Distributed Resource Reservation for Beacon Based MAC Protocols

Transcription

1 Distributed Resource Reservatio for Beaco Based MAC Protocols Frak Leipold, Jörg Eberspächer To cite this versio: Frak Leipold, Jörg Eberspächer. Distributed Resource Reservatio for Beaco Based MAC Protocols. Fi Arve Aagese; Svei Joha Kapskog. Networked Services ad Applicatios - Egieerig, Cotrol ad Maagemet, 6164, Spriger, pp , 2010, Lecture Notes i Computer Sciece, < / >. <hal > HAL Id: hal Submitted o 20 Aug 2014 HAL is a multi-discipliary ope access archive for the deposit ad dissemiatio of scietific research documets, whether they are published or ot. The documets may come from teachig ad research istitutios i Frace or abroad, or from public or private research ceters. L archive ouverte pluridiscipliaire HAL, est destiée au dépôt et à la diffusio de documets scietifiques de iveau recherche, publiés ou o, émaat des établissemets d eseigemet et de recherche fraçais ou étragers, des laboratoires publics ou privés.

2 Distributed resource reservatio for beaco based MAC protocols Frak Leipold 1 ad Jörg Eberspächer 2 1 Sesors, Electroics & Systems Itegratio EADS Iovatio Works Germay frak.leipold@eads.et 2 Istitute of Commuicatio Networks Techische Uiversität Müche joerg.eberspaecher@tum.de Abstract. Wireless coected devices become icreasigly popular i a large variety of applicatios. Cosumer electroics most certaily is the field with the most wireless iovatios i the past years; but also other areas, such as medical equipmet, vehicular o-board etworks or maiteace services, experiece a icreasig demad for wireless commuicatio. Additioally the etworks should just work ad require as little maiteace as possible. Hece future WLAN ad WPAN must be self-cofigurig, self-healig ad distributed to provide flexible usage. Assumig a homogeeous distributed MAC protocol with a beaco based reservatio mechaism, a radio resource reservatio algorithm is developed to fulfil the delay ad data rate requiremets from the devices. It uses a game theoretic approach to achieve ifrastructure-less desig ad still provides fair resource allocatio. Chages i the radio chael, failig devices or liks ad mobile odes are detected ad a reorgaisatio of resources is calculated. 1 Itroductio The costatly icreasig performace demads for wireless high speed coectios is the motor to further ehace existig techologies ad to ivet ew ad iovative commuicatio iterfaces. For wireless local area etworks (WLAN) several cadidates have bee developed i the past years. The recetly approved IEEE stadard achieves up to 600 Mbits/s, by usig MIMO (multipleiput-multiple-output) techology. Highdataratescaalsobeeasilyachievedbyusigultrawidebad(UWB)as the badwidth of more tha 500 MHz provides plety resources. The WiMedia stadard [1] is capable of 480 Mbit/s with oe sigle chael. The recetly published upgrade eve allows 1024 Mbit/s. This is doe with oe sigle atea (hece o MIMO), which implies less complex trasceivers. Oe sigificat differece betwee WiFi ad high-speed UWB is the trasmit rage. Due to the strict power limitatios of at most dbm/mhz for UWB, defied from the FCC [2], ad mostly followed by other regulatory orgaisatios,

3 2 high data rate UWB has a typical rage of about 10 meters. It also uderlies more itese spatial limitatios give by walls ad other obstacles. This is ofte iterpreted as a hadicap of UWB, but it actually also implies two sigificat advatages. First it reduces the cogestio i the wireless chael. For istace the IEEE b/g techology is curretly ofte used i home applicatios. But the stadard has oly three o-overlappig chaels (respectively four i Japa). Thus i apartmet houses, where most flats have their ow WiFi access poit (AP), the chaels are very crowded as the techology ca achieve 20 to 100 meters trasmit rage eve i idoor scearios. With UWB the radio sigals are almost cofied to the idividual apartmets ad the cotetios for the wireless chaels get relaxed. There will be less wireless systems sharig the commo resources. The secod advatage is the reduced risks of beig eavesdropped. The attacker must be i very close proximity to successfully receive the UWB sigals. Eavesdroppig becomes much more difficult ad the risk of fallig victim to a attack declies. Eve though WiMedia is a fully distributed algorithm, without ay coordiatig odes such as a WiFi-access-poits, the etwork still has service providig access poits, for istace a ode that is coected to a LAN ad operates as a bridge betwee the LAN ad the WiMedia etwork to eable access to the Iteret. These WiMedia (or UWB) APs do ot have ay special role i terms of the PHY or MAC protocol. Usig UWB for WLAN applicatios requires a larger umber of APs. For the office sceario about each room requires at least oe AP, depedig o the size of the room. This makes the etwork maagemet more complex. For ow usually the WiFi etworks are maaged by maiteace persoel, but with a large umber of APs to maage, this becomes a very complicated task. Several automated resource maagemet algorithms for WLAN systems have bee proposed yet. For WiMedia little work has bee doe o this subject. With its completely distributed algorithms the existig approaches ca ot be used efficietly. I a earlier work [3] a iteger optimizatio algorithm was preseted to calculate the optimal AP placemet ad resource allocatio pla for WiMedia etworks. It is very complex ad time cosumig algorithm that will ru durig the desig phase of the etwork. But durig the operatio of the etwork chages to the wireless chael will occur ad a fast adaptatio to the ew coditios must be made. Depedig o the size of the etwork a cetralised optimisatio may take too log. A distributed approach is preferred. Our applicatios target wireless o-board etworks for public trasportatio vehicles, such as aircraft. Cabi maagemet systems, cotaiig readig lights, sigs, speakers, small displays, shall be coected wireless to reduce productio ad maiteace costs ad to make the cabi layout more flexible. UWB curretly is the most promisig techology that provides a very robust radio chael for the aircraft eviromet [4], high data-rates ad o licesig problems. Furthermore the etwork shall have self-healig capabilities ad adapt quickly to failures ad chages. Therefore the resource maagemet algorithm

4 3 should be distributed ad eable the etwork to operate with the best possible cofiguratio eve whe some parts are failig completely. The etwork maagemet of the public trasportatio sceario is ot very differet from a office eviromet or home cosumer electroics. Therefore the gaied results i this paper ca be easily applied to other situatios, where devices must be coected to service access poits. The rest of this paper is orgaised as followed. Sectio 2 describes features of WiMedia relevat to this work. Sectio 3 describes briefly game theory ad some related research. I sectio 4 the distributed resource maagemet algorithm is preseted. Fially sectio 5 cocludes this paper. 2 Uique WiMedia features The WiMedia stadard ad also other possible stadards that use a distributed beaco based MAC implemetatio, has some uique features, which are of value for the resource maagemet. WiMedia is completely ifrastructure-less, meaig all odes have the same physical ad MAC capabilities. Ulike i other commo protocols there is o coordiator or MAC level access poit. The protocol uses periodic superframes cotaiig a beaco period ad a data period; see Figure 1(a). All odes allocate a beaco slot i the beaco period, regardless if they have to trasmit user data or ot. This slot is fix over time ad chages oly i rare coditios. For DRP (distributed reservatio protocol) chael access the beacos are used to aouce trasmissios ad reserve parts of the data period, the so called MAS (media access slots). Coflicts are idetified early ad collisios ca be mostly avoided. The oly collisios may occur i the beaco slot allocatio process. The beacos cotai details of all upcomig trasmissio i the data period. This meas a sigle ode kows whe a eighbour commuicates with aother ode. This feature makes WiMedia ot susceptible to the hidde statio problem ad it does ot eed RTS/CTS messages. With the beacoig a hard limitatio of WiMedia applies. The beaco period has oly 96 beaco slots with two slots reserved for special fuctios. This meas o more tha 94 odes must be withi rage o the same chael. The group of odes withi trasmit rage of a subject ode is called beaco group (BG). Each ode icludes the members of the BG ad the respective beaco slot ids ito its ow beaco; hece it is broadcastig this iformatio to the surroudig odes. Furthermore the stadard also defies the exteded beaco group (EBG), which is the set of odes represetig the eighbour s eighbours. Figure 1(b) shows the BG ad the EBG of a ode. A ewly activated ode picks a beaco slot that is ot occupied by a ode of the BG or from the EBG. This rule implies that for ay give ode the beaco slot ids of the BG members must be uique, otherwise they would cause collisios. For the EBG members the ids ca be idetical, because the subject ode ca ot liste to them, but must ot trasmit at the same time as it would cause collisio at the ode that has the subject ode ad the EBG-ode i its BG.

5 4 (a) Superframe structure (b) Beaco group ad exteded beaco group Fig. 1. WiMedia beaco groups: Figure (a) shows the beaco period ad data period of a superframe. I (b) the relatio of the beaco group ad the exteded beaco group are show. The cosequece of overlappig areas i WiMedia is the extesio of the EBG. Whe a border ode is withi rage of two APs, it will add parts of the secod AP odes to its BG, which agai appear i the EBG of the first AP. Hece overlaps icrease the exteded beaco group size ad therewith limit the maximal ode desity. The beaco group size is directly related to the ode desity ad trasmissio rage as slots may oly be occupied by oe ode. The exteded beaco group size additioally depeds o the activatio sequece of the etwork, as for a sigle ode the eighbour s eighbours beaco slots may be idetical. 3 Game theory for resource maagemet For the distributed resource maagemet algorithm preseted i this work, oe key feature is to have a fair resource allocatio. Nodes should ot act selfishess ad occupy wastefully available resources. The objective of the algorithm is to maximize the overall efficiecy of the etwork. Hece odes must ot oly take their ow situatio ito accout, but also those of the surroudig odes. The game theory is capable of defiig fair ad distributed algorithms. It origiates from ecoomics i the 40 s, but has also bee applied to biology, egieerig, political sciece, computer sciece, ad philosophy. The problem is

6 5 defied i a strategic game with a set of ratioal players. Each player tries to maximise its payoff fuctio by choosig the best strategy ad also by takig the actios of the other players ito accout. Usually the players strategies will result i a equilibrium, for istace the Nash equilibrium, where o player ca chage the ow strategy without decreasig its payoff. May differet types of games have bee developed: cooperative or o-cooperative, symmetric or asymmetric, zero-sum or o-zero-sum, simultaeous or sequetial, just to ame a few. Game theory has bee proposed for resource maagemet i wireless etworks for all kids of techologies. For IEEE the authors of [5] preset a game to share the available radio chaels. The payoff fuctio based o trasmissio delay, chael access legth ad throughput. I [6] a algorithm for OFDM based commuicatio is described that miimizes the trasmissio power, while still achievig the QoS requiremets of the system. For IEEE etworks [7] presets a game defiitio to cotrol the amout of badwidth give to ew coectios, with respect to delay, throughput ad other QoS parameters i the etwork. Eve though the amout of available resource maagemet algorithms based o game theory is huge, o oe hadles a comparable system to that of a Wi- Media etwork. The distributed beacoig mechaism ad DRP chael access scheme is uique. The priority for a WiMedia etwork lies i the efficiet MAS allocatio. To utilise the WiMedia features a ew resource maagemet algorithm is required. 4 Algorithm The resource maagemet algorithm assumes APs with fix locatios. Furthermore the etwork cosists of a wired backboe that coects the APs. The APs provide the service for wireless ed devices to get access to the wired back boe etwork. The MAC mechaism for the WiMedia odes shall be DRP. For the aircraft sceario DRP is essetial, as it provides guarateed resources to the ed devices. The QoS requiremets for ed devices are defied i required data rate ad maximum time delay. Assumig oly oe MAS is reserved per ed device ad take ito accout the 64 ms of a superframe, the delay of a message over the UWB lik ca be early 64 ms. For applicatios with smaller delay requiremets, the resource maagemet algorithm must reserve two or more MAS per ed device with maximum gaps betwee the MAS. The possible delays will be reduced. Despite the metioed DRP reservatio scheme the etwork may also use the alterative chael access PCA (prioritized cotetio access) for o critical applicatios. MAS that are ot completely used by the owig device ca be released ad used for PCA, a CSMA like access scheme. This way the uused MAS sectios ca still be used for applicatios where collisios ca be tolerated. The followig two lists show iput ad output parameters of the algorithm:

7 6 Iput parameters Lik quality Devices per AP BG/EBG size Badwidth utilisatio Delay requiremets Output parameters Chael allocatio MAS reservatio The lik quality is give i RSSI (received sigal stregth idicator) or LQE (lik quality estimator). Both are defied i the WiMedia stadard ad should be accessible from the applicatio layer. It is assumed that a optimal cofiguratio of the etwork is kow. This ca be obtaied from calculatios i the desig phase of the etwork. These calculatios are usually very itesive ad take a log time. Therefore they ca ot be performed i-time durig the operatio of the etwork. Efficiecy i a WiMedia etwork usig DRP ca be defied as the cofiguratio with the miimal used MAS, which agai implies the miimal usage of resources ad maximum remaiig badwidth. I the followig otatio O stads for the umber of owed MAS from ode. A ode ows a MAS whe it has reserved this MAS for trasmissios or receptio i the beaco period. R stads for the umber of reserved beaco slots i the beaco period of ode. This ca be ow MAS, MAS owed by surroudig odes or MAS marked as occupied from alie odes. Furthermore the raised idices o ad e idicate if the value is the optimal solutio, or the effectively curret allocatio. Fially the secod raised idices r or i show if the value must be a iteger or a real. The iteger otatio represets the umber of owed or reserved MAS. The real otatio idicates how much of that MAS will actually be used accordig to the data rate requiremets. Two examples will demostrate the otatio: O o,i are the owed MAS slots by ode for the optimal solutio ad iteger coutig, R e,r are the effectively used MAS of ode i real coutig. The effectiveess of the odes is defied i two differet ways. For the ed devices the effectiveess is: E ED = { O o,r O e,r 0, if data rate or delay requiremets fail (1) A effectiveess of 1 is the optimal solutio; for sub-optimal solutios the umerator icreases ad the effectiveess closes zero. The otatio i real values is used, to get a differece i the equatios for icreased trasmit data rates, but still oly usig oe MAS. For istace a ode with low badwidth requiremets must always reserve at least oe MAS. This ca mea that eve for the lowest possible ad highest possible data rate still oly oe MAS is required. To reward odes switchig to a higher trasmit data rate ad thereby less occupy the chael, a real otatio for the umber of owed MAS is used. A higher data rate E ED will chage the efficiecy of that ed device to the positive. For APs the effectiveess uses the sum of the umber of used MAS (iteger coutig). A AP has to maitai reservatios to the ed devices, which ca

8 7 oly be doe i complete MAS slots. Cosider a ode with low badwidth requiremets, but strict timig delays. Optimally it could operate with two MAS slots. But due to already existig reservatios o combiatio of two MAS might be available to satisfy the timig requiremets. Hece a combiatio of three MAS must be used. If usig the real coutig of MAS, as for the ed devices, o differece of the AP effectiveess would be measurable; the summed effectively used MAS portio would stay the same. But with the iteger coutig the extra MAS is measurable. The effectiveess of a AP is defied as: E AP = R o,d R e,d R o,d, if % of the available MAS required AR e,d 0, if more tha 100% of the available MAS required The parameter A i the umerator for the case of % MAS utilisatio must be greater 1. It artificially degrades the efficiecy ad prevets a AP of beig overloaded ad shifts AP assigmets from ed devices to surroudig APs if possible. The utility fuctio of each ode is the sum of the efficiecy of all odes (ed devices ad APs) withi rage. Each ode tries to maximize the utility fuctio. For a ed device the utility fuctio is: U = E ED For a AP the utility fuctio is: U = E AP + o D + o D E ED o E ED o + p A + p P (2) E AP p (3) E AP p (4) D is the set of ed devices o the same chael that are i rage of ode. I aalogue P is the set of APs o the same chael i rage of ode. Each ode (AP ad ed device) has a algorithm implemeted to calculate the MAS reservatios, depedig o the selected chaels. This algorithm takes the APs as subject odes ad iterates the ed device odes i rage ad o the same chael. The ed device with the earliest delay requiremet will be served first ad gets assiged the closest MAS to or below the delay. This is repeated, util all odes have eough MAS to satisfy the delay, afterwards radom MAS are assiged, to match possible data rate requiremets. Each ode must have at least oe MAS per superframe. To provide the required iformatio of the surroudig odes, the QoS requiremets, eighbourhood relatio ad curretly superframe structure is exchaged o a two-hop distace. Each ed device ca choose which AP i rage it selects. The decisio uses the utilisatio fuctio. A ed device compares the existig utilisatio value ad the to be obtaied value whe chagig the AP by calculatig the MAS reservatios ad utilisatio value of the ew setup. A access poit chage is oly doe, whe a utilisatio gai exists.

9 8 Aalogous the APs also recalculate the MAS reservatio ad utility value, with chagig chael selectio. Fially the processes for the ed devices ad APs of the distributed resource maagemet algorithm based o game theory are as followed: Ed device 1. Sca all chaels for APs 2. Calculate MAS reservatios ad U for chaels with APs i rage 3. Select the chael with the best U 4. Periodically resca other chaels Access poit 1. Select chael (o AP or AP with worst RSSI) 2. Try to icrease U by ivestigatig possible chael chage 3. Select chael with best U 4. Periodically recalculate Mobile odes are also supported. Therefore the optimal solutio is calculated without them. Later, the mobile odes are icluded i the system as a regular ed device. They will leads to a costatly higher utilisatio value, but still the algorithm fids the same fair solutio. 5 Coclusios I this work a resource maagemet algorithm based o game theory for high data rate UWB etworks was preseted. The etwork cosists of service access poits, which provide access for wireless ed devices to a wired etwork, such as LAN or Iteret. The algorithm achieves a fair distributio of resources ad adapts to chages i the radio chael or topology. Network efficiecy is based o the umber of used MAS slots, which also ecompass badwidth ad timig requiremets. The MAS reservatio algorithm is essetial for scearios with high MAS utilisatio, as the efficiecy of the overall algorithm is tightly coupled to the efficiecy of the MAS reservatio for large loads. The distributed approach does ot require a dedicated ode to compute the etwork parameters. Hece whatever parts of the etwork are failig, the etwork still tries to coect as may odes to the APs as possible. The algorithm ca be used i various applicatios: o-board etworks, office ad home WLAN or ad-hoc commuicatio. Refereces 1. ECMA Iteratioal, ECMA-368: High Rate Ultra Widebad PHY ad MAC Stadard - 3d Editio, December FCC, Report ad Order. FCC 02-48, April F. Leipold ad S. Bovelli, Requiremets for Radio Resource Maagemet i Multicell WiMedia Networks, i ICT-MobileSummit, 2009.

10 4. J. Chuag, N. Xi, H. Huag, S. Chiu, ad D. Michelso, UWB Radiowave Propagatio withi the Passeger Cabi of a Boeig Aircraft, i Vehicular Techology Coferece, VTC2007-Sprig. IEEE 65th, April L. Berlema, G. Hiertz, B. Walke, ad S. Magold, Radio resource sharig games: eablig QoS support i ulicesed bads, Network, IEEE, vol. 19, July-Aug Z. Ha, Z. Ji, ad K. Liu, Power miimizatio for multi-cell OFDM etworks usig distributed o-cooperative game approach, i Global Telecommuicatios Coferece, GLOBECOM 04. IEEE, vol. 6, Nov.-3 Dec D. Niyato ad E. Hossai, Radio resource maagemet games i wireless etworks: a approach to badwidth allocatio ad admissio cotrol for pollig service i IEEE , Wireless Commuicatios, IEEE, vol. 14, Feb