Real-Time Traffic in Ad-hoc Sensor Networks

Real-Time Traffic in Ad-hoc Sensor Networks Luciano Bononi Computer Science Department University of Bologna 4127 Bologna, Italy Email: bononi@cs.unibo.it Lorenzo Donatiello Computer Science Department University of Bologna 4127 Bologna, Italy Email: donat@cs.unibo.it Marco Furini DSSCQ Department University of Modena and Reggio Emilia 421 Reggio Emilia, Italy Email: marco.furini@unimore.it Abstract The usage of sensors is emerging in the nowadays scenario where safety related applications are more and more used. These applications have timing constraints that need to be met by the control system. Thus, guaranteeing timeliness properties is considered a key challenge for research on wireless sensor networks. Among the different components that need to address this problem, we focus on the communication subsystem and, in particular, on the MAC protocols that handles transmissions. To provide timeliness behavior, the employed MAC protocol has to provide a time-bounded service. The contribution of this paper is to investigate the performance of two wireless MAC protocols that can be employed in wireless sensor networks: IEEE 82.11e and. The former is part of the well know IEEE 82.11 standard, whereas the latter has characteristics that well fit the requirements of wireless sensor networks. Results show that two protocols have similar performances for networks with limited devices and/or traffic, whereas when devices and/or traffic increases, has to be preferred. I. INTRODUCTION Sensor networks are used in many areas, from defence and surveillance to health or intelligent homes, and applications based on wireless sensor networks are emerging nowadays [1], [2], [3]. Examples include safety related applications (e.g., fire monitoring), environment observation (e.g., highway traffic monitoring), health control (e.g., medical care), and weather forecast (e.g., flooding control). Many of these sensor-based applications have timing constraints that require a sensor to: i) sense a physical phenomenon (e.g., temperature), ii) process the sensed value, and iii) communicate with other sensors, in a bounded and guaranteed time. The importance of timing constraints in wireless sensor networks is growing, and guaranteeing a timeliness behavior is considered a key challenge for research on wireless sensor networks [4]. The respect of the application timing constraints needs to be addressed by the whole system that handles the application: from wireless networking protocols to operating systems, from middleware services to data management, from programming models to theoretical analysis [5]. Looking at the communication sub-system, the network should guarantee an upper bound to the worst-case transmission time in order to support sensor-based applications that have timing constraints [6]. In fact, communication in a timely fashion is fundamental to correctly support real-time applications. Since timely communication is typically supported at the Medium Access Control (MAC) layer, the employed MAC protocol has to provide a time-bounded service [7]. Many new MAC protocols are currently being proposed and tested for wireless sensor networks, but most of them focus on energy efficiency, and not on timing constraints [8]. Therefore, MAC protocols with timing features are desirable. One can think of using real-time MAC protocols designed for ad-hoc networks, but although wireless sensor network and ad-hoc networks have several common characteristics, there are few important differences: in ad-hoc networks, nodes cooperatively organize themselves into a communicating radio network, moving and adapting to changes in the environment, whereas in sensor networks sensors slightly move or do not move at all. However, this does not mean that wireless sensor networks have fixed topology; in fact, changes may occur because of the changing nature of the wireless link and the death of sensors due to technical problems or battery exhaustion. Another important difference is that communication among nodes in ad-hoc networks might be done in a multi-hop way, each node being a router for other nodes communications, whereas in sensor networks the communication among nodes is likely to be done in multi-hop transmissions, as the node s embedded radio interface might be not powerful enough for the signal to reach the destination in one hop; therefore intermediate nodes are used to relay messages. In the literature, different wireless MAC protocols have been proposed to deal with real-time transmission. For instance, IEEE 82.11e is an amendment of the well-know IEEE 82.11 and aims at supporting real-time transmissions in a wireless environment; TBMAC [9] provides a delay bounded service using CDMA technology to avoid transmission collisions: the delay bounded service is provided requiring all the mobile hosts to have a common clock and the system needs to be aware of the exact geographic position of each mobile host; TPT (Token Passing Tree) [1] supports real-time applications in ad-hoc networks in which terminals have low mobility (e.g., indoor scenario); [11] have similar characteristics, but provides a better service than TPT. The contribution of this paper is to investigate the performance of two wireless MAC protocols that can be employed in wireless sensor networks: IEEE 82.11e and. The former has been chosen as it is part of a widely diffused standard, whereas the other has been chosen as its characteristics well fit with requirements of wireless sensor networks (e.g.,

multi-hop transmissions, low mobility of devices, and possible sudden variation of the network topology due to devices that stop working). Through a simulation study we analyze the goodput and the MAC delay; the goodput represents the number of bits per second that a source is able to deliver to the destination, whereas the MAC delay is the time that elapses between the arrival of a packet (of data) at the head position of the output queue and the arrival at its destination. Results reveal practical design suggestions: for networks with limited devices or with limited traffic the two protocols achieve similar results, whereas for networks with a considerable number (> 2) of stations or traffic, should be preferred. The remainder of this paper is organized as follows. In section II we briefly overview IEEE 82.11e and ; Section III presents the evaluation of the two protocols and their comparison. Conclusions are drawn in section IV. II. BACKGROUND In this section we briefly overview the characteristics of IEEE 82.11e and of. A. IEEE 82.11e IEEE 82.11e is an approved amendment to the IEEE 82.11 standard that defines a set of Quality of Service enhancements for wireless LAN applications through modifications to the Media Access Control (MAC) layer. It has been designed to support delay-sensitive applications. This version introduces the concept of traffic categories with the Enhanced Distribution Coordination Function (EDCF). Each station has eight traffic categories (i.e., every packet to transmit has its own priority). Using EDCF, stations try to send data after detecting the medium is idle and after waiting a period of time defined by the corresponding traffic category called the Arbitration Interframe Space (AIFS). A higher-priority traffic category will have a shorter AIFS than a lower-priority traffic category. To avoid collisions, a station waits for a random number of time slots (known as a contention window) before attempting to transmit data. If another station transmits before the countdown has ended, the station waits for the next idle period. The length of the contention window is defined with CWmin and CWmax. EDCF also defines the maximum time a station can transmit once it accesses the network (with the parameter TXOP). A detailed description of IEEE 82.11e can be found at http://ieee82.org/11/. Here, we highlight that this protocol provides probabilistic (and not absolute) timing guarantees to the applications. B. [11] is a distributed real-time MAC protocol that supports real-time and generic applications in ad-hoc networks. is designed to be used in scenarios where devices have limited mobility in a limited space (e.g., indoor scenario). operates in slotted networks, provides two types of service (delay bounded and best-effort), and allows concurrent access transmissions by using CDMA technology to avoid collisions. Each node is required to have two different output queues: one for the real-time and one for the non realtime traffic (the former has higher transmission priority than the latter). Briefly, the protocol organizes all the devices as a virtual ring and uses an access mechanism to control transmissions and to integrate real-time and non real-time traffic. The access mechanism is based on a control signal that travels into the virtual ring. Every time this signal visits a station, it gives transmission authorizations. A station can always transmit as long as it has transmission authorizations. The number of such authorizations is defined by the call admission procedure and may vary from station to station. The transmission is done using CDMA and transmitter-based spreading codes. Once a packet is transmitted, it travels from its source to its destination hop by hop. The control signal travels into the network but can be seized by a station that has real-time traffic to transmit and still have transmission authorizations since the previous visit of the control signal. When released (if captured) or when passed (if not captured) the control signal provides new transmission authorizations to the visited station. The access mechanism provides a time-bounded service (i.e., an upper bound to the network access time) and is proved to be fair (i.e., stations have the same opportunity to access the network). Since the protocol operates in a wireless environment, is provided with a mechanism that manages the dynamic variations and failures of the virtual ring. Briefly, the mechanism works as follows: when the control signal visits a station, this one can become an ingress station with a certain probability. An ingress station stops all the transmissions in the network and listens to the wireless channel for a predefined period of time in order to discover new stations that want to join the existing network. If no stations are asking to join the network, the ingress station informs all the other stations that transmissions are no longer stopped; conversely, if new stations want to join the network, the ingress station is in charge of re-building the virtual ring embedding the new requesting stations. After building the virtual ring, transmissions are allowed. This mechanism does not compromise timing guarantees, as the upper bound also include the time to manage dynamic variation of the network topology. A detailed description of can be found in [11]. Here, we highlight that the protocol provides absolute timing guarantees to the applications. III. EVALUATION In this section we investigate the performance of IEEE 82.11e and of. As previously mentioned, these protocols have been chosen as the former is an amendment of

the widely used IEEE 82.11, whereas the latter is a protocol that supports real-time applications in ad-hoc networks where nodes have limited mobility in a limited space without requiring additional node information like the exact geographic position, and it allows multi-hop transmissions (characteristics that well fit the requirements of wireless sensor networks). The goal of this investigation is to study the goodput and the MAC delay while varying the number of stations present in the network and the amount of real-time traffic to transmit. We recall here that with goodput we refer to the number of bits per second that a station is able to deliver to their destination (considering only data bits and not header bits), and with MAC delay we refer to the time that elapses between the arrival of a packet at the head of the transmission queue (i.e., the packet is the first one that will be transmitted by the station) and the arrival to its destination station. For simulation we use the discrete event simulation environment OMNeT++, which is free of charge for teaching and research use [12]. The simulation settings for IEEE 82.11e are the following: Channel data rate of 11 Mbps; Packet length of 124 bits; Network slot length of 2µs; Non real-time traffic rate is unlimited (always traffic to transmit); Real-time traffic has priority 7 (i.e., highest priority); Non real-time traffic has priority (i.e., lowest priority); AIFSN and TXOP are equal to 3 and (for non realtime traffic), and to 2 and 1.54 milliseconds (for realtime traffic). Note that these are the default values for the channel access; CWmin and CWmax are equal to 31 and 123, respectively. Note that these values depends on the physical layer, and that we consider the DSSS physical layer specifications. The simulation settings for are the following: Channel data rate of 11 Mbps; The number of slots circulating into the virtual ring is equal to the number of stations; Network slot length of 53 bytes; Probability of becoming an ingress station when receiving the control signal is equal to 4%; The number of real-time authorizations for every round of the control signal round is equal to 5; The number of non real-time authorizations for every round of the control signal is equal to 5; Applications have always non real-time traffic to transmit (non real-time traffic data rate is unlimited). As in most of the work on the subject, the simulation results are reported assuming the network to be error-free during transmission. A. Goodput Figure 1 shows the real-time traffic goodput of the network while varying the number of stations in the network, and while Goodput (kbps) 25 2 15 1 5 IEEE 82.11e 1 2 3 4 5 6 Fig. 1. Real-time traffic goodput achieved when every station of the network Goodput (kbps) 25 2 15 1 5 IEEE 82.11e 2 4 6 8 1 Fig. 2. Real-time traffic goodput achieved in a network of 25 stations where every station has unlimited non real-time traffic to transmit. (95% confidence interval) considering each station of the network to generate real-time traffic with a data rate equal to 45 Kbps. It can be observed that is able to handle all the real-time traffic generated by stations, whereas IEEE 82.11e does not when the number of stations in the network goes beyond 2. Figure 2 presents a similar investigation, but here the realtime traffic goodput of the network is measured while varying the real-time traffic in a network of 25 nodes. It is not surprisingly to observe that is able to handle all the generated real-time traffic, whereas IEEE 82.11e is not. This means that when using IEEE 82.11e, nodes will see their real-time output queue to grow. Results of both investigations show that IEEE 82.11e and achieve comparable real-time goodput when the network traffic load is low (either due to small network or to limited real-time traffic data rate). When the network traffic load increases, IEEE 82.11e is not able to handle the realtime traffic, whereas is. In addition to the real-time traffic goodput, it is interesting to observe the goodput of the non real-time traffic, as the network

Goodput (kbps) 16 14 12 1 8 6 4 2 IEEE 82.11e 1 2 3 4 5 6 Fig. 3. Non real-time traffic goodput achieved when every station of the network has unlimited non real-time traffic and 45 kbps of real-time traffic to transmit. likely supports both real-time and non real-time applications. Again the non real-time goodput is measured while varying both the number of stations in the network and the real-time traffic data rate. Figure 3 shows the non real-time traffic goodput of the network while varying the number of stations in the network, and while considering each station of the network to generate real-time traffic with a data rate equal to 45 Kbps. For both protocols, the goodput decreases while increasing the number of stations in the network. However, IEEE 82.11e achieves very low goodput, whereas has much higher goodput. This behavior shows that the mechanism that controls the transmission in IEEE 82.11e leaves no room for the non real-time traffic when the real-time traffic to handle is considerable. Therefore IEEE 82.11e is not fair in handling real-time and non real-time traffic. Conversely, results show that the mechanism that controls transmission in ensures a fair access to the network by both the real-time and the non real-time traffic. Results are confirmed with a simulation that computes the non real-time traffic goodput of the network while varying the real-time traffic in a network of 25 nodes. B. Mac Delay Figure 4 shows the real-time MAC delay while varying the number of stations in the network, and while considering each station of the network to generate real-time traffic with a data rate equal to 45 Kbps. IEEE 82.11e provides a much lower delay than WRT- Ring. Hence, it supports applications with stringent real-time constraints. However, it is to note the IEEE 82.11e does not support multi-hop transmissions and therefore it requires all the nodes to see each other within the network (transmissions are single hop source-destination only). This explains the lower MAC delay. Conversely, in a packet needs to perform multi-hop transmissions to reach its destination. Since wireless sensor networks require multi-hop transmissions, we build a benchmark scenario where the multi-hop feature is built 2 18 IEEE 82.11e 16 IEEE 82.11e MH 14 12 1 8 6 4 2 1 2 3 4 5 6 Fig. 4. Real-time MAC delay achieved when every station of the network 1 9 IEEE 82.11e IEEE 82.11e MH 8 7 6 5 4 3 2 1 2 4 6 8 1 Fig. 5. Real-time MAC delay: achieved in a network of 25 stations where every station of the network has unlimited non real-time traffic to transmit. upon the IEEE 82.11e protocol (i.e., the network nodes are considered as composing a virtual ring). Figure 4 also presents a series (label IEEE 82.11e MH), which represents results of IEEE 82.11e when simulating multi-hop transmissions. It can be observed that when the number of stations goes beyond 3, IEEE 82.11e provides higher MAC delays than. Figure 5 shows the real-time MAC delay while varying the real-time traffic present in a network of 25 nodes. The MAC delay is steady for, whereas it increases for IEEE 82.11e MH while increasing the real-time traffic load. Results confirm that IEEE 82.11e is effective in supporting real-time traffic when transmissions are done in single hop; conversely, when transmissions are done in multi-hop the MAC delay increases a lot and has to be preferred. Figure 6 shows the non real-time MAC delay while varying the number of stations in the network, and while considering each station of the network to generation real-time traffic with a data rate equal to 45 Kbps. The MAC delay of is lower than the one experienced while using IEEE 82.11e. Similar results are obtained while varying the real-time traffic present in a network of 25

14 12 1 8 6 4 2 IEEE 82.11e 1 2 3 4 5 6 Fig. 6. Non real-time MAC delay achieved when every station of the network 9 8 7 6 5 4 3 2 1 IEEE 82.11e 2 4 6 8 1 Fig. 7. Non real-time MAC delay achieved in a network of 25 nodes where achieved when every station of the network has unlimited non real-time traffic to transmit. nodes (see Figure 7). C. Summary of Results Results obtained while analyzing the goodput and the MAC delay experienced by IEEE 81.11e and by under different network conditions show that both protocols can be employed in wireless sensor networks if the number of devices is limited and/or the real-time traffic is limited too. In particular, for a number of devices below 2 and with a global network real-time traffic below 1 Mbps both protocols can support the traffic generated by the real-time applications. What differs in this scenario is the performance of the non real-time traffic, which is better handled by. For network with a higher number of stations (> 2) and/or higher real-time network traffic (> 1Mbps), achieves better performances in terms of goodput. Results obtained while analyzing the MAC delay showed that IEEE 82.11e could support applications with real-time requirements tighter than, providing that all the devices of the network are within the range of communication (i.e., single hop transmission). Conversely, in generic wireless sensor networks conditions where multi-hop transmissions are required, achieves lower MAC delay. IV. CONCLUSIONS This paper focused on the support of applications with timing constraints in wireless sensor networks. In particular, it presented an investigation of the performances of two MAC protocols that can be employed to manage real-time traffic: IEEE 82.11e and. The investigation revealed that supports both real-time and non real-time traffic in a fair way, whereas IEEE 82.11e penalizes the non real-time traffic in favor of the real-time traffic. The investigation also showed that IEEE 82.11e might have problems in supporting real-time applications if the real-time traffic load goes beyond a certain threshold. To summarize, results showed that in wireless sensor networks with limited devices and/or real-time traffic both protocols can be employed, whereas when devices and/or traffic increases, has to be preferred. ACKNOWLEDGMENT This work has been partially supported by Italian MIUR funds (project PRIN-26 NADIR: Design and analysis of distributed, and QoS-aware protocols and algorithms for Wireless Mesh Networks). Authors would like to thank Luca Cibelli and Angelo Trotta for their work on the simulator. REFERENCES [1] S. Y. Lin Ruizhong, Wang Zhim, Wireless sensor networks solutions for real time monitoring of nuclear power plant, in Fifth World Congress on Intelligent Control and Automation, 24., vol. 4, June 24, pp. 3663 3667. [2] I. K. I.F. Akyildiz, Wireless sensor and actor networks: research challenges, Ad Hoc Networks, vol. 2, no. 4, pp. 351 367, 24. [3] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, Wireless sensor networks: A survey, Computer Networks, vol. 38, pp. 393 422, 22. [4] T. Watteyne, I. Augé-Blum, and S. Ubéda, Dual-mode real-time mac protocol for wireless sensor networks: a validation/simulation approach, in InterSense 6: Proceedings of the first international conference on Integrated internet ad hoc and sensor networks. New York, NY, USA: ACM, 26, p. 2. [5] T. Facchinett and G. Buttazzo, Integrated wireless communication protocol for ad-hoc mobile networks, in International Workshop on Real-Time Networks (RTN), Catania, Italy, June 24, pp. 43 46. [6] J. A. Stankovic, Research challenges for wireless sensor networks, SIGBED Rev., vol. 1, no. 2, pp. 9 12, 24. [7] K. C. Chen, Medium access control of wireless lans for mobile computing, IEEE Network, vol. 8, no. 5, pp. 5 63, 1994. [8] E. Egea-Lopez, J. Vales-Alonso, A. S. Martinez-Sala, J. Garcia-Haro, P. Pavon-Marino, and M. V. B. Delgado, A wireless sensor networks mac protocol for real-time applications, Personal Ubiquitous Comput., vol. 12, no. 2, pp. 111 122, 28. [9] R. Cunningham and V. Cahill, Time bounded medium access control for ad hoc networks, in Proceedings of ACM Principles of Mobile Computing, 22. [1] R. Jianqiang, J. Shengming, and H. Dajiang, A token passing tree MAC scheme for wireless ad hoc networks to support real-time traffic, in NETWORKING : Proceedings of the IFIP-TC6/European Commission International Workshop on Mobile and Wireless Communication Networks. London, UK: Springer-Verlag, 2, pp. 47 57. [11] L. Donatiello and M. Furini, Ad hoc networks: A protocol for supporting QoS applications, in Proceedings of the IEEE International Parallel and Distributed Processing Symposium. IEEE Computer Society, 23. [12] Omnet++ discrete event simulation system, www.omnetpp.org.