Reliable optimised flooding in ad hoc networks

University of Wollongong Research Online Faculty of Informatics - Papers (Archive) Faculty of Engineering and Information Sciences 2004 Reliable optimised looding in ad hoc networks J. Lipman University of Wollongong, jlipman@uow.edu.au P. Boustead University of Wollongong, boustead@uow.edu.au J. F. Chicharo University of Wollongong, chicharo@uow.edu.au Publication Details his article was originally published as: Lipman, J, Boustead, P & Chicharo, JF, Reliable optimised looding in ad hoc networks, Proceedings of the IEEE 6th Circuits and Systems Symposium on Emerging Technologies: Frontiers of Mobile and Wireless Communication, 31 May-2 June 2004, vol 2, 521-524. Copyright IEEE 2004. Research Online is the open access institutional repository for the University of Wollongong. For further information contact the UOW Library: research-pubs@uow.edu.au Reliable optimised looding in ad hoc networks Abstract Information dissemination (looding) forms an integral part of routing protocols, network management, service discovery and information collection (sensing). Given the broadcast nature of ad hoc network communications, information dissemination provides a challenging problem. Blind looding in ad hoc networks results in the broadcast storm problem. To limit the broadcast storm problem, mechanisms for optimised looding have been proposed. However, this optimisation reduces the inherent level of redundancy. the minimum spanning tree (MST) algorithm using local one hop topology in a distributed manner as the basis of a more reliable optimised looding mechanism called, reliable minimum spanning tree (RMST) lood is proposed. RMST utilises unique properties of MST graphs that allow for broadcast transmissions to be replaced by unicast transmissions. Unicast transmission is inherently more reliable than broadcast transmission as it utilises link layer acknowledgement and retransmission, thereby improving the reliability of a lood and reducing the broadcast storm problem. Simulation is used to show that RMST is able to achieve equivalent reliability in terms of packet delivery compared to blind looding. Importantly, RMST is able to achieve signiicantly beter performance than MPR and equivalent performance to LMSTFlood in terms of reducing the broadcast storm problem. Disciplines Physical Sciences and Mathematics Publication Details his article was originally published as: Lipman, J, Boustead, P & Chicharo, JF, Reliable optimised looding in ad hoc networks, Proceedings of the IEEE 6th Circuits and Systems Symposium on Emerging Technologies: Frontiers of Mobile and Wireless Communication, 31 May-2 June 2004, vol 2, 521-524. Copyright IEEE 2004. his conference paper is available at Research Online: htp://ro.uow.edu.au/infopapers/193 IEEE 6th CAS Symp. on Emerging Technologies:Mobile and Wireless CO”. Shanghai,China; May 3 I-June 2,2004 Reliable Optimised Flooding in Ad hoc Networks Justin Lipman and Paul Boustead and Joe Chicharo Telecommunications and Information Technology Research Institute University of Wollongong, Wollongong Australia Email: Cjustin,paul,joe~hicharo}@titr.uow.edu.au Absfmc-Information dissemination (flooding) forms an integral part of routing protocols, network management, service discovery and information collection (sensing). Given the broadcast nature of ad hoc network commnnications, information disseminationprovides a challenging problem. Blind flooding in ad hoc networks results in the broadeast storm problem. To limit the broadcast storm problem, mechanisms for optimised flooding have been proposed. However, this optimisation reduces the inherent level of redundancy. We propose to apply the M i n i ” Spanning lkee W T ) algorithm using local one hop topology in a distributed manner as the basis of a more reliable optimised flooding mechanism called, Reliable Minimum Spanning ’ h e (RMST)flood. RMST utilises unique properties of MST graphs that allow for broadcast transmissions to be replaced by unicast transmissions. Udcast transmission is inherently more reliable than broadcast transmission as it ntilises link layer acknowledgement and retransmission, thereby improving the reliability of a flood and reducing the broadcast storm problem. We show through simulation that RMST is able to achieve equivalent reliability in terms of packet delivery compared to Blind flooding. Importantly, RMST is able to achieve SigniSeantly better performance than MPR and equivalent performance to LMSTFlood in terms of reducing the broadeast storm problem. Keywords: Ad hoc Network, Flooding, Broadcasting I. INTRODUCTION An ad hoc network is a collection of wireless mobile nodes forming a temporary network lacking traditional centralised administration. Mechanisms for information dissemination in ad hoc networks, such as Blind flooding, form an integral part of communication. Blind flooding is seen as a reliable [l] as all nodes participate in rebroadcasting a message at least once. This redundancy provides an inherently high degree of fault tolerance. However, this results in the Broodcast Storm Pmblem [2]. Numerous optimised flooding mechanisms [3][41[51[6][71 have been proposed to limit the broadcast storm problem. However, limiting the broadcast storm problem reduces the inherent redundancy found in Blind flooding making optimised flooding mechanisms less reliable. We compare the performance of optimised flooding mechanisms and Blind flooding at reliably delivering a message in the presence of increasing background traffic. We show that Blind flooding is remarkably robust and is able to reliably deliver messages, however: it suffers from the broadcast storm problem. Optimised flooding mechanisms aimed at reducing the broadcast storm problem prove to be less reliable in the presence of background traffic than Blind flooding. Optimised flooding mechanisms rely upon selected 0-7803-7938-1/04 /$20.00 ?ZOO4 IEEE nodes to rebroadcast messages during a flood. Given the use of unreliable broadcast transmissions in optimised flooding, a problem arises when nodes responsible for rebroadcasting a message do not receive the message. There exists some work on reliable flooding in wireless networks, however much of this work does not relate to ad hoc networks with an lEEE 802.11 MAC. Related reliable flooding mechanisms [8][9][101[11] provide only limited optimisation and require significant overhead to ensure reliability. Some of these mechanisms require that acknowledgements are returned to the source of a flood, however: this is not always necessary depending upon the application. This is particularly so in a typical ad hoc network where the source of a flood may not know of the existence of non local nodes. There also exist mechanisms [10][11][12] that consider changes to the IEEE 802.1 1 MAC layer. However we do not focus upon the later in this paper. In ad hoc networks, there exists a need for optimised flooding mechanisms that limit the broadcast storm problem yet provide reliability in terms of packet delivery. In this paper we propose Reliable Minimum Spanning Tree (RMST) flooding. RMST is a reliable and optimised flooding mechanism that benefits from the unique nature of the localised Minimum Spaqning Tree (MST) as used in [6][7]. RMST utilises unicast transmission (with link layer acknowledgement and retransmission), which provides a more reliable transport mechanism than broadcast transmission. Reliability is improved at each trhsmitting node, thus RMST distributes the load of ensuring flood reliability among all nodes. We show that RMST is comparable with existing optimised flooding mechanism at reducing the broadcast storm problem. More importantly, RMST shows comparable reliability, in terms of packet delivery, to Blind flooding and greatly improves upon reliability ‘provided by existing optimised flooding mechanisms in the presence of background traffic. This paper is organised as follows. Section ll explores the use of distributed MST and proposes Reliable Minimum Spanning Tree flooding. Section III provides a performance evaluation based on realistic simulations. Section IV concludes. 11. PROPOSED RELIABLE FLOODINGMECHANISM The Minimum Spanning Tree (MST) graph [13], shown in Figure l(a), is a.coMected graph that uses the minimum 521 ,.>b_UdM.W mwbrudm, Fig. 1. Centralised and Distributed MST Fig. 2. RMST Roodutilising IEEE 802.1 I .total edge length. This results in a graph with one less edge than the number of vertices. The MST is traditionally used in networks for determining broadcast trees using global topology information. In [6] and [71 the authors propose the use of MST with restricted one hop topology as the hasis of a distributed optimised flooding mechanism. This allows for an optimal broadcast set (BSET) of nodes with minimal .transmission range to be determined as with the centralised approach. More importantly, the resulting distributed MST graph (Figure I@)) does not exhibit the tree like structure of the centralised MST (Figure I(a)). It can be seen by comparing Figure l(a) and Figure l@) that centralised MST 2 distributed MST as described in [14]. Thus many of the performance benefits (reducing the broadcast storm problem) of centralised MST are maintained with the addition of fault tolerance not found in the centralised MST approach. However, there exists a significant problem in ' broadcast environments where a broadcast transmission may be lost due to packet corruption, packet collision or hidden node transmissions. Therefore it is possible that nodes may not receive a broadcast transmission. Furthermore those nodes that do not receive a broadcast transmission may be required to receive a transmission. This is especially true in the case of optimised flooding mechanisms, where selected nodes are responsible for retransmission. Given that optimised flooding mechanisms greatly reduce the redundancy found in Blind flooding, there may be situations where a packet may be lost and a flood may not propagate due to reduced redundancy. RMST is a reliable and optimised flooding mechanism that computes a local MST based upon one hop neighbour knowledge in a distributed manner as is done in [61 and [71. The MST allows nodes to determine the closest neighbouring nodes that must be included within any transmissions, to ensure a connected graph, thereby ensuring a flood propagates throughout an ad hoc network. The distributed MST results in a connected graph with a neighbour degree greater than one but less than six and an average neighbour degree of less than 2.04 nodes [14]. If the prior broadcasting node is removed, the average neighbour degree is reduced to 1.04 nodes. This low neighbour degree results in a reduced BSET of neighbouring nodes to which a broadcasting node must transmit a message. The resulting small BSET allows for IEEE 802.11 broadcast transmissions (as used by existing flooding mechanisms) to be replaced with IEEE 802.11 unicast transmissions. Unicast 522 micast and link layer ARC! transmission is a more reliable transport mechanism than broadcast transmission as it implements a RTS/CTS exchange at the MAC layer prior to transmission in order to reduce problems associated with the hidden node problem. More importantly, unicast transmission utilises a frame retransmission mechanism at the MAC layer based upon a positive acknowledgement scheme (ARQ). Thus, a transmitting node will retransmit a frame if it does not receive a positive acknowledgement from the destination node. The IEEE 802.11 ARQ is not completely reliable and packet loss is still possible. However it provides a more reliable transport mechanism than broadcasting and requires no modifications to the MAC layer. The number of retransmissions before a timeout o c c m may be adjusted, but is generally 4-7 retransmissions. If a node fails to retransmit a message to a destination node, it is able to detect the failure and may utilise an alternative scheme (such as broadcasting) to continue dissemination. Figure 2 shows the distributed MST graph for a topology of nodes. Nodes obtain their local topology through the exchange of beacon messages. Nodes B and D are node A's determined MST neighbours and must be included in any transmissions from node A. In LMSTFlood, node A would adjust its broadcast transmission power to include the distance of its furthest MST neighbour. However, in RMST, node A will first unicast a message to its furthest MST neighbour. The unicast is shown by a black directed line. If this unicast is successfull it will then unicast the message to the next furthest node, in this case node B. The reasoning for unicasting to the furthest node is a result of the limited transmission distance and the possibility of a node moving out of broadcast distance in a highly mobile environment. In Figure 2, both unicat messages are successfully delivered. However, when node B transmits to node F and node D transmits to node E both packets are lost or corrupted. Therefore, at the link layer, both nodes then retransmit as shown by the dashed grey directed limes until an ACK is received or the maximum number of retransmissions is reached. Each node, upon receiving a broadcast message, calls RMSi"(message). The algorithm determines if tlie message has been seen before. If not, then a BSET is determined by supplying the MST with the node's one hop topology. The previous broadcasting node and all neighbouring nodes that may have heard the previous broadcast are removed from the BSET. If the BSET is not an emptyset, then the utilises transmission power control thus limiting the number required transmission power to reach the remaining nodes of nodes affected by a broadcast, whereas MF'R does not. Figure 3 shows the percentage of nodes that receive a in the BSET is determined and the message rebroadcast. The MST algorithm used is based upon Prim's algorithm [IS]. message as the CBR packet rate is increased. In a NULL MAC environment, delivery is assumed to be 100%.However, in GloMoSim the use of a more realistic IEEE 802.11 MAC Algorithm RMST(message) 1. if not seen message before and transmission medium. results in nackets being lost due to 2. BSET MST(I-bop Neighbours) collision, corruption and fading. We utilise three CBR source3. i +previous broadcasting node destination pairs in the simulation to create background aaffic 4. H +nodes that recieved orevious broadcast that may effect the delivery performance of the flooding mech5. BSET + BSET - * anisms. The source-destination pairs are selected randomly and BSET t BSET - H 6. 7. for each node i in BSET UDP packets of 512 bytes are transmitted between nodes using 8. Tpower transmission.power(i) the AODV routing protocol. Each source begins transmitting 9. Unicast(Message, TpoIUe7.) data at a random time prior to the initiation of a flood. From Figure 3, it can be seen that Blind flooding and 111. RESULTS RMST provide the best delivery performance and are only We utilise the GloMoSIM 2.03 simulation environment with slightly affected bx background traffic. Blind flooding provides two different MAC layers. An ideal NULL MAC layer is reliability through redundant broadcasts, but suffers from the used to create an environment with no medium contention broadcast storm problem as shown in Figures 4 - 6. However, nor hidden-node scenario. The transmission medium is error RMST being optimised limits the broadcast storm problem. free. A bidirectional link between two nodes is assumed upon RMST achieves comparable delivery to Blind flooding as it reception of a beacon message. In the NULL MAC layer, a first utilises unicast transmissions which are more reliable than order radio model [I61 is assumed. In this model the first order broadcast transmissions. LMSTFlood and MPR suffer in deradio dissipates Eelec = 50nJJbit to run the circuitry of a livery as broadcast packets are affected more significantly transmitter or receiver and a further Eamp= 100pJ/(bit*m2) by background traffic as both mechanisms rely upon specific for the transmitter amplifier. Equation 1 is used to calculate the nodes receiving a broadcast. In the case of LMSTFlood, nodes costs of transmitting.a k-bit message a distance d. Equation 2 are able to determine whether they are required to rebroadcast is used to calculate the costs of receiving a k-hit message. The by calculating their local MST. But if a node does not receive radios have power control and consume the minimal required a broadcast message then it effectively halts the flood in that energy to reach the intended recipients. The second MAC layer direction. MPR (source based) attaches a relay list to the tested is the IEEE 802.11 MAC layer as implemented in Glo- broadcast message, thus if the message is not received by moSIM, however this has been modified to allow transmission one or more of the relays, then it may effectively cancel the power control for broadcast and unicast packet transmission propagation of the flood at that point. as required. The simulation area is 600 meters by 600 meters. Figure 4 shows the power consumed by each mechanism Nodes are placed in a random topology within this area. to complete a flood. RMST utilises more energy to complete Nodes have a maximum transmission range of 100 meters. a flood than LMSTflood. This is expected as RMST must A node within each random topology is selected randomly as perform more transmissions (Figure 5) than LMSTFlood, the source of a flood. The topologies generated are not fully resulting in more duplicate packets (Figure 6). Compared connected thus some topologies may result in a partitioned ad to Blind flooding and MF'R, RMST shows significantly hoc network. The total number of nodes reachable for each better pelformance in terms of reducing the broadcast storm topology is determined so as to account for partitioning. problem. The use of transmission power control in RMST when unicasting allows for a reduction in duplicate packets ET=(^, d ) = Eelcc* k E.,, * k * d2 (1) received and limits the number of nodes that will bear a transmission thereby reducing power consumption. ~~ ~ - ~ - + Simulations are run 50 times with a different seed for each run. The final results are averaged and 95% confidence intervals are displayed in each graph. Blind flooding, MPR (source based) and LMSTFlood are the comparison flooding mechanisms. Blind flooding is selected as it is a brute force approach with a high degree of reliability, but suffers from the broadcast storm problem. MPR and LMSTFlood were selected as they are both optimised flooding mechanisms that reduce the broadcast storm problem in ad hoc networks. LMSTFlood Thus, the combined use of unicast transmission and distributed minimum spanning tree enables RMST to achieve comparable reliability to Blind flooding, surpassing existing optimised flooding mechanisms. Additionally, RMST effectively limits the broadcast storm problem outperforming MPR and acheiving comparable performance to LMSTFlood. IV. CONCLUSIONS Various mechanisms for reliable flooding have been proposed in literature. However, they either suffer from significant overhead to disseminate and determine if a flooded message 523 Fig. 3. Broadcast Reachability with Background CBR traffic Fig. 6. Duplicate Packets Recieved to Blind flooding. REFERENCES was received or they require modifications to the IEEE 802.11 MAC layer to improve broadcast delivery between nodes. In this paper, we have introduced Reliable Minimum Spanning Tree (RMST) flooding. RMST is a distributed and more reliable optimised flooding mechanism that benefits from the unique nature of the distributed minimum spanning tree and requires no modification to the IEEE 802.11 MAC layer. RMST utilises unicast packet transmission, which provides a mnre reliable transport mechanism than broadcast packet transmission as used by existing optimised flooding mechanisms. We show that RMST compared to LMSTFlood, MPR and Blind flooding is able to reliably deliver packets given three source-destination pairs generating CBR traffic. In fact the performance of exisiing optimised -flooding mechanisms was shown '0 suffer in the presence of CBR traffic as packet rate was increased. RMST shows comparable performance at reducing the broadcast storm problem when compared to existing optimised flooding mechanisms. Significa&, RMST shows equivalent packet delivery performance when compared wm- 2 54 yl - Im IYJ Fig. 5 . Packets transmitted 524 ml [I] S . Basagni, A. D. Myers, and V. R. Symtiuk. Mobility-independent flooding for real-time, multimedia applications b ad hoc networks. In P m . 1999 IEEE Emerging Technologies Symposium on wireless Communicotionr & System, Richardson,TX. April 12-13 1999. [21 Sze-Yao Ni. Yu-Chee Tseng, Yuh-Shyan men,and Jang-ping She". 'The broadcast storm problem in B mobile ad hoe network. In PmceedingS of the jjih annual A C M E E E international conference on Mobile computing and networking, pages 151-162. ACM Press. 1999. [3] A. Qayyum, L. VlennoG and A. Laouiti. Multipoint relaying: An efficient technique far floodingin mobile wireless networks. 35th Annual Hawaii International Conference on System Sciexes, 2001. [4] Yn-Chee Tseng, SmYao Ni. and En-Yu Shih. Adaptive approaches to relieving broadcast storms in a wireless multihop mobile ad hoc network. In f"femOtiOMl Conference on Distributed System, pages 481A88, 2001. [51 1. Cartigry, F. Ingelrest, and D. Simplot. Rng relay subset Roodingprot m l s in maneu. In I n t e m i o M l Jooumol of Foundations of. Comsuter . Science Vol. 14, No. 2. pages 25S265. April 2003. [6] lustin Lipman, Wul Boustend, Joe Chicham, and John Judge. Optimised Roodingmechanism in ad hoc networks. In Pmceedings ofthe Workshop 0" the InfeVEf, ? b k C O - W t i C d O U Md PmUSsing ( w I n P ' o 3 ) , pages 27-33, Coolangata, Australia, December 2003. [7] I. CaRigny, F. Ingelrest, D. Simplot-Ryl, and I. Stojmenovic. Localized lmst and m g based minimum energy broadcast protocols in ad hoc netwarb. In PNC. IEEE INFOCOM. Son Francisco, CA, USA, 2003. [SI S. Alagar and S. Venkatesan. Reliable broadcast in mobile wkless networks. In Pmceedings Milirmy Communicnrionr Conference, pages vol. I, pp. 236240, 1995. 191 E. Wmni and G. P. Rossi. Reliable broadcast in mobile multihov v a ~ h netw&ks. In Mobicom 97, Budopest, Hungmy pages pp.34- 4!2,'1997. [IO] 1. Tounilhes. Robust broadcast: Improving the reliability of broadcast uansmissions on csmdca. 10 IEEE Intemtional symposium on perSOMI, Indoor and Mobile Radio Communications, pages vol. 3, pp. I 1 11-11 15. 1998. [ I l l Ken Tang and Mario &la. Mac layer broad& support in ieee 802.11 wireless In pmceedjngs Mi/itnry Communica,ionr Conference, pages 5-548, 2ooO. [I?.] Ivan Stojmenovic, Mahtab Seddigh, and lovisa Zunic. Dominating sets and neighbor elimination-based broadcasting algorithms in wireless netwwb. In IEEE Trmmtions on Pomllel Md Distributed System, pages Vol. 12. No. 12. December 2001. [I31 G. Toussaint. The relative neighbourhwd graph of finite planar set. Panem Recognitioq pages vol. 12, no. 4, pp. 261-268, 1980. 1141 Ning ti, Jennifer C. Hou. and h i Sha. Design and analysis of an mst-based topology conml algorithm. In P m . oflEEE Infocom, 2003. 1151 R. Prim. Shortest connection networks and some generalisations. The Bell System Technical b u m / , pages vol. 36, pp. 3389-1401, 1957. [I61 W a d i Rabiner Heinzelman. Anantha Chandrakasan, and Hari Balakrishnan. Energy-efficient communication protocol far wireless microsensor networks. In Proceedings of the Hawaii Internarioml Conference on System Sciences, pages 1-10. January ZWO. ~~ ~~