CDTS: Coordinator Data Traffic Shunt model for Zigbee networks

Transcription

1 CDTS: Coordinator Data Traffic Shunt model for Zigbee networks Chinyang Henry Tseng 1*, Shaiuhuey Wang 2, Bor-Shing Lin 1, Tong-Ying Juang 1, Xiao-Ru Ji 1 1 National Taipei University, New Taipei City, Taiwan {tsengcyt, bslin, juang }@mail.ntpu.edu.tw, s @webmail.ntpu.edu.tw 2 Chung Yuan Christian University, Jungli, Taiwan anglenew@cycu.edu.tw Abstract. Zigbee, a wireless sensor network, is expected to have a explosive growth in the use of wireless control and monitoring applications because it has the advantages of low cost and easy deployment. As Zigbee is applied to more emerging applications, scalability becomes a critical issue: all sensor data is sent to the coordinator before forwarding to the sink node, and obviously the coordinator becomes the bottleneck. This paper proposes Coordinator Data Traffic Shunt (CDTS) performing data traffic shunt feature to reduce Coordinator s traffic. CDTS group consists of CDTS routes and forwards data directly to Sink instead of to the coordinator. To ease CDTS router implementation without modifying Zigbee standard, CDTS layer is added between Media Access Control (MAC) and network (NWK) layers in Zigbee stack. CDTS layer intercepts packets and redirect them to the Sink node without involving the coordinator. We implement CDTS routers in TI CC2530 Zigbee platform and NS2 simulation. Experiment results demonstrates CDTS provides better packet deliver ratio and lower coordinator loading than original Zigbee stack. Thus, CDTS successfully resolves Zigbee bottleneck problem and is fully compatible with current Zigbee stack design. Keywords: Zigbee, Coordinator, Coordinator Data Traffic Shunt (CDTS), Personal Area Network (PAN) 1 Introduction ZigBee is a wireless network standard and is expected to have an explosive growth in wireless control and monitoring applications because of its low cost and low power consumption. Popular applications include home and building automation, Industrial control, Embedded sensing, Medical data collection and warning systems. ZigBee standard becomes complete by adding four main components: network layer, application layer, ZigBee device objects (ZDOs) and manufacturer-defined application objects for customization and total integration. ZDOs, the most significant improvement, are responsible for keeping device roles, management of requests to join a network,

2 device discovery and security. Zigbee supports star, tree and mesh network topologies and, for all topologies, ZigBee must have one coordinator device responsible for a Personal Area Network (PAN) creation, parameter control and fundamental maintenance. Zigbee uses Ad-hoc On-demand Distance Vector (AODV) routing protocol to automatically construct an ad-hoc network as routes needed. Nodes not in use are in the sleep mode but can be woken up to the active mode around 3 seconds when they are needed for data collection and transmission. This on-demand characteristic results in low power consumption and long battery life. Fig. 1. Sensor data traffic in Zigbee network In Zigbee, a Coordinator forms a PAN and is responsible for forwarding all the sensed data in that PAN to Sink, the backend server. As Zigbee is applied to more emerging applications requiring intensive data transmission, obviously, the Coordinator becomes the bottleneck and scalability issue becomes critical. With a larger scale of the Zigbee network, if the coordinator cannot handle the large amount of data, data loss and transmission will occur and even worse, the whole network becomes inoperable. The failure of network operation is unacceptable for time-sensitive applications, especially for real-time monitoring systems. For example in Fig. 1, Zigbee sensor nodes constantly send large amount of sensor data and therefore the Zigbee coordinator must have high data capacity. In addition, if the sensor data comes from critical biomedical systems, such as cardiovascular system, data transmission failure may cause emergency warning misses and life support delay. For these time critical applications, traditional Zigbee transmission mechanism needs to be improved to enhance the transmission reliability of coordinator as the Zigbee scale and the amount of sensed data becomes larger and larger.

3 To solve the Coordinator bottleneck problem, this paper proposes Coordinator Data Traffic Shunt (CDTS) which performs sensor data traffic shunt feature to reduce Coordinator s traffic. CDTS group consists of CDTS routers linked with the sink node through other Direct Communication Links (DCLs), such as RS232 or Bluetooth. CDTS group forwards data directly to the sink node instead of the coordinator. To ease the integration of CDTS feature in a Zigbee router without modifying Zigbee standard, CDTS layer is added between Media Access Control (MAC) and network layers in Zigbee stack. The CTD layer intercepts the packets from network layer and modifies the destination address to be the CDTS router itself instead of the coordinator. CTDS group redirects sensor traffic to the sink node without involving the coordinator. We implement CDTS routers in TI CC2530 Zigbee platform and NS2 simulation. Experiment results show CDTS provides better packet deliver ratio and scalability as well as lower coordinator loading than original Zigbee stack while increasing the throughput at the sink node. Thus, CDTS successfully resolves Zigbee bottleneck problem and is fully compatible with current Zigbee stack design. This paper is organized as follows: Section 2 describes analysis, requirements, and related works. Section 3 presents the model design. Section 4 describes experiment results. Section 5 concludes the work. 2 Analysis and Requirements 2.1 Zigbee Network Analysis In order to modify Zigbee data process procedures, it is required to analyze Zigbee stack processes of network and data initiation. First, in order to form a Zigbee Personal Area Network (PAN), this Zigbee PAN has a coordinator node, which is in charge of PAN initiation. The coordinator first selects a PAN ID and broadcasts beacon requests with this ID, with which allows other routers and end devices can join this PAN network. Then routers broadcast beacon requests for other remote Zigbee nodes such that all Zigbee nodes join this PAN network. Thus, all nodes are capable to reach and send their sensor data to the coordinator. Fig. 2. Zigbee ACK processing in APS and MAC layers

4 When nodes begin to send sensor data to the coordinator, they usually require to establish clusters binding with the coordinator. Zigbee application support sublayer (APS) is responsible to establish the clusters, and the APS layer at the destination is required to send back APS acknowledge (ACK) messages to avoid data loss, as illustrated in Fig. 2, (step 1, 3, 5, 7). Because of high data loss rate in wireless environments, Zigbee MAC layer, IEEE , requires all Zigbee intermediates nodes sending MAC ACK frames to prevent data loss for APS data flows, as shown in Fig. 2 (step 2, 4, 6, 8). After receiving APS ACK, Zigbee nodes can complete cluster binding with the coordinator for sending sensor data. Thus, if Zigbee data process procedures are modified, cluster binding process and its APS ACK flow have to be maintained. 2.2 Requirement. In Zigbee PAN networks, Zigbee nodes sending sensor data with cluster bindings with the coordinator. If the cluster data traffic increases a lot, the coordinator will easily become the network bottleneck. In addition, the high traffic data load may cause the coordinator running out of battery or malfunction, which will paralyze the whole Zigbee PAN network. Thus, coordinator traffic loading becomes a major Zigbee network scalability issue, denoted as R 1. To address this scalability issue, the desired solution should satisfy the following three requirements, denoted as R 1A, R 1B, R 1C, respectively. First, minimizing the coordinator traffic loading is required, denoted as R 1A, even though the sensor data traffic toward the sink node is increasing. Second, the flexibility of the desire solution deployment is required, denoted as R 1B, so the PAN network data traffic topology can be easily expanded. Third, since Zigbee data traffic delivering relies on cluster bindings with the coordinator, seamless maintenance of cluster binding is required, denoted as R 1C. Because Zigbee embedded solutions usually do not provide source codes for the implementations of lower layers, such as MAC, Network (NWK), and APS layers, the desired solution is required to be compatible with these existing implementations without modifying them, denoted as R 2. For example, IEEE , AODV (Ad hoc On-demand Distance Vector) are Zigbee standard solutions for MAC and NWK layers respectively so they cannot be modified, denoted as R 2A. Besides, Zigbee sink applications are usually customized built according to application demands because they are usually black boxes for us. Thus, the desired solution is required to be fully compatible with all kinds of sink applications without modifying them, denoted as R 2B. 2.3 Related works Here we discuss related works that intend to enhance data delivering for Zigbee networks. Direct Diffusion [1] is a datacentric and query driven communication protocol. Its sink node must send queries to retrieve sensor data. This design may save periodic data sending but miss desirable data, and it requires the sink node cooperation and extra data queries. LEACH [2] is a two level data transfer protocol, and all its data must be sent through cluster heads. Xianghua Xu et al. [3] proposed an

5 enhanced tree-based routing algorithm for reducing AODV route discovery overhead. Ruixia Liu et al. [4] proposed a new cluster routing protocol, called AODV cluster, to reduce energy consumption in a large scale of AODV network. These two works require modify AODV design in order to enhance AODV performance in Zigbee networks. Kartinah Zen et al. [5] proposed a load balancing scheme for child and parent coordinators in case some coordinators have connected with excessive neighbors. Zheng Sun et al. [6] proposed a new routing protocol to balance routing overhead in Zigbee networks based on AODV design. Kuei-Li Huang et al. [7] proposed a load balancing mechanism, called controller assisted distributed (CAD) for Zigbee networks to balance network loading between PAN networks. Most of these works require modify current Zigbee stack design. Although they can enhance Zigbee performance in certain scenarios, they cannot provide a generic solution for the bottleneck issue resulting from current Zigbee coordinator design. 3 CTDS Design 3.1 CDTS Group To address R 1 in section 2.2, we introduce Coordinator Data Traffic Shunt (CDTS) group to solve this problem. CDTS group consists of CDTS routers, which have direct communication links (DCLs) linking with the sink node. While receiving data packets from end devices toward the coordinator, CDTS routers forward the packets to the sink node by DCLs instead of the coordinator, presented in Fig. 3. Fig. 3. CDTS Group

6 Because CDTS routers can handle lots of data flows in parallel without going through the coordinator, the CDTS group is able to mitigate network traffic burden of the coordinator, and thus R 1A is resolved. In addition, by having more CDTS routers, a CTDS group can handle more data traffic from Zigbee nodes, and thus R 1B is resolved. 3.2 CDTS Layer To address R 2 in section 2.2, we introduce CDTS layer to solve this problem. CDTS layer is designed for CDTS routers in order to redirect the data packets to the sink without modifying Zigbee MAC and NWK layer. As illustrated in Fig. 4, CDTS layer intercepts data packets between MAC and NWK layers for perform data traffic shunt. It does not modify MAC and NWK layers in Zigbee stack so R 2A is satisfied. Fig. 4. CDTS Layer in Zigbee stack While CTDS layer processes the data packets toward the sink, it modifies the PAN destination address to be the CDTS router itself instead of the coordinator, as illustrated in Fig. 5. Because the destination becomes itself, NWK layer forwards the data packets to its upper layer, APS layer, instead of sending them to the coordinator. Fig. 5. CTDS Layer modifies PAN destination address

7 As illustrated in Fig. 6, the APS layer starts to build a cluster binding with the Zigbee node of source PAN address and sends back a ACK message without notifying the coordinator. While receiving sensor data by the binding cluster, the CDTS router then forwards the data to the sink by its DCL. Thus, the CDTS router seamlessly handles data traffic for the coordinator so R 1C is resolved. In addition, the sink applications receive data packets as usual without being modified so R 2B is resolved. Fig. 6. Seamless cluster binding at the CDTS router 4 Evaluation To evaluate our model, we implement the model in TI CC2530 Zigbee platform and Network Simulator 2 (NS2) simulation. The simulation is based on IEEE and traffic sources are Constant Bit Rate (CBR). The evaluation of our work is based on the following performance metric: Packet Deliver Ratio (%): packets sent from Zigbee sensor nodes / packets received at the sink node. Coordinator Loading (bps): packets processed at the coordinator in bps. Sink Throughput (bps): packets received at the sink node in bps. 4.1 Embedded system We use the CC2530 ZigBee Development Kit with ZigBee2007/PRO. The development kit comes with four ZBDC51 module and ZBDC51MB development boards. We use IAR Embedded Workbench software to compile the source code to create a hex file which downloads through the burner to RAM of the CC2530 chip. The communication between the host computer and the CC2530 Zigbee platform is RS232 serial port at baud rate of K bps. The experiment network topology is 4 node square in 100*100 meter size. The experiment duration is set to 30 seconds. The number of data transmission rate from 7.5k to 47.3 kb/s according to the payload size.

8 Coordinator loading (bps) Packet Delivery Ratio ( %) 100% 90% 80% 70% 60% Originator Zigbee stack CDTS Sink throughput ( bps ) Fig. 7. Packets Delivery Ratio As illustrated in Fig. 7, packet delivery ratio in the original Zigbee stack drops a lot while sink throughput increases from 50 K bps, but the same ratio in CTDS stays above 80%. This shows CDTS group can perform great data traffic shunt so packet deliver ratio does not drop as the throughput demand increases. However, since all data packets go to the coordinator in the original Zigbee stack, the ratio cannot sustain while the throughput demand increases. Thus, packet deliver ratio provides direct evidences that CDTS group does a great job for data traffic shunt for Zigbee networks Originator Zigbee stack CDTS Sink throughput (bps) Fig. 8. Coordinator loading As illustrated in Fig. 8, coordinator loading in the original Zigbee stack increases a lot while sink throughout increases from 60K bps, but the same loading value in CDTS stays below 20 bps. Obviously, increasing sink throughput results in extremely high network burden at the coordinator in the original Zigbee stack since all traffic must go through the coordinator. In CDTS, because of data shunt feature, increasing sink throughput does not result in the heavy network burden at the coordinator. This indicates that the coordinator is not the network bottleneck any more, and it can last much longer time, which means the whole PAN network can survive longer time.

9 Packets Delivery Ratio (%) 4.2 NS2 We also perform similar experiments in NS2 simulation environment for scalability tests. The network size is 200*200m with transmission range as 30 meter. The duration of the simulation is set to 100 seconds. The CDTS group has 3 CDTS routers, and the CBR rate is 28 K bps at each Zigbee senor node. As illustrated in Fig. 9, packet delivery ratio in the original Zigbee stack starts to drop while the number of nodes increases from 4, but the same ratio in CDTS remains relatively high. This shows CDTS model is much more scalable than the original Zigbee stack because of CDTS data shunt feature. 100% 90% 80% 70% 60% 50% 40% originator Zigbee stack CDTS Number of nodes Fig. 9. Packets Delivery Ratio with 3 CDTS routers 5 Conclusion Zigbee has lots of emerging sensor application because it has the advantages of low cost and easy deployment. Scalability becomes its major problem due to its coordinator design. First, we analyze Zigbee stack design and identify two groups of requirements for the desirable solution. Second, we propose Coordinator Data Traffic Shunt (CDTS) model and show CDTS successfully resolve the requirements. CDTS group performs data traffic shunt feature that forwards sensor data to the sink node without going through the coordinator. CDTS layer in the CDTS routers intercept packets and change the PAN destination address to be CDTS router itself such that APS layer can easily establish cluster binding from the sensor node toward the CDTS router itself instead of the coordinator. This clear and effective design of CDTS model thus provide a scalable solution for Zigbee sensor network and is fully compatible with existing Zigbee stack. At last, we implement CDTS model in TI CC2530 platform and NS2 simulation. Experiment results show CDTS provides better packet deliver ratio and scalability as well as lower coordinator loading than original Zigbee stack while increasing the throughput at the sink node. Thus, CDTS successfully resolves Zigbee bottleneck problem and is fully compatible with current Zigbee stack design.

10 Reference 1. C. Intanagonwiwat, R. Govindan and D. Estrin, Directed diffusion: A scalable and robust communication paradigm for sensor networks, Proceedings of the 6th annual international conference on Mobile computing and networking, pp , August R. C. Shah and J. M. Rabaey, Energy aware routing for low energy ad hoc sensor networks, Wireless Communications and Networking Conference, pp , March Xianghua Xu, Daomin Yuan, Jian Wan, An Enhanced Routing Protocol for Zigbee/IEEE Wirless Networks, Second International Conference on Future Generation Communication and Networking, pp , December Ruixia Liu, Qiang Guo, Yong Fu, Xianglong Kong, A New Cluster Routing Strategy Based on ZigBee, International Conference on Web Information Systems and Mining, pp , October Kartinah Zen, Halikul Lenando, Mohamad Nazim Jambli, Load balancing based on nodes distribution in mobile sensor network, International Conference on IT in Asia (CITA), Pp. 1-6, July Zheng Sun, Xiao-guang Zhang, Dianxu Ruan, Hui Li, Xun Pang, A Routing Protocol based on Flooding and AODV in the ZigBee Network, International Workshop on Intelligent Systems and Applications, pp. 1-4, Kuei-Li Huang, Chien-Chao Tseng, Jui-Tang Wang, Tsung-Hsi Yang, A Controller- Assisted Distributed (CAD) Load Balancing Scheme for ZigBee Networks, International Conference on Parallel Processing Workshops, pp. 1-5, September 2011.