a version of by means of adap ta tion layers related to UMTS and NS-2, as we had an available work in progress from sev - 802.11, respec tively. The RAMON node model makes eral univer si ties regard ing the HO between UMTS and use of two submodules: the Ra dio Resources Control WLAN from which we started our work. The imple - (RRC) and the Mobil ity Manage ment (MM), as illus - menta tion of the sim ula tion model was done as part of trated in figure 8. To cal cu late the at ten u a tion of the a thesis devel oped at CICESE (Vaquera, 2004). radio sig nal, the model employs the Friss free space prop - The UMTS-TDD exten sion manual for NS-2 a ga tion model (1/r2 where r is the dis tance) for short (Todini et al., 2003) (pre lim i nary ver sion), pro vides a dis tances and an ap prox i ma tion to Two Ray Ground de scrip tion and enough in struc tions to un der stand its (1/r4) in long distances. The approx i ma tion assumes im ple men ta tion and how to de velop sim u la tions. that flat earth reflection occurs.

Figure 8. Recon fi gu rable node with two interfaces in NS-2

176 RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM J. Sánchez-García, Luis A. Villaseñor-Go nzález, Mario E. Vaquera-Flores and Raúl Aquino-Santos

The rout ing pro to col em ployed for this in fra struc ture – At t = 30.0, the MN changes its direc tion and moves network is NOAH (no Ad-Hoc) (Widmer, 2004). The towards the other WLAN (AP1), thus exe cut ing a cost and number of users can be estab lished for every hori zon tal HO. The second WLAN is denoted by network; the bandwidth of every network is di vided node 6 in fig ure 10. between the number of users of that specific network. – At t = 50.0, the MN changes again direc tion to wards The node is autoconfigurable and before sim ula tion the the UMTS to perform an ascen dan t verti cal HO user can de fine sev eral pa ram e ters like weight_Cost, passing from WLAN to UMTS. The UMTS BS is rep - weight_Dis tance, weight_Band Width, weight_RSSI, the resented by node 3 in . num ber of ver i fi ca tions (num_checks) be fore ex e cut ing a HO pro cedure to avoid the ping-pong effect, and the During sim ula tion, the MN opens a TCP connec - RSSI_Thresh old that will be used to trigger the HOs. tion by means of an FTP appli ca tion; packets are sent The RSSI and RSSI_Thresh old pa ram e ters were in cluded from the FTP server (i.e. the corre spon dent node de- into the RAMON model and are part of the contri bu - noted as node 1 in figure 10) towards the MN. Data tions of this work. trans fer be gins at t = 0.2 sec. There are two in terme di - ate nodes repre sent ing the Internet cloud, as illus trated in figure 10; these two routers are denoted with the b) Network scenario emplo yed for HO node num bers 0 and 2. All the nodes have an asso ci ated hierarchical address. The simu l ation scenari o consid ers a topol ogy com- The Agent Adver tise ment messages are broadcasted posed by one UMTS and two WLAN networks; the every second , and the reset time of the mon itor ing, link home agent (HA) is located at the UMTS network BS and connec tion tim ers is 0.5 sec. (the NS-2 UMTS mod ule only sup ports the home net- work to be located at the UMTS network) and the for - VII. Results and numeric anal y sis eign agents are the routers from the WLANs, as it is il- lustrated in figure 9. The simu l ation begins with the The mea sured pa ram e ters in the sim u la tion were: MN con nected to the UMTS net work and moving to a point to point delay, number of packets lost and HO WLAN network with a constant speed of 3 km/h de lay. In ad di tion, this sec tion pres ents a com par a tive (0.8333 m/s). The sim ula tion time is 70 sec. and at t = anal y sis be tween the re sults ob tained for ver ti cal (as- 0.1 sec. the MN (node 4 in the ani ma tor of NS-2, figure cend ing and de scend ing) and hor i zon tal HOs. 10) begins to move to wards the first WLAN access point (AP0 or node 5). The follow ing paragraph s de- scribe the movements made by the MN, which are il- lustrated in figure 9:

NC – The MN per forms a descen dan t verti cal HO crossing UMTS 0.1.0 HA from UMTS to WLAN. 2.0.0

0.0.0 MN Secondary 2.0.1 routers of Internet FTP server HA 1.0.0 MN UM TS NC AP1 AP0 FA FA WLAN 4.0.0 3.0.0

23 21 Router Router HA FA 2 WLAN 2 FA WLAN AP0

AP1 Figure 9. Mobility scenario with one UMTS and two WLANs Figure 10. NAM visua li za tion screen

RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM 177 Handoff Between a Wire less Local Area Network (WLAN) and a Wide Area Network (UMTS)

a) Sim u la tion pa ram e ters lost6 in any of the sim ulated HO scenar ios, except when the oper a ti on was und er abno rmal cond ition s; Fig ure 10 il lus trates the Net work An i ma tor (NAM) vi- for exam ple a person runnin g at a speed of 20 m/s (72 su al iza tion of the net work to pol ogy im ple mented for km/h), and carry in g his mobil e termi n al. Perfor m ance simu la tion in NS-2. The lower section of figure 10 re sults for this last scenario are not presented in this shows the gen erated traf fic between the agents and the work. mobile node (MN). Ta ble 1 shows the pa rame ters emplo yed for simu la - 1. Handoff from UMTS to WLAN tions; these values were obtained from (Vaquera, 2004) and repre sent a typ ical exam ple, where the most im- The HO delay can be visu al ized in table 2; the delay to por tant pa ram e ter is con sid ered to be the as so ci ated complete the HO is mea sured from the time the HO is network cost defined by the weight_Cost parameter. triggered, until the time the MN finds itself con nected to the new network. Table 1. Simu la tion Para me ters Figure 11 shows the results obtained for the delay of the packets from the CN to MN (end to end); this fig- Parameter Value ure illus trates the delay before and af ter the HO is made between UMTS and the first WLAN. It is ob- num_checks 1 served that the delay is larger when the MN is con- weight_Cost 0.7 nected to the UMTS, than when it is con nected to the WLAN. weight_Distance 0.1 The received sig nal strength from the UMTS BS and weight_Bandwidth 0.1 the WLAN AP0 are shown in figure 12 and fig ure 13, re- weight_RSSI 0.1 spectively. Figure 14 shows the re sults of the RSSI that the MN mea sures from the UMTS BS; this in forma tion RSSI_Threshold –79.4 dBm is presented as a function of the dis tance between the BS and the MN. An overlap is observed in (i.e. between distances 64 and 70 meters), this is due to the fact that b) Per for mance re sults the MN reaches a maxi mum distance at around 69 mts. and then it begins to come back closer to the UMTS BS. In this sec tion we present the perfor man ce results for the point to point delay and the HOs delay. During nor mal op er a tion con di tions there were no pack ets 6 Packets trans mitted from the BS, or the AP, to the MN.

Table 2. HO Delay from UMTS to WLAN

Event Time State Event 23.720026 Connect HO Trigger from distance 57.156366 23.720026 HO HO Start from BS 2.0.0 23.768811 HO Sent MIPT_REG for: 2.0.0 23.768811 HO Sent MIPT_REG for: 3.0.0 23.784171 HO Received REG_REPLY 23.784171 HO HO Execution: from 2.0.0 to 3.0.0 23.784171 HO Detach from 2.0.0 23.784171 IDLE Attach to 3.0.0,MH has new coa 3.0.0 23.801215 Connect Sent MIPT_REG for: 3.0.0 Total --- 0.081189 HO total time

178 RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM J. Sánchez-García, Luis A. Villaseñor-Go nzález, Mario E. Vaquera-Flores and Raúl Aquino-Santos

Handoff simulation (from 0 to 27 second) from UMTS to AP0 WLAN UMTS RSSI 0.7 -46 Connected to UMTS -48 0.6

-50 s

d 0.5 n o c

e -52 m s B

d n

i 0.4

n i y

I a -54 l S e S D 0.3 R -56

0.2 -58 Connected to AP0 WLAN 0.1 -60

-62 0 0 10 20 30 40 50 60 70 0 50 100 150 200 250 300 350 400 450 Time in seconds Packages

Figure 11. Simu la tion results of the packet delay metric Figure 12. RSSI of the UMTS BS

AP0 WLAN RSSI UMTS RSSI -78.5 -46

-48

-50

-79 -52 m m B B d d

n n i i

I

I -54 S S S S R R -56 -79.5

-58

-60

-80 -62 15 20 25 30 35 40 45 50 55 60 40 45 50 55 60 65 70 Time in seconds Distance in meters

Figure 13. RSSI of the WLAN AP0 Figure 14. RSSI of UMTS BS as a function of distance

Figure 16 shows the graph of the RSSI mea sured by 2. Handoff between WLANs the MN from the WLAN AP1, dur ing simu la tion.

The end to end delay of data pack ets sent from the CN 3. Handoff from WLAN to UMTS to MN is shown in figure 15; the delay measure ments are taken between t1 = 43.0 and t2 = 49.0 to include the The packet end to end delay (i.e. CN to MN) is hori zon tal HO proce dure between AP0 and AP1. Table shown in figure 17; these results were obtained some 3 shows the HO de lay mea sured from the time the HO second s before and af ter the HO from the second AP is triggered to the time when the MN is connected to WLAN to UMTS. Table 4 shows the HO delay mea - the new net work. sured from the time the HO is triggered to the time when the MN is connected to the UMTS again.

RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM 179 Handoff Between a Wire less Local Area Network (WLAN) and a Wide Area Network (UMTS)

Handoff simulation (from 43 to 49 second) from AP0 WLAN to AP1 WLAN AP1 WLAN RSSI 0.12 -78.5 HO Connected Connected to AP0 WLAN to AP1 WLAN 0.1 s d n o c e s

0 .08 -79 n i

m y a B l d e

n D i

0.06 I S S R

0.04 -79.5

0.02

0 -80 0 100 200 300 400 500 600 700 800 900 1000 25 30 35 40 45 50 55 60 65 70 Packages Time in seconds

Figure 15. Packet delay during the horizontal HO Figure 16. RSSI of the AP1 WLAN

Handoff simulation (from 58 to 63 second) from AP1 WLAN to UMTS Complete simulation of the connection handoff (70 seconds) 0.7 0.7 Connected Connected Connected to UMTS to UMTS to UMTS 0.6 0.6 s s d 0.5 d 0.5 n n o o c c e e s s

n n i 0i .4 0.4

y y a a l l e e D D 0.3 0.3

0.2 0.2 HO Connected Connected Connected to AP1 WLAN to AP0 WLAN to AP1 WLAN

0.1 0.1

0 0 0 100 200 300 400 500 600 0 1000 2000 3000 4000 5000 6000 7000 Packages Packages

Figure 17. Packet delay during ascendant vertical HO Figure 18. Packet delay during simula tion

4. Com par i son be tween HOs The packets that went through the UMTS BS, when the MN was connected to that agent, are the most de- This section pres ents a compar i son and analy sis of the layed. Since the WLANs effec tive bit rate is larger than HO results derived by simu l ation ; perfor m ance results that of UMTS, it was found in the sim ula tions that the are eval uated in terms of the point-to-point packet de- pack ets suffer less delay when transmit ted through the lay and HO de lay. WLAN APs.

Point-to-point packet delay

Figure 18 shows the delay of the packets (packet size was set to 1000 bytes) during the entire simu la tion.

180 RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM J. Sánchez-García, Luis A. Villaseñor-Go nzález, Mario E. Vaquera-Flores and Raúl Aquino-Santos

HO delay pack ets from the new agent while there are pending pack ets (from the old agent) still on the air or ca ble. Table 5 shows the measured delays during the HOs. It The “Detach” op era tion denoted in table 2, table 3 can be observed that the HO delay is larger between and table 4, does not im ply disablin g the inter face, WLANs, where it was expected to be smaller, but this and the MN can still receive packets from the previ - delay is random and is due to the Dwell and Link timers ous BS or AP. employed . This can be confirmed by observ ing the de- Figure 19 presents the graphics obtained for the lay between “HO Start” and the first “Sent measured RSSI at the MN from the UMTS BS, and MIPT_REG” in table 2, ta ble 3 and table 4, which are from the access points lo cated in both WLANs, during equiva lent to 0.048785, 0.164219 and 0.009106 sec. re- the entire simu la tion. The solid line repre sents the spectively. The use of timers can be consid ered as a UMTS signal that MN receives and is present at all problem that reduces the effi ciency, but is nec essary to times, the cross and aster isk lines repre sent the signal s avoid the “ping-pong” effect. of AP0 and AP1 respec tively. It can be observed that the Table 6 shows the HO delay at the transport layer MN receives them with a RSSI greater than -80 dBm, (TCP). It is observed that the MN can receive packets good enough to con nect with the network from which from two connec tions at the same time; this happens it receives the stron ger signal . The verti cal lines in di- when the MN (during a HO) is begin nin g to receive cate the time at which the HOs are triggered.

Table 3. HO Delay from WLAN to WLAN

Event Time State Event 45.284171 Connect HO Trigger from distance 100.184815 45.284171 HO HO Start from BS 3.0.0 45.448390 HO Sent MIPT_REG for: 3.0.0 45.448390 HO Sent MIPT_REG for: 4.0.0 45.574707 HO Received REG_REPLY 45.590894 HO Received REG_REPLY 45.590894 HO HO Execution: from 3.0.0 to 4.0.0 45.590894 HO Detach from 3.0.0 45.590894 IDLE Attach to 3.0.0,MH has new coa 4.0.0 45.600000 Connect Start Monitor Total --- 0.315829 HO total time

RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM 181 Handoff Between a Wire less Local Area Network (WLAN) and a Wide Area Network (UMTS)

Table 4. HO delay from WLAN to UMTS

Event Time State Event 60.590894 Connect HO Trigger from distance 103.447349 60.590894 HO HO Start from BS 4.0.0 60.600000 HO Start Monitor 60.600000 HO Start HO Detection:ho_timer=0 60.600000 HO Sent MIPT_REG for: 2.0.0 60.640328 HO Received REG_REPLY 60.640328 HO HO Execution:from 4.0.0 to 2.0.0 60.640328 HO Detach from 4.0.0 60.640328 IDLE Attach to 2.0.0,MH has new coa 2.0.0 60.698995 Connect Sent MIPT_REG for: 2.0.0 Total --- 0.108101 HO total time

Table 5. HOs Delay

HO Delay (sec.) HO Event 0.081189 UMTS to WLAN 0.315829 WLAN to WLAN 0.108101 WLAN to UMTS

Table 6. HO delay at the trans port layer (TCP)

HO Direction Time From UMTS to WLAN Last package received in MN by part of BS 25.4102 sec First package received in the MN by part of AP0 24.8473 sec Absolute difference 0.5629 sec From WLAN to WLAN Last package received in MN by part of AP0 46.1370 sec First package received in the MN by part of AP1 45.9917 sec Absolute difference 0.1454 sec From WLAN to UMTS Last package received in MN by part of AP1 60.1731 sec First package received in the MN by part of BS 60.1703 sec Absolute difference 0.0028 sec

182 RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM J. Sánchez-García, Luis A. Villaseñor-Go nzález, Mario E. Vaquera-Flores and Raúl Aquino-Santos

UMTS RSSI, AP0 and AP1 WLAN -45 FDD model, as a result, future work will con sider the UMTS HO HO AP0 imple men ta tion of an FDD based UMTS network. AP1 -50 Connected Connected Connected Connected to UMTS to AP0 WLAN to AP1 WLAN to UMTS HO Refe rences

-55 Gumming L., Chi Z. HCRAS: A Novel Hybrid Internet wor - m B

d -60 king Archi tec ture Between WLAN and UMTS Cellular n i

I nd S Networks (2 , 2005) IEEE Consumer Commu nica tions S R -65 and Networking Conference. 2005, 374-379 pp. HyoJin K., JooSeok S. A Novel Vertical Handoff Scheme -70 Based on Mobility Speed in Integrated WLAN and UMTS Networks. IEICE Transac tions on Communi ca tions , E90-B -75 (7):1844-1847. July 2007. Matusz P., Machan P., Wozniak J. Analysis of Profita bi lity of -80 0 10 20 30 40 50 60 70 Time in seconds Inter-system Handovers between IEEE 802.11b and UMTS (28th Annual, 2003, Gdansk, Poland) Procee dings Figure 19. UMTS BS and APs RSSI during simula tion of the 28th Annual IEEE Interna tional Conference on Local Computer Networks (LCN’03). Gdansk, Poland, Oct 20th-24th, 210-217 pp. VIII. Conclu sions O’Hara B., Petrick A. IEEE802.11 Hand book, a Desig ner’s Companion . IEEE Press. First Issue. 1999. 173 p. Dif fer ent al ter na tives for handoff be tween WLAN and Pahlavan K., Krish namurthy P., Hatami A., Yliant tila M., UMTS were ana lyzed. In this work we imple mented a Makela J.P., Pichna R., Vallstron J. Handoff in Hybrid HO al gorithm while trying to re duce complex ity and Mobile Data Networks. Personal Communi ca tions, IEEE, provide the best advan tages (no coupli ng MIP based) of 7(2):34 -47. 2000. other previ ously reported HO ap proaches. The verti cal Prasad R., Muñoz L. WLANs and WPANs Towards 4G Wireless . and hori zon tal HOs are performed without los ing any The Artech House. First Issue. Norwood. 2003. 276 p. ongo ing TCP connec tions and without losing packets. Perkins C.E. Mobile IP. IEEE Communications Maga zine, A novel al gorithm was proposed , imple mented and evalu ated to supp ort hori zon tal and verti cal handoffs 40(5):66-82. 2002. between UMTS and WLAN networks; the HO algo - Perkins C.E., Johnson D.B. Mobi lity Support in IPv6 (Annual, rithm relies on several met rics, includ ing the RSSI. Pre- 1996, New York, USA) Proceedings of the Annual Inter- vious works impl ement vari ou s algo rithms, some of national Conference on Mobile Computing and Networ- th th them even with high compu ta tiona l cost, as those that king, MOBICOM. Rye, New York, USA.Nov. 10 12 , make use of neu ral net works, but still they rely on the 27-37 pp. 1996. RSSI metric. Perkins C.E., Johnson D.B.. Mobi lity Support in IPv6. It was concl uded that the inter con nec tion archi tec - draft-ietf-mobi leip-ipv6-24.txt. 170 p. [on line] 2003. ture between UMTS and WLAN by means of MIP rep - [Consulted on January 29, 2008]. Avai lable on: resent a via bl e alter n ative to enabl e roami ng between http://net.infocom.uniroma1.it/downloads/umts_ networks. The HO delay (with TCP) doesn’t sig nifi - manual.pdf. cantly affect the file transfer services, data and email, Vacirca F., Todini A. UMTS Module for ns. [on line] 2003 but it may affect real time ap plica tions, since it some - [Consulted on January 29, 2008]. INFOCOM Depart - times exceed the maxi m um all owed dela y consid ered ment, Rome Univer sity “La Sapienza”, Rome, Italy. In accept able for voice and video appli cations (500 msec.). version of ns-2.1b9a. Avai lable on: However, if the HOs do not oc cur frequen tly, the in tro- http://net.infocom.uniroma1.it/reti_files/reti_down - duced HO delay may be con sidered not so annoying for loads.htm. voice or video applications. Vaquera-Flores M.E. Handoff between IEEE802.11 Wireless Local The perfor man ce results in this work con sidered a Area Network Access Points Environ ment and a Wide Area TDD based UMTS network; however, recent exten - Network (WAN). Master of Science Disserta tion. CICESE, sions to the NS-2 UMTS model have imple mented the Ense nada, Mexico. August 2004. 117 p.

RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM 183 Handoff Between a Wire less Local Area Network (WLAN) and a Wide Area Network (UMTS)

Widmer L. Extensions to the ns Network Simulator [on line]. 2004 [Consulted on January 29, 2008]. Infrastruc ture Routing Type (no ad-hoc). March 1st 2004. Available on: http://icapeople.epfl.ch/widmer/Mobi leIP/ns-exten sion/ 3GPP TR 25.942 V3.1.0 1999. RF System Scenarios . Issue 1999, 111 p. 3GPP TR 23.923 V3.0.0 2000. Combined GSM and Mobile IP Mobility Handling in UMTS IP CN. Issue V3, 76 p.

About the authors Jaime Sánchez-García. Received the enginee ring degree in electro nics and communi ca tions from IPN-ESIME, Mexico (1976); a M.Sc. in elec tronics and telecom mu ni ca tions from CICESE Mexico (1979) and the D.Sc. in elec trical enginee - ring (major in commu nica tions) from The George Washington University (2001). Since 1979, he holds a research and faculty position at CICESE’s elec tronics and telecom mu ni ca tions department. He won the 1st place in III Ericsson Yearly Award (1988) in México. His publi ca tions include several IEEE arti cles and inter na tional confe rences. Current research interests include wire less and sensor networks, software radio, radio channel modeling and multica rrier CDMA. He is an IEEE and a IEICE member. Luis A. Villaseñor-Go nzález. Received an enginee ring degree in electro nics from UABC, Mexico (1993); M.Sc. in elec tronics and telecom mu ni ca tions from CICESE, Mexico (1997); and PhD in electrical enginee ring from the University of Ottawa in 2002. He is currently a research professor at the CICESE research center. He colla bo rated as a network research engineer at the Communi ca tions Research Centre in Ottawa, Canada. At CRC he was involved in a variety of research activi ties in network techno lo gies for the Government of Canada between 1999 and 2003. His current research interests include Mobile Ad-hoc networks, wireless communi ca tions networks, QoS protocol archi tectures, perfor- mance analysis and evaluation of Internet techno lo gies and computer networks. He is currently a member of the IEEE. Mario E. Vaquera-Flores. Received a M.Sc. in elec tronics and telecom mu ni ca tions from CICESE, Mexico (2004). Raúl Aquino-Santos. Received the elec trical enginee ring degree from the University of Colima (1988); M.Sc. in elec tronics and telecom mu ni ca tions from CICESE, Mexico (1990); PhD degree from The University of Sheffield, England (2004). From 1990 to 1996 he was in charge of The Telecom mu ni ca tions and Networks Department of the University of Colima, and from 1996 to 2000 he was the director of the Faculty of Telema tics. He has imparted several national and inter na tional courses and confe rences related to infor ma tion tech no lo gies, tele com mu ni ca tions, networks and multi- media. He has national and interna tional publi cations, and has supervised several thesis and projects in undergra duate and postgra duate programs since 1990.

184 RIIT Vol.X. Num.2. 2009 167-184, ISSN1405-7743 FI-UNAM