Routing and quality of service support for mobile ad hoc networks

Computer Networks 51 (2007) 3142 3156 www.elsevier.com/locate/comnet Routing an quality of service support for mobile a hoc networks Anelise Munaretto a,c, *, Mauro Fonseca b,c a CPGEI/UTFPR, Programa e Pós-grauação em Engenharia, Elétrica e Informática Inustrial, Av. 7 e setembro, 3165, 80230-901 Curitiba (PR), Brasil b PPGIA/PUC-PR, Bloco 2 Parque Tecnológico 2 anar, Rua Imaculaa Conceição, 1155, Prao Velho, CEP-80215-901 Curitiba (PR), Brazil c LIP6/Université e Paris VI, 104 avenue u Présient Kenney, 75016 Paris, France Receive 30 November 2005; receive in revise form 17 December 2006; accepte 28 December 2006 Available online 1 February 2007 Responsible Eitor: V.R. Syrotiuk Abstract OLSR is an optimization over classical link state protocols tailore for mobile a hoc networks. In this paper, we propose the QOLSR protocol which inclues quality parameters to the stanar OLSR. Three variants of QOLSR are introuce, taking into account the elay measurement together with the hop count metric. Then, we analyze new heuristics for the multipoint relay selection, an evaluate our propose approaches comparing them with the stanar OLSR protocol. Simulation results show that an increase loa-balancing an a reuce roppe packets rate are achieve ue to the inclusion of the elay information. Ó 2007 Elsevier B.V. All rights reserve. Keywors: A hoc networks; Routing protocols; QoS support; Delay metric 1. Introuction * Corresponing author. Aress: CPGEI/UTFPR, Programa e Pós-grauação em Engenharia, Elétrica e Informática Inustrial, Av. 7 e setembro, 3165, 80230-901 Curitiba (PR), Brasil. Tel.: +55 41 3310 46 86. E-mail aresses: anelise@cpgei.cefetpr.br (A. Munaretto), mauro.fonseca@ppgia.pucpr.br (M. Fonseca). The highly ynamic nature of a mobile a hoc network results in frequent an unpreictable changes in the network topology, increasing the complexity of routing among noes. Such challenges make routing probably the most active research topic within the MANET area. Besies the challenges associate with mobility, more ifficulties are introuce by the specific characteristics of the wireless channel. Broacast is the basic moe of operation over a wireless channel where, in general, each transmitte message can be receive by all neighbors locate within one-hop from the sener. In terms of traffic classification, in unicast the MAC layer is suppose to filter the messages an eliver them to higher layers those whose aress matches with the noe. When, broacast is use all neighbors that receive 1389-1286/$ - see front matter Ó 2007 Elsevier B.V. All rights reserve. oi:10.1016/j.comnet.2006.12.010

A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 3143 the message will forwar them to higher layers regarless of its aress, while the routing will forwar the messages to other noes. Such broacast traffic is often use by routing protocols for path iscovery. In this case, the simplest implementation of the broacast operation is by naive flooing, which may cause the broacast storm problem ue to reunant re-broacast [1]. Numerous routing protocols an algorithms have been propose. Their performance uner various network environments an traffic conitions have been extensively stuie an compare [2]. MANET routing protocols are typically subivie into two main categories: reactive on-eman routing protocols an proactive routing protocols. Reactive on eman routing protocols establish the route to a estination only when there is a eman for it. In proactive routing protocols routes are calculate before neee. Such protocols can be erive from either legacy Internet istance-vector or linkstate protocols. For mobile a hoc networks, the proactive routing protocols are table-riven, where each noe tries to maintain routing information about every other noe in the network at all times. An example of a proactive protocol is OLSR [3], which is an optimization over the classical link state protocol [4] (e.g. OSPF [5]). OLSR is now officially efine by the RFC 3626 of the IETF [6] in the Mobile A hoc NETworks (MANET) working group. It performs hop-by-hop routing, i.e., each noe uses its most recent information to route a packet. Therefore, each noe selects a set of its neighbor noes as MultiPoint Relays (MPRs) [7]. In the OLSR protocol, only the noes selecte as MPRs are responsible for forwaring control traffic, intene for iffusion to the entire network. MPRs provie an efficient mechanism for flooing control traffic that reuces the number of require transmissions. The MPRs are also responsible for eclaring the link state information over the network. However, no QoS information is taken into account, leaing to a non optimal path selection in terms of QoS requirements. In this paper we propose a QoS-enhancement for the OLSR protocol, which we efine as the QOLSR protocol. The QOLSR protocol extens the stanarize OLSR protocol [3], introucing QoS metrics to wireless an mobile a hoc networks. While the hop istance may be a vali metric for wire an stationary networks, it oes not consier the specifics of wireless links nor noe movement. We introuce three QOLSR variants of OLSR with ifferent traeoffs: QOLSR1, QOLSR2 an QOLSR3. The three variants QOLSR1, QOLSR2 an QOLSR3 are heuristics esigne for the selection of MPRs base on QoS parameters. The heuristic use in QOLSR1 selects as MPR the neighbor noe that can reach the largest number of noes (such a noe will have what we efine as the maximum reachability or egree). If there are two or more neighbor noes with the same reachability, then QOLSR1 prioritizes the neighbor with the smallest elay. The next heuristic is the one use in QOLSR2, which prioritizes the noe with the smallest elay when selecting the MPR noe. If there are two or more neighbors with the smallest elay, then the noe to be chosen as MPR will be the one with the largest reachability. The last propose heuristic, use in QOLSR3, selects as MPR the neighbor noe with the smallest elay among the neighbors that are, at most, within two hops from the initial noe. We emonstrate that QOLSR3 fins the optimal shortest path, in terms of elay, using only partial knowlege of the network topology. The remainer of this paper is organize as follows. A etaile specification of the QOLSR protocol is presente in Section 2, while in Section 3 we present its performance evaluation. In Section 4 we conclue the paper. 2. QOLSR 2.1. OLSR OLSR [3] is a proactive routing protocol that shares the stability of link state algorithms [4] an the avantage of having the routes immeiately available when neee. In pure link state protocols, all links within neighbor noes are eclare an control messages are flooe over the entire network. The OLSR protocol is an optimization of pure link state protocols (e.g. OSPF [5]) for mobile a hoc networks. First, it reuces the size of the control packets: instea of all links, it eclares only a subset of links within those neighbors that are in the MPR set [8]. Secon, it minimizes the flooing of control traffic by using only those noes within the MPR set to broacast its messages. Therefore, only MPRs of a noe rebroacast packets. This technique significantly reuces the number of retransmissions in a flooing or broacast proceure [7]. A etaile escription of the OLSR protocol can be foun in the RFC3626 [3].

3144 A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 2.2. Problem statement In the literature, as in [9], compulsory or guarantee services are sometimes also referre to as preicte or eterministic services, respectively. Deterministic service offers har QoS guarantees to all packets belonging to a ata stream. Preicte service is base on statistical bouns an provies weaker QoS guarantees than eterministic service, but stronger ones than best-effort service. Proviing a QoS service as oppose to a best-effort service is a very complex problem in MANETs [10]. A network s ability to provie QoS epens on the intrinsic characteristics of all network components, from transmission links to the MAC an network layers [11]. Such particular MANET characteristics generally lea to the conclusion that this type of network provies a weak support to QoS. Wireless links have time-varying capacity an larger loss rates than wire links. Topologies are highly ynamic with frequent link outage. Ranom access-base MAC protocols commonly use in this environment have no QoS support, e.g. the IEEE 802.11 family. Moreover, MANET link layers typically run in unlicense spectrum, making it more ifficult to provie strong QoS guarantees. This scenario inicates that har QoS is extremely ifficult to guarantee in a MANET. If the noes have high mobility then even statistical QoS guarantees may be har to achieve ue to the lack of sufficient accurate knowlege of the network states. There is a growing consensus [12,13] that aaptive quality of service moels present the only viable approach to aressing the technical challenges associate with wireless networks. A more realistic approach for QoS provisioning in an a hoc network is base on an aaptive QoS moel where applications must aapt to the time-varying nature of the network. In [14], the efinition of the QoS moel for a MANET inclues a set of parameters that can aapt an application to the current quality of the network. The propose QOLSR protocol introuces quality parameters to the stanar OLSR protocol. Since the quality parameters are available before the routing table calculation, the QoS constraints can be verifie before the route selection. The applications can be aapte really knowing the quality that is expecte to be foun over the network. Therefore, an aaptive cross-layer QoS moel, incluing the application layer, must be propose to aapt the applications. However, this moel is out of the scope of this work. An aaptive QoS routing protocol proviing more appropriate metrics is alreay a consierable improvement in terms of achievable performance for mobile an wireless environment. In orer to esign a QoS routing protocol for MANETs base on a link state approach one shoul: (i) mitigate control traffic, which is the main rawback of proactive protocols; (ii) avoi oscillations an instabilities, which are typically generate by frequent changes on selecte routes; (iii) improve overall performance when compare with the stanar metric base on hop count. 2.3. Our proposal No aitional control messages are inclue for QOLSR. We benefit from Hello an TC messages to evaluate an maintain QoS information through the network. Nevertheless, to carry this QoS information, the format of Hello an TC messages must be change, which increases the size of these messages. In spite of that, the final control traffic loa is smaller than if new control messages were inclue. In Section 3 we evaluate the propose protocol in terms of control traffic. The QOLSR protocol aims to introuce more appropriate metrics for mobile a hoc networks. Many QoS parameters, e.g. banwith, one-way elay, en-to-en elay, energy, cost, jitter, packet loss probability, stability, etc., can be relevant in analyzing a wireless an mobile environment. However, the impact of each QoS parameter epens on the application requirements. We present how we measure these metrics an how these measurements are use as QoS metrics by QOLSR. These measure values are inclue in the QOLSR protocol functionalities. QOLSR oes not require any changes in the format of IP packets. Thus, any existing IP stack can be use an the protocol only interacts with routing table management. As in stanar OLSR, link state information is generate only by noes selecte as MPRs. This information is then use for route calculation in all noes. 2.4. Metric measurements This section presents QoS metrics an how we have esigne the measurement of these metrics. 2.4.1. Delay metric The first QoS parameter to be analyze is the oneway elay metric. Due to the fact that Hello messages are broacast in a controlle manner through the

A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 3145 network, there is no acknowlegement of broacast messages. Consequently, we consier one-way elay rather than en-to-en elay. The measurement of the one-way elay avois the increase of traffic loa typically impose by the aition of extra messages to the QOLSR protocol in orer to estimate the en-to-en elay. However, for successful measurement of the one-way elay, a single global time axis is require. Clocks in a hoc networks, wireless networks an sensor networks, when synchronize via GPS [15], NTP [16], or any of many efficient synchronization protocols for wire as well as wireless meia [17 19], allow the assumption about an approximate, but sufficient, single global time axis. Thereby, base on the above consierations, we assume synchronize clocks. Finally, the assumption of a synchronize network simplifies the one-way elay calculation in the ubiquitous environment. During the neighbor iscovery performe by QOLSR, each noe generating a Hello message inclues its creation time. When a neighbor noe receives this message, it calculates the ifference between such a time an the current time, which represents the measure one-way elay. This measure one-way elay inclues the queuing time, the transmission time, the collision avoiance time, an the control overhea time. Moreover, such a measurement can rastically vary over time. However, it represents a relative measurement since is one using broacast messages rather than unicast messages as use by ata packets. Therefore, we opt for an estimate value avoiing the increase of the traffic control. Furthermore, we use the weighte average metho to accommoate varying elays by using an aaptive algorithm. Our metho is base on the RTT (Roun-Trip-Time) estimation in congestion avoiance an control [20] for TCP [21], which is useful for calculating weighte averages across multiple intervals. average elay ¼ a ol average elay þð1 aþmeasure elay where average_elay: represents the new weighte average, ol_average_elay: stores a weighte average, which is slowly change base on measure_ elay, measure_elay: compute from the elapse time between sening a Hello message an receiving it, 0 6 a 6 1: if a 0 then the formula makes weighte average immune to changes lasting a short time. Otherwise, if a 1 then the formula makes weighte average respon quickly to elay changes. In orer to make a traeoff between elay variations an path oscillations, we efine a threshol value, e.g. 10%. Then, a new average_elay is only compute if an only if the last measurement, i.e. measure_elay, is more than such a threshol value. Otherwise, we conserve the olest average elay without changing the elay measurement. More iscussion about the path oscillation problem is presente in Section 2.7. 2.4.2. Banwith metric The banwith measurement is epenent on the link layer technology. The IEEE 802.11 technology is wiely use in testbes an simulations for wireless a hoc network research. Nevertheless, a hoc networks present even greater challenges than infrastructure wireless networks at the MAC layer [10]. For the sake of simplicity, an for calculating the available banwith, we consier the meium access control scheme escribe in IEEE 802.11b. The metho propose in this section calculates the available banwith between a given noe an its neighbors base on the work of Gerla an coworker [22], which by its turn consiers the acknowlegement time of the ata packets for measuring the banwith metric. Our propose metho inclues both ata packets an signalling traffic (e.g. Hello messages an TC messages in the OLSR protocol), since they also use the available banwith an must be taken into account. Suppose noe i an its neighbor noe j, then we efine the available banwith between them as follows: Bw ði;jþ ¼ð1 uþthroughput packet ði;jþ ; where u is the link utilization. The throughput seen by one packet of S bits can be calculate as: Throughput packet S ¼ t q þðt S þ t CA þ t overhea ÞRþ P R r¼1 B T where t q is the MAC queuing time, t S the time to transmit S bits, t CA the collision avoiance time, t overhea the control overhea time (e.g. RTS, CTS, etc), R the number of necessary transmissions, B T the backoff time for retransmission r. However, as shown in [22], this formula reveals some unesirable characteristics such as packet size

3146 A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 epenence an high variance ue to ranom per packet effects. To increase the statistical robustness of the measurements, a packet winow of 16 or 32 samples (packets) is shown to be aequate. To illustrate, authors in [22] show where the high variance per packets measurements are aggregate on a winow of 32 packets. In this case, the Link_throughput is the aggregate measurement of the Throughput_packet. The ile_time is the uration time uring which a noe is ile an the winow_uration represents the winow size. The ile_time an winow_uration are calculate to prouce the link utilization factor an the permissible throughput measurement as: Permissible throughput ile time ¼ Link throughput winow uration The a hoc network size is use to classify MANETs [10]. The scale of an a hoc network can be classifie as small-scale (i.e., 2 20 noes), moerate-scale (i.e., 20 100 noes), large-scale (i.e., more than 100 noes), an very large-scale (i.e., more than 1000 noes). Many works [23] have presente possible manners to calculate the boun on throughput for a hoc networks. However, the achievable throughput can change ue to network conitions [24]. Moreover, the available banwith unergoes fast time-scale variations ue to channel faing an error from physical obstacles [25]. These effects are not present in wire networks. Then, to make estimation of available banwith an accurate throughput calculation in wireless networks are challenging tasks. Furthermore, the wireless channel is also a share-access meium, an the available banwith also varies with the number of hosts contening for the channel an their bit rates [26]. As a result of these challenges an ue to the instability of the banwith value in IEEE 802.11 networks, we perform our simulations presente in this paper, consiering only the elay metric rather than the banwith metric. We exten upon earlier work in [27] by using a more realistic measurement of the quality of the network. 2.5. Selection of multipoint relays This section iscusses an presents QoS-base MPRs selection heuristics. We start efining the terminology use an presenting the stanar MPR selection efine in RFC3626 [3]. Afterwars, we propose new heuristics consiering QoS parameters when selecting MPRs. 2.5.1. Terminology Accoring to the RFC3626 an base on [28,29], if we consier elay metrics, the following terminology will be use in escribing QOLSR algorithms an heuristics: Neighbor of an interface: a noe is a neighbor of an interface if the interface (on the local noe) has a link to any one interface of the neighbor noe. 2-hop neighbors reachable from an interface: the list of two-hop neighbors of the noe that can be reache from neighbors of this interface. MPR set of an interface: a (sub)set of neighbors of an interface selecte such that through these selecte noes, all strict two-hop neighbors from that interface are reachable. N(x): N(x) is the subset of neighbors of the noe x, which are neighbors of its interface. N2(x): the set of two-hop neighbors reachable from the noe x. D(x,y): the egree of a one-hop neighbor noe y where y is a member of N(x), an is efine as the number of symmetric neighbors of noe y, excluing all members of N(x). Shortest path: a path with minimum elay, calculate by the source noe base on its known partial network topology. Optimal shortest path: the shortest path between two noes in the whole network topology. Any noe in the network can be selecte as an intermeiate noe in the optimal shortest path. 2.5.2. Stanar MPR selection Fining a MPR set with minimal size falls in the category of ominating set problems, which are known to be NP-complete. Demonstrations an proofs were etaile in [7]. The information neee to calculate MPRs is the set of one-hop neighbors an two-hop neighbors. The propose heuristic in [3] to calculate multipoint relay set of noe x is as follows: Step 1: Start with an empty multipoint relay set MPR(x). Step 2: Calculate D(x,y), "noes y 2 N(x). Step 3: First, select those one-hop neighbor noes in N(x) as multipoint relays which provie the only path to reach some noes in N2(x), an

A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 3147 a these one-hop neighbor noes to the multipoint relay set MPR(x). Step 4: While there still exist some noes in N2(x) that are not covere by the multipoint relay set MPR(x): Step 4a: For each noe in N(x) which is not in MPR(x), calculate the number of noes that are reachable through it among noes in N2(x) an which are not yet covere by MPR(x). Step 4b: Select the noe from N(x) as a MPR that reaches the maximum number of uncovere noes in N2(x). Step 5: To optimize, remove each noe in MPR(x), one at a time, an check if MPR(x) still covers all noes in N2(x). The thir step permits to select some one-hop neighbor noes as MPRs which must be in the MPR(x) set, otherwise MPR(x) will not cover all two-hop neighbors. So these noes will be selecte as MPRs in the process, sooner or later. The heuristic use in the stanar OLSR protocol computes a MPR set of carinality at most log n times the optimal multipoint relay number, where n is the number of noes in the network [30]. The stanar OLSR heuristic limits the number of MPRs in the network, ensuring the overhea to be as low as possible. However, in QoS routing, by such a MPR selection mechanism, the goo quality links may be hien to other noes in the network. There is no guarantee that OLSR fins the optimal shortest path with respect to the elay metric. By example. From Fig. 1 we construct the Table 1 consiering the stanar OLSR heuristic: Base on the propose heuristic, noe m will select its MPRs as follows: The Step 3 of the heuristic oes not apply because all two-hop noes are reachable through more than one one-hop neighbor. Going to Step 4 of the heuristic, we have a tie, because noe a an noe j have the same reachability an the same egree. Then, we suppose that the noe a is ranomly selecte as an MPR. Therefore, MPR(m) = a. Then, the noes from N2(m) which are now covere by a noe in the MPR set are remove. The Step 4 is repeate while there exist noes in N2(m) which are not covere by at least one noe in the MPR set. s t v r i j Finally: MPR(m) =a, e, h, c. h a g x Now a constraints to the graph in Fig. 1; these are epicte in Fig. 2. When m is builing its routing table for estination x using the classical MPR heuristic, it will select the route (m,a,x) whose banwith is 15 kbps an the elay is 130 ms. The optimal shortest-wiest path between m an x is (m,j,x). It has 280 kbps as banwith an 24 ms as elay. The ecision on how each noe selects its MPRs is essential to etermine the optimal QoS route in the network. In the MPR selection, links with best QoS resources shoul not be omitte. We present in this section three heuristics for the MPR selection: QOLSR1, QOLSR2 an QOLSR3 consiering the minimum elay path. 2.5.3. QOLSR1 This heuristic is similar to the stanar MPR selection except when we have a tie, because noe a an noe j have the same reachability an the same egree. Then, rather than ranomly selecting an MPR, our heuristic chooses the one-hop m k Noe selecting MPR One-hop neighbor 2-hop neighbor Fig. 1. Example for the multipoint relay selection. Table 1 Example for multipoint relay selection Noe One-hop neighbor Two-hop neighbor MPR m a, b, c,, e, f, g, h, i, j s, v, x, y, z, w, q, k, r, t a, e, h, c f b e c q y w z

3148 A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 s t (100k, 10m) (1k, 600m) v (10k, 1 00m) (5k, 190m) (300k, 2m) (500k, 10 m) (1k, 410m) r i (500k, 10m) (1k, 250m) j h (400k, 10m) (900k, 9m) a g (15k, 1 10 m ) (280k, 15m) (90k, 5m ) (500k, 10m) (20k, 10m) x (400k, 10m) (30k, 20m) (300k, 20m) (1k, 10m) m (70 0k, 60 m) neighbor with the minimum elay path. The heuristic use in QOLSR1 prioritizes the minimum shortest path when selecting the MPR. Such a heuristic moifies the stanar MPR selection presente in Section 2.5.2 by aing the Step 4c as follows: Step 4c: In case of a tie in the Step 4b, select the one-hop neighbor with the minimum elay path as MPR. There is no guarantee that QOLSR1 fins the optimal shortest path with respect to the elay metric. By example. From Fig. 2 we construct the Table 2 consiering the QOLSR1 heuristic: Applying this heuristic, when m is builing its routing table, for estination y, it selects j, e, c, an h MPRs. It will select the route (m,c,y) whose elay is 50 ms. The optimal shortest path between m an y is (m,b,y). It has 20 ms as elay. ( 10 k, 60 m ) k Noe selecting MPR One-hop neighbor 2-hop neighbor Constraints = (bw, elay) f b (150k, 10m) (10k, 400m) (300k, 40m) (10 0k, 90m) (130k, 20m) (100k, 100m) (4k, 10m) e c (400k, 30m) (100k, 10m ) q y (100k, 1 0m) (80k, 1m) (40 0k, 3 m ) (200k, 100 m) (100k, 10m) (120k, 50m) (300k, 20m) Fig. 2. Example for the multipoint relay selection consiering QoS constraints. w z 2.5.4. QOLSR2 The heuristic use in QOLSR2 prioritizes the minimum elay path metric rather than the reachability or the egree. After, if there is more than one MPR with the same minimum elay path, then the MPR with the maximum reachability shoul be chosen. Such a heuristic moifies the Step 4.b an as the Step 4c of the stanar MPR selection presente in Section 2.5.2 as follows: Step 4b: Select the one-hop neighbor with the minimum elay path as MPR. Step 4c: In case of a tie in the Step 4b, select the noe from N(x) as a MPR that reaches the maximum number of uncovere noes in N2(x). There is no guarantee that QOLSR2 fins the optimal shortest path with respect to the elay metric. By example. From Fig. 2 we construct the Table 3 consiering the QOLSR2 heuristic: Applying this heuristic, when m is builing its routing table, for estination z, it selects i, h, b, j, e an c MPRs. It will select the route (m,c,z) whose elay is 140 ms. The optimal shortest path between m an z is (m,,z). It has 100 ms as elay. 2.5.5. QOLSR3 QOLSR3 selects as the MPR the neighbor noe the one with the smallest elay among the neighbors that are, at most, within two hops from the initial noe. Thus, it moifies Step 4b, inclues the new Step 4c, an removes Step 5 from the MPR selection algorithm of the stanar OLSR protocol, as presente in Section 2.5.2. The propose algorithm, incluing the above moifications, is presente next: Step 4b: Select the noe of N(x) as a MPR, which has the minimum elay path consiering N(x) an N2(x). Step 4c: In case of a tie in the Step 4.b, select the noe which reaches the maximum number of uncovere noes in N2(x). Claim 1. Consier the case when each ege is given an arbitrary nonnegative elay (length). In this case, the total elay of this path is efine to be the sum of the elays of its eges. In Fig. 3, we efine p[s,] to be a shortest path from s to ; that is: Table 2 MPR selecte in the QOLSR1 Table 3 MPR selecte in the QOLSR2 Noe MPR Noe MPR N j, e, c, h m i, h, b, j, e, c

A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 3149 v 1 v i v j s u 1 u 2......... u u i+1 u u i-1 j-1 j+1 u k u i u j Fig. 3. Optimal shortest path from s to. p ¼ðu 1 ;...; u i 1 ; u i ; u iþ1 ;...; u j 1 ; u j ; u jþ1 ;...; u k Þ is an optimal shortest path, for k P 3. For any intermeiate noe u i (i 5 1) in p that is not selecte as MPR by its previous noe u i 1, we can fin a noe v i selecte as MPR by u i 1 such as the path (u 1,...,u i 1,v i,u i+1,...,u k ) has at least the same total elay. Proof. Let p =(u 1,...,u i 1,u i,u i+1,...,u k ), k P 3 an optimal shortest path from u 1 to u k, as epicte in Fig. 3. Suppose that on the optimal shortest path, the noe u i is not selecte as MPR by its previous noe u i 1. We can assume that for each noe on the path, its next noe in the path is its one-hop neighbor, an the noe two hops away from it is its two-hop neighbor. Base on the basic iea of the MPR selection that all two-hop neighbors of a noe shoul be covere by this noe s MPR set, u i 1 must have another neighbor v i, which is selecte as its MPR, an is connecte to u i+1. Let p 0 =(u 1,..., u i 1,v i,u i+1,...,u k ), k P 3. Accoring to the criteria of MPR selection specifie on QOLSR3, wherein the metric is the elay over two-hops, then u i 1 selects v i instea of u i as its MPR because: elay ðui 1 v iu iþ1 Þ 6 elay ðui 1 u iu iþ1 Þ ð1þ From Eq. (1) we have elay(p 0 ) 6 elay(p). Base on this assumption, if path p is an optimal shortest path, path p 0 is also an optimal shortest path. Claim 2. There is an optimal shortest path in the whole network such that all the intermeiate noes are selecte as MPR by their previous noes. Proof. By a recurrence. Let p =(s,u 1,...,u i 1,u i, u i+1,...,u j,...,u k,), j < k an optimal shortest path as epicte in Fig. 3. i. By Claim 1, the first intermeiate noe u 1 is selecte as MPR by source s. Then, we can fin a noe v 1 selecte as MPR by s such that the path p 0 =(s, v 1,..., u i 1, u i, u i+1,..., u j,...,u k,) has the same total elay of the optimal path. Then, p 0 is also an optimal shortest path. So, the source s MPRs are on the optimal shortest path. ii. We assume that all the noes u 1,...,u i 1, u i,u i+1,...,u k are selecte as MPRs by their previous noe in the path p. We prove that the next hop noe of u i on p is u i s MPR. Suppose that u j+1 is not an MPR of u j. As above, by using the Claim 1, we can fin a noe v j+1 selecte as MPR by u j such that the path p 0 =(s,u 1,...,u i 1,u i,u i+1,...,u j,v j+1,...,u k, ) has at least the same total elay of the optimal shortest path; then p 0 is also an optimal shortest path. So, in an optimal shortest path, the (j + 1)th intermeiate noe is the MPR of the (j)th intermeiate noe. Base on i. an ii., all intermeiate noes of an optimal shortest path are MPRs of previous noes. By Claim 2, there is an optimal shortest path such that all intermeiate noes are the MPRs of previous noes on the same path. So the optimal shortest path for the whole network topology is inclue in the partial topology the noe knows. Using a shortest path algorithm, as Dijkstra s algorithm [28], we can compute the optimal shortest path in the partial network topology. We can conclue that QOLSR3 fins the optimal shortest path. QOLSR3 fins the optimal shortest path using only partial knowlege of the network topology. The heuristic use in the QOLSR3 fins exactly optimal MPRs that guarantee minimum elay path. Nevertheless, there is no guarantee that this heuristic fins the minimum MPR set an also there is no optimization of the number of MPRs as one in Step 5 of QOLSR1 an QOLSR2 heuristics. Table 4 summarizes all propose heuristics for the MPR selection comparing to the stanar MPR selection of OLSR. In Section 3, we simulate three propose QOLSR selection heuristics using the elay metric an

3150 A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 Table 4 Propose MPRs heuristics Heuristic First metric Stanar MPR selection Secon metric Degree None Yes QOLSR1 Degree Delay Yes QOLSR2 Delay Degree Yes QOLSR3 Delay Degree No compare them with the OLSR protocol. In following section, we present the routing table calculation performe by QOLSR. 2.6. Routing table calculation In traitional ata networks, routing protocols usually characterize the network with a single metric such as hop-count or elay, an use the shortest path algorithms for path computation. For the elay metric, each arc (i, j) in the path p where p =(i,j,k,...,q,r) is assigne a real number elay ij. When the arc (i,j) is nonexistent, then elay ij = 1. Let elay(p)=elay ij + elay jk + + elay qr.the routing problem is to fin a path p* between i an r so that elay(p*) is the minimum. In such a case, we use Dijkstra s shortest path algorithm. 2.7. Path oscillation problem Minimization of the MPR set The consistent elivery of QoS guarantees requires stability of the ata path. In particular, while it is possible that after a path is first selecte, network conitions change an result in the appearance of new better paths, such changes shoul be prevente from unnecessarily affecting existing paths. In particular, switching over to a new (an better) path shoul be limite to specific conitions, e.g., when the initial selection turns out to be inaequate or extremely expensive. This aspect is beyon the scope of QoS routing an belongs to the realm of path management, which is outsie the main focus of this paper. However, because of its potentially significant impact on the usefulness of QoS routing, we briefly outline a possible approach to path management, accoring to [31]. In orer to reuce oscillations between paths, we introuce a threshol value. The basic iea is to trigger path selection only when there is a significant change in the value of metrics compare to last metric measurement. The notion of significance of a change can be base on an absolute scale or a relative one. Accoring to [31], an absolute scale means partitioning the range of values that a metric can take into equivalence classes an triggering an upate whenever the metric changes sufficiently to cross a class bounary. A hysteresis mechanism may be ae to suppress upates when the metric value oscillates aroun a class bounary. OLSR efines a Link Hysteresis strategy escribing quality of the link that mitigates oscillations among paths. A relative scale, on the other han, triggers upates when the percentage change in the metric value excees a preefine threshol. Inepenent of whether a relative or an absolute change trigger mechanism is use, a perioic trigger constraint can also be ae. This constraint can be in the form of a hol-own timer, which is use to force a minimum spacing between consecutive upates. Alternatively, a transmit timer can also be use to ensure the transmission of an upate after a certain time has expire. Such a feature can be useful if link state upates avertising banwith changes are sent unreliably. The QOLSR path selection, propose in this paper, uses a relative scale, setting the threshol value to a preefine percentage. The strategy is: the metric measurement is only taken into account if an only if such a measurement excees a preefine threshol. Such a strategy mitigates oscillations an instabilities, which are typically generate by frequent changes on selecte routes. Nevertheless, since the path oscillation problem involves path management, it is not completely solve only using our proposal, which o not inclue a path management. 3. Performance analysis The performance of the QOLSR protocol is stuie with simulations. QOLSR has been implemente with OPNET moeler [32]. It is particularly popular in the a hoc networking community, an many protocols use in a hoc networks have been implemente, incluing Wireless LAN MAC layer IEEE 802.11 protocol, an some routing protocols such as AODV, DSR an TORA. Our implementation inclues the OLSR moule over the Wireless Moule an QOLSR functions. We evelope the OLSR protocol as specifie in RFC3626. The implementation is completely moular an esigne in compliance with other MANET protocols specifie for raio/wireless moels. In this way, such a moularity guarantees an easy

A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 3151 evaluation among MANET protocols an increases the accuracy in results while we use stanar OPNET moels. The MAC layer an PHY layer of the OPNET wireless moel are esigne base on the IEEE 802.11 stanar [33]. Although the MAC implementation is not complete, it supports most of IEEE 802.11 stanar functionalities, such as the back-off, eference, RTS/CTS, NAV, frame exchange sequence, fragmentation, access point functionality, basic service set, lost frame retransmission, an uplicate frame etection, etc. Both DCF moe an PCF moe are supporte. In aition, OPNET supports the 802.11b physical layer moel [33]. Both PLCP preamble an PLCP heaer are implemente an the multirate support is provie. There are three choices for the physical layer configuration in IEEE 802.11 stanar: frequency hopping, infra re, an irect sequence. In our simulation scenario, we have selecte the Direct Sequence Sprea Spectrum (DSSS) technology. The configuration of the mobile stations is mostly base on the efault parameter values propose for the DSSS system in IEEE 802.11b stanar [34], like the length of RTS, CTS, ACK, MAC heaer, the slot time, the SIFS, the DIFS, the PLCP preamble, etc. To provie the higher rates, 8-chip complementary coe keying (CCK) is employe as the moulation scheme. 3.1. Simulation: mobile network scenario Table 5 Parameters use in the simulations Parameter Value Transmission rate (Mbps) 11 QOLSR a Varying between 0.3 an 0.7 Packet size (bytes) 64 Inter-arrival time (s) Varying between 1 s an 0.0625 s Number of noes 30 Mobility moel Ranom waypoint mobility moel Network area (m m) 1500 m by 300 m Default istance (m) 100 Maximum istance (m) 300 The main parameters use in simulations are given in Table 5. Networks of 30 noes are generate, where noes roam in an area of 1500 m by 300 m. The mobility moel use is the ranom waypoint mobility moel efining the way users move in the simulate area [35]. Network mobility is varie by changing the maximal noal spee v max = 10 m/s. The spee value is varie between 0 an 10 m/s to moel ifferent mobility. In such a moel, each noe picks a ranom estination in the efine area, sample a spee value. Once the noe arrives at its estination, it pauses for a time p = 100 s. The network loa is varie by changing the inter-arrival time of packets accoring to Table 5 parameters. The uration of the simulation is 300 s. Network traffic is generate by a traffic generating source, where the source an the estination are chosen ranomly. This traffic generating source generates packets with 64 bytes. The time between successive traffic generations varies following a specifie value. Such a value is varie changing the packet interarrival time between 1 an 0.08333, i.e. sources generate packets at a rate varying between 1 an 360 packets/s, varying the network loa ynamically. All noes are traffic generate sources. In orer to assess the improvement achieve in QoS support capability by incorporating our algorithms to the original OLSR protocol, we implement three ifferent heuristics for the MPR selection functionality. The implemente heuristics consiering the minimum elay path are QOLSR1, QOLSR2 an QOLSR3 as etaile in Section 2.5.2. Then, we compare simulation results from the original best-effort OLSR protocol an our three QOLSR variants. Simulation results presente are base on a single metric for the routing calculation, i.e. the minimum en-to-en elay. Nevertheless, we have esigne the routing ecision (i.e. the metric use for the routing calculation) as well as QOLSR heuristics as simulation parameters. Then, for each simulation we can specify the metric use an the QOLSR heuristic propose. Other implementations have been one using multiple-metrics, i.e. banwith an elay, for the routing calculation in [27]. The reason that we choose the elay constraint in our simulations, as iscusse in Section 2.4.2, to calculate the available banwith for each noe is still a challenging task. We believe that the elay constraint reflects a more realistic measurement of the quality of the network. 3.2. Simulation results The next figures show the relative performance of the routing algorithms OLSR, QOLSR1, QOLSR2, an QOLSR3. For each simulate routing algorithm, the QOLSR parameter a was varie within the set [0.3,0.4, 0.5,0.6, 0.7]. Results for a = 0.8 an a = 0.9 are not shown because for these values the algorithms become unstable an strongly scenarioepenent.

3152 A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 Fig. 4 epicts the number of receive packets/sec for each simulate protocol. From the figure we can see that uner light traffic, i.e. 50 packets/s, the number of receive packets is about the same for all cases. This occurs because uner light traffic the elay measurements present no significant variation to inuce the selection of a new path. Therefore, as the main ifference between QOLSR an OLSR is the inclusion of the elay metric, there is no consierable ifference between their performance for the case of light traffic. In the case of more than 100 packets/s, the avantages of the propose QOLSR protocols become apparent. Note that the greater is a, the lower is the weight of the last measure elay, i.e. the instantaneous elay. In this case, the instability cause by the instantaneous elay has little influence in the network behavior. For heavy traffic, i.e. more than 350 packets/s, QOLSR3 with a = 0.7 elivers fewer receive packets than QOLSR3 with a = 0.3. This can be explaine because when a = 0.3, an uner heavy traffic, the weight of the last elay measurement is large an therefore the protocol change routes very often. The increase in the traffic increases the en-to-en elay, giving rise to path instability. For heavy traffic, the best results were achieve in the case of QOLSR1 with a = 0.7, where the improvement when compare with stanar OLSR is aroun 12%. Finally, as a benchmark, in the figure we also plot a line representing the maximal values, i.e. the ieal values, where 100% of the transmitte packets are receive with zero roppe packets. Fig. 5 shows the average number of roppe packets/s for each simulate protocol. The number of roppe packets/s is consierably ecrease for all QOLSR variants when compare with stanar OLSR. The best performance was achieve for QOLSR1 with a = 0.7, where the packet loss is 58% smaller than for OLSR. Table 6 summarizes the gains of the QOLSR variants with respect to stanar OLSR. Fig. 6 presents the average packet elay for each protocol. At an arrival rate of uner 200 packets/s, QOLSR1 an QOLSR2 achieve a smaller average elay when compare with OLSR. In the plot we zoom in on the range between 30 an 220 packets/ s which make more clear the gain of the propose QOLSR protocols for that range. Uner light traffic, the best results are achieve by QOLSR1, followe by QOLSR2. From the figure we can also see that the relative performance changes epening on the traffic conition (after 220 packets/s). We have emonstrate in Section 2.5.5 that QOLSR3 fins the optimal path with respect to the elay metric. However, in our simulations we fin that QOLSR3 generates more instability in the paths because such a protocol reacts more to the traffic variation, increasing the total average elay. Moreover, the goo performance achieve 400 350 OLSR Maximal QOLSR1 (alpha=0.7) QOLSR2 (alpha=0.7) QOLSR3 (alpha=0.3) QOLSR3 (alpha=0.7) 300 Receive (packets/sec) 250 200 150 100 50 0 0 50 100 150 200 250 300 350 400 Transmitte (packets/sec) Fig. 4. Comparison of the average of receive packets/s.

A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 3153 80 70 OLSR QOLSR1 (alpha=0.7) QOLSR2 (alpha=0.7) QOLSR3 (alpha=0.7) 60 Droppe (packets/sec) 50 40 30 20 10 0 0 50 100 150 200 250 300 350 400 Transmitte (packets/sec) Fig. 5. Comparison of the number of roppe packets/s. Table 6 Traffic roppe: the gain achieve for each simulate QOLSR protocol comparing with OLSR Protocol Gain achieve (%) QOLSR1 (33.18 62.65) QOLSR2 (26.68 40.77) QOLSR3 (14.58 40.07) by OLSR in Fig. 6 can be explaine by the fact that in OLSR the number of roppe packets is larger than in the QOLSR protocols, an that such roppe packets are not taken into account in the average elay calculation. Then, when a packet is not roppe an is elivere with a large elay, the total average elay is increase. However, by 0.22 0.2 0.18 0.16 0.03 0.028 0.026 0.024 0.022 OLSR QOLSR1 (alpha=0.6) QOLSR1 (alpha=0.7) QOLSR2 (alpha=0.4) QOLSR2 (alpha=0.7) QOLSR3 (alpha=0.5) QOLSR3 (alpha=0.6) 0.14 0.02 0.018 Delay (sec) 0.12 0.1 0.016 0.014 40 60 80 100 120 140 160 0.08 0.06 0.04 0.02 0 50 100 150 200 250 Transmitte (packets/sec) Fig. 6. Comparison between three propose QOLSR algorithms of the average packet elay.

3154 A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 100 90 80 OLSR QOLSR1 (alpha=0.7) QOLSR2 (alpha=0.4) QOLSR2 (alpha=0.7) QOLSR3 (alpha=0.3) QOLSR3 (alpha=0.5) Cumulative probability (%) 70 60 50 50 45 40 40 35 30 30 25 20 20 15 10 5 10 0 0.01 0.012 0.014 0.016 0.018 0.02 0.022 0.024 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Delay (s) Fig. 7. CDF of packets elay. the same fact, QOLSR increases the network reliability. In orer to investigate in etail why the average elay is reuce, we consier the cumulative istribution function (CDF) of packet elays. Fig. 7 shows the CDF of packet elays for a traffic loa of 300 packets/s. In the figure zoom in on the range of 0 50%. From the figure we can see that all QOLSR protocols present better results than the stanar OLSR protocol. We note that QOLSR3 achieves the largest elay reuction, about 44% less than OLSR. In the range of 60 95% in the CDF, the smallest elays are achieve by QOLSR1 with a = 0.7. For more than 95%, the best results are achieve by QOLSR2. The avantage of the three QOLSR variants over the stanar OLSR protocol becomes evient when traffic gets heavy. Specifically, QOLSR3 achieves the best result ue to its inherent improvement in the loa-balance over the whole network. However, this improve loa istribution is achieve through the increase of the MPR set, reflecte in a increase in the control traffic over the network. For the case of light traffic, the performance results for the three propose schemes are very similar, since in this case the routes are more stable. However, we can say that QOLSR1 an QOLSR2 have slightly better performance uner light traffic because they minimize the MPR set an consequently generate less traffic control than QOLSR3. 4. Conclusions This article presente a QoS-base routing protocol for mobile an wireless a hoc networks. In orer to inclue quality parameters in the routing information, QoS measurements were applie. Methos to calculate elay an banwith measurements were propose. The elay metric is calculate between each noe an its neighbors having irect an symmetric links. The banwith measurements are calculate using IEEE 802.11b as the meium access control protocol. However, the throughput is very instable ue to network conitions. Accurate throughput calculation is still a challenging task. Therefore, we consiere elay as a metric rather than banwith as a metric. The implications of routing metrics on path computation were examine an the rationale behin the selection of QoS metrics were iscusse. Heuristics for multipoint relays selection were propose. The heuristic use in stanar OLSR fins a MPR set with minimal size. However, there is no guarantee that OLSR fins the optimal path consiering QoS constraints. Thus, three variants that allow QOLSR to fin the minimum elay path were propose. In orer to inclue quality requirements in the MPRs selection, an also in routing information, elay measurements are applie. We emonstrate that the QOLSR3 heuristic fins the

A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 3155 optimal shortest paths using only partial knowlege of the network topology. The performance of the propose QOLSR variants were examine through computer simulations. The three QOLSR variants achieve better performance when compare with the stanar OLSR protocol. Finally, the propose work can be use to aapt protocol functionalities such as route selection. It woul even be possible to support aaptive applications, such as multimeia, which are sensitive to network changing conitions. Acknowlegements This work is supporte by CNPq. The authors woul like to thank Richar Demo Souza for his revision of the text, an the eitor an the anonymous reviewers for their contributions that enriche the final paper. References [1] S.-Y. Ni, Y.-C. Tseng, Y.-S. Chen, J.-P. Sheu, The broacast storm problem in a mobile a hoc network, in: ACM/IEEE SIGCOMM, Seattle, 1999. [2] E. Beling-Royer, C. Toh, A review of current routing protocols for a-hoc mobile wireless networks, IEEE Personal Communications (1999) 46 55. [3] T. Clausen, P. Jacquet, Optimize Link State Routing Protocol (OLSR), IETF, RFC 3626, October 2003. [4] M. Steenstrup, Routing in Communication Networks, Prentice Hall International Inc., Englewoo Cliffs, 1995. [5] J. Moy, OSPF Version 2, IETF, RFC 2328, April 1998. [6] Mobile A-hoc Networks IETF working group, Available from: http://www.ietf.org. [7] A. Qayyum, L. Viennot, A. Laouiti, Multipoint relaying technique for flooing broacast message in mobile wireless networks, in: HICSS: Hawaii International Conference on System Sciences, Hawaii, 2002. [8] T. Clausen, P. Jacquet, A. Laouiti, P. Muhlethaler, A. Qayyum, L. Viennot, Optimize link state routing protocol, in: IEEE INMIC, Pakistan, 2001. [9] J. emeer, A. Eberlein, Towars an avance qos architecture, in: Multimeia Internet Broacasting, Springer-Verlag, 2001. [10] I. Chlamtac, M. Conti, J.-N. Liu, Mobile a hoc networking: imperatives an challenges, Elsevier A Hoc Journal 1 (1) (2003) 13 64. [11] R. Ramanathan, M. Steenstrup, Hierarchically-organize, multihop mobile wireless networks for quality-of-service support, ACM/Baltzer Mobile Networks an Applications 3 (1) (1998). [12] G. Bianchi, A. Campbell, R. Liao, On utility-fair aaptive services in wireless networks, in: IEEE IWQOS 98, Napa Valley, CA, 1998. [13] M. Naghshineh, M. Willebeek-LeMair, En-to-en QoS provisioning in multimeia wireless/mobile networks using an aaptive framework, IEEE Communications Magazine 35 (11) (1997) 72 81. [14] N. Nikaein, C. Bonnet, A glance at quality of service moels in mobile a hoc networks, in: DNAC 2002, Paris, 2002. [15] T. Logson, The Navstar Global Positioning System, Van Nostran Reinhol, New York, 1992. [16] D.L. Mills, Internet time synchronization: The network time protocol, in: Zhonghua Yang, T. Anthony Marslan (Es.), Global States an Time in Distribute Systems, IEEE Computer Society Press, 1994. [17] K. Roemer, Time synchronization in a hoc networks, in: ACM MobiHoc, Long Beach, 2001. [18] W. Su, I. Akyiliz, Time-iffusion synchronization protocol for sensor networks, IEEE/ACM Transactions on Networking 13 (2) (2005) 384 397. [19] A. Munaretto, M. Fonseca, K.A. Agha, G. Pujolle, Virtual time synchronization for multimeia a hoc networks, in: IEEE VTC 2004-Fall, 2004, Los Angeles. [20] V. Jacobson, M.J. Karels, Congestion avoiance an control, in: ACM SIGCOMM, Stanfor, 1988. [21] J. Postel, Transmission Control Protocol, IETF, RFC 793, September 1981. [22] M. Kazantziis, M. Gerla, En-to-en versus explicit feeback measurement in 802.11 networks, in: IEEE Symposium on Computers an Communications, Taormina, 2002. [23] P. Gupta, P.R. Kumar, The capacity of wireless networks, IEEE Transactions on Information Theory IT-46 (2) (2000) 388 404. [24] M. Grossglauser, D. Tse, Mobility increases the capacity of ahoc wireless networks, IEEE/ACM Transactions on Networking 10 (4) (2002) 477 486. [25] S.H. Shah, K. Chen, K. Nahrstet, Available banwith estimation in IEEE 802.11-base wireless networks, in: 1st ISMA/CAIDA Workshop on Banwith Estimation, San Diego, 2003. [26] M. Heusse, F. Rousseau, G. Berger-Sabbatel, A. Dua, Performance anomaly of 802.11b, in: IEEE INFOCOM, San Francisco, 2003. [27] H. Bais, A. Munaretto, K.A. Agha, G. Pujolle, QoS for a hoc networking base on multiple metrics: banwith an elay, in: The Fifth IFIP International Conference on Mobile an Wireless Communications Networks, Singapore, 2003. [28] M.J. Atallah, Algorithms an Theory of Computation Hanbook, CRC Press, 1999. [29] Z. Wang, J. Crowcroft, Quality-of-service routing for supporting multimeia applications, IEEE Journal of Selecte Areas in Communications 14 (7) (1996) 1228 1234. [30] A. Qayyum, L. Viennot, A. Laouiti, An efficient technique for flooing in mobile wireless networks, Research Report 3898, INRIA, June 2000. [31] G. Apostolopoulos, D. Williams, S. Kamat, R. Guerin, A. Ora, T. Przygiena, QoS routing mechanisms an OSPF extensions, IETF, RFC 2676, August 1999. [32] OPNET Moeler, Available from: http://www.opnet.com. [33] OPNET Technologies, Wireless LAN Moel Description, November 2000. [34] IEEE 802.11 Stanar, November 1999. [35] J. Broch, D.A. Maltz, D.B. Johnson, Y.-C. Hu, J. Jetcheva, A performance comparison of multi-hop wireless a hoc network routing protocols, in: ACM/IEEE MobiCom, Dallas, 1998.

3156 A. Munaretto, M. Fonseca / Computer Networks 51 (2007) 3142 3156 Anelise Munaretto is currently Associate Professor at the Feeral Technologic University of Paraná (UTFPR), Curitiba-PR, Brazil. She was grauate from the Pontifical Catholic University of Paraná (PUC-PR), Curitiba-PR, Brazil with a B.Sc. in computer engineering in 1994. She receive the M.Sc. an Ph.D. egrees in computer networks from the University Pierre et Marie Curie (Paris VI), Paris, France, in 2001 an 2004, respectively. Her research interests inclue routing an quality of service in mesh/sensor/a hoc networks, WiMax, an wireless LAN. Mauro Fonseca is currently Associate Professor at the Pontifical Catholic University of Paraná (PUC-PR), Curitiba-PR, Brazil. He receive the B.Sc. egree in computer engineering from PUC-PR, in 1994 an gaine the M.Sc. egree in networks an istribute systems from the the Feeral Center of Technological Eucation of Paraná (CEFET-PR), Curitiba-PR, Brazil, in 1997. He receive the Ph.D. egree in computer science from the University Pierre et Marie Curie (Paris VI), France, in 2003. His research interests are in service management frameworks an architectures for networks an beyon.