Hop-by-Hop based Reliable Congestion Control Protocol for Wireless

Document Sample
Hop-by-Hop based Reliable Congestion Control Protocol for Wireless Powered By Docstoc
					Hop-by-Hop based Reliable Congestion Control Protocol for Wireless Sensor Networks
Hyun-tae Kim Ravi Sankar Jung-sik Lee and In-ho Ra
School of Electronic and Information Eng. Kunsan National University Kunsan, Jeonbuk, Korea Email:{leejs, ihra@kunsan.ac.kr} JEIT-BK21 Department of Electrical Eng. Chonbuk National University University of South Florida Jeonju, Jeonbuk, Korea Tampa, Florida, USA Email:camelk@dcs.chonbuk.ac.kr Email:sankar@eng.usf.edu
Abstract— In a wireless sensor network, due to huge amount of small data packets, network congestion is easily happened. So it requires an efficient control protocol to provide reliable data transfer and energy conservation while avoiding network congestion. In this paper, we propose a HRCCP (Hop-by-Hop based Congestion Control Protocol) to guarantee both a reliable transfer in data accuracy and minimum consumption of energy waste by using Hop-by-Hop sequence number and DSbACK (Delayed and Selective ACK, Buffer Condition) scheme. Finally, it shows that data reliability and energy efficiency is enhanced by the proposed HRCCP protocol with the simulation results performed on TinyOS based network platform announced by UC Berkely.

I. INTRODUCTION Wireless sensor network (WSN) is one of the emerging research areas that provides designated services such as disaster prevention, environment monitoring, medical monitoring, habitat monitoring, military surveillance, inventory tracking, intelligent logistics, and structural health monitoring by a system dealing with various types of sensing data acquired from sensors widely deployed in a target area. The main characteristic of sensor networks is that it requires specific methodologies for aggregating and processing huge amount of sensing data in real-time. It uses an intelligent clustering method to deliver the sensed data originating from each independent sensor deployed in the same interest area along the shortest path on the self-organized interconnection topology for the dedicated wireless sensor network [1]. In general, hundreds or thousands of different types of tiny sensors are randomly deployed in a wide target area for environmental monitoring and meaningful data for a query generated by a server should be delivered to a sink node in time so that WSNs do not suffer from high delay. Usually, low energy single-hop or multi-hop wireless sensor network is used for the purpose of reliable real-time data transfer [2]. Due to many limitations on a sensor such as power, computational capacity, and memory, many research studies have been performed on how to maximize the entire life-time of a wireless sensor network in order to reduce the unnecessary energy consumption by activating just a few sensor nodes with high energy level among others deployed in the same event area or by not delivering duplicated data packets based on aggregation scheme that may save a lot of energy wastes from unnecessary data transfers. But, in a real situation on data delivery using traditional transport protocol, how to deliver

useful data reliably is very important since a data transfer failure before reaching to a destination node results in its retransmission, and energy is lost due to sensor nodes participating in data delivery to the failure node. Typically, wireless sensors are required to be deployed densely and the number of sensor nodes in a wireless sensor network is large compared to ad-hoc network. In general, sensor nodes are prone to failures and the packet loss rate in a wireless sensor network becomes very high because of deploying huge amount of wireless sensors and also due to fading, noise, and interference problems on wireless signal. Eventually, high possibility of data loss would invoke data retransmission that results in network congestion. Network congestion is a phenomenon where the end-to-end packet delay is getting longer because the overloaded and saturated packets under repeated retransmissions degrade the network performance, therefore reliable data transfer in time would be hardly provided. That is to say, if there are fewer on-going packets and fewer retransmissions, energy can be saved. Therefore, effective congestion control and loss recovery approach can be considered as effective solution to this problem [3]. In summary, the problem of transport control protocols for wireless sensor networks is how to effectively control congestion control and how to guarantee reliability while simultaneously conserving max energy. Consequently, to handle the problem mentioned above, a method for delivering packets reliably in time and in sequence according to packet sequence number is required to reduce the total amount of retransmitted packets through a wireless sensor network. This paper aims to present an energy conservation by effective congestion control in wireless sensor networks by minimizing unnecessary retransmissions while guaranteeing reliable data transfer based on Hop-by-Hop sequence number and DSbACK(Delayed and Selective ACK, Buffer Condition) scheme. The remainder of the paper is organized as follows. Section II describes problem definitions and disadvantages of existing methods. Section III presents the features, functional procedures of our work in detail. Section IV provides the results of performance evaluation compared to PSFQ and SenTCP performed on TinyOS. Finally, Section V concludes the paper. II. PREVIOUS WORK AND PROBLEM DEFINITION In wireless sensor networks, methods for handling problems

on reliable transfer, energy-conservation, and congestion control are treated as most important ones since the resulting network performance and service qualities might be quiet different according to the adopted methods by which the total number of retransmissions, buffer overflow/underflow in an intermediate node, and data accuracy at sink node are determined. Specifically, packet loss is common in WSNs due to bad wireless channel quality, sensor failure, and sudden network saturation. Hence wireless sensor networks should guarantee certain reliability in any corresponding layer of the protocol stack or cooperatively cross each layer through the enforced loss recovery policy in order to abstract correct information. In general, for providing reliability in wireless sensor networks, end-to-end feed back method based on negative acknowledgement (NACK) is widely used that relies on the reduction of delivery probability of a sensor node to achieve higher energy conservation over a wireless sensor network. In a situation where a receiver detects packet losses by checking holes in the sequence number space and requests retransmission of missing packets by using NACK, it is possible to save lots of energy, if piggy-back NACKs are understood as carrying implicit acknowledgement for packets delivered in sequence. But, as shown in Fig. 1, a serious problem can be caused when the NACK for a missing packet is sent from the sink node to the source node along the hop-to-hop down link over a wireless sensor network because the network traffic overhead can be increased by many duplicate NACKs that are traveling back to the same source node along the routing path and not closely relevant to actual retransmission of the missing packet. This problem gets worse when multiple paths are available from a source to a sink node and when the number of intermediate nodes on an enforced path is increased.
Source Node 01 Node 02 Sink

the error recovery procedure of PSFQ.
Source Node 01 Node 02 Sink

1 1 2 3
NACK

1

2 2 2 3 3

Fig. 2. Congestion control process of PSFQ

1 1 2 2 3 4
NACK

1 2

4
K NAC

4
NACK

CODA (Congestion Detection and Avoidance in Sensor Networks) is an upstream congestion protocol that attempts to detect congestion by monitoring current buffer occupancy and wireless channel load [5]. It removes buffer overflow problem by using open-loop hop-by-hop backpressure scheme when the buffer occupancy or wireless channel load exceeds a threshold-based value. Similarly, it decides whether congestion happens by estimating the status of a channel load so that it is possible to perform sophisticated congestion control. But the process for calculating the current local congestion degree of an intermediate node is complex and time consuming work that requires high energy costs. On the other hand, SenTCP is a possible choice for reducing such a high computing complexity by jointly using average local packet service time and average local inter-arrival time in order to estimate current load congestion degree in each intermediate sensor node. It provides higher throughput and good energy-efficiency because it obviously reduces packer dropping. However, SenTCP can handle only congestion and guarantee no reliability in data transfer [6]. To handle the problem mentioned above, this paper presents a hop-by-hop based reliable congestion control protocol that can reduce both the computing complexity of estimate algorithm and the increase in end-to-end transfer delay caused by applying congestion control. It proposes a method for accomplishing energy consumption saving and data transfer reliability by using DSbACK feedback message that contains useful control information such as hop-by-hop sequence number, current delay in local packet transfer, and buffer status.

Fig. 1. NACK based end-to-end congestion control process.

To solve this problem, the PSFQ (Pump Slowly, Fetch Quickly) protocol has presented where an intermediate node dose not forward subsequent packets after a packet loss until that packet is completely recovered [4], and all such delayed packets are entered in a buffer at the intermediate node waiting to be forwarded further. It takes an advantage of enabling fast detection and recovery from missing packets, and accordingly it can produce an effect on energy saving by not sending subsequent packets and by eliminating unnecessary NACKs. But it causes another problem on increasing the possibility of buffer overflow in intermediate nodes and extending the end-to-end transfer delay for a block of packets. Fig. 2 shows

III. DSBACK BASED CONGESTION CONTROL In this section, the basic concept and operational procedures of the proposed method are described in detail. It also explains how to add hop-by-hop sequence number to end-to-end sequence number and how to control data transfer and congestion using DSbACK based feedback message. A. Hop-by-hop sequence number In general, the transport control protocol for WSNs should provide congestion control mechanism and guarantee reliability. Usually, the most data streams are sent from sensor

nodes to sink so congestion might occur around the sink node and hop-by-hop mechanism is widely used for congestion control and loss recovery since it can reduce missing packets and conserve energy. In this paper, we propose a congestion control method to speed up the delay time in end-to-end transfer using a pair of end-to-end sequence number and hop-by-hop sequence number by which loss recovery is performed on missing packets at hop-by-hop level.
Source Node 01 Node 02 Sink

1 1 1 1 2 2 3 3
NACK

1 1

3 2 3 2

also might be greatly decreased. A DSbACK is sent to the previous node when the following conditions are satisfied: 1) an error packet is received, 2) a timer is out 3) the degree of buffer occupancy exceeds the predefined threshold. Fig. 4 shows an example of transmission control process based on DSbACK feedback message. A DSbACK message sent back to a sending node offers essential information on arriving packets in delay to adjust the sending rate of packets at a source node and to remove packets stored in a sender’s buffer. And it also informs the status of a packet in delivery when the packet encounters a failure or when it is discarded by buffer overflow. In the proposed method, it decreases the transfer rate exponentially every time when a DSbACK feedback message is sent from an intermediate node whereas it increases the transfer rate linearly when network congestion becomes lessened.

2 4 2 3 2 3

IV. PERFORMANCE EVALUATION In this section, the simulation results performed on a test bed using TinyOS[7] by UC Berkeley, TOSSIM[8], and component based nesC[9] language are described. A. Simulation environment The network topology used for simulation is a 4 X 4 grid WSN shown in Fig. 5 and simulation parameters are shown in Table I.

Fig. 3. Transmission control using hop-by-hop sequence number

End-to-end sequence number is used to reassemble disorderly arrived packets at the sink node in sequence and hop-by-hop sequence number is used only between hops, which eliminates unnecessary retransmissions requested by end-to-end loss recovery procedure. With the hop-by-hop sequence number, an intermediate node can forward a received packet immediately to the next upstream node further and reduce the overall transfer delay by requesting the retransmission only for the missing packet to the previous node. B. DSbACK based congestion control In the proposed method, for loss recovery and congestion control, DSbACK feedback message is sent to the previous intermediate node by piggyback and information such as the degree of buffer occupancy and the range of packet loss is included in it. That is to say, DSbACK message can reduce the total number of transmissions between adjacent intermediate nodes by combining information on congestion control with loss recovery so that the total communication costs at

Fig. 5. A 4 X 4 gird network represented by TinyViz
S o u rc e N ode 01 Node 02

TABLE I SIMULATION PARAMETERS

1

1 1 1

2 3 4

2 3 4
A DSb CK

2 3 4 2

2 3 4 5
A DSb CK

Parameter Type No. of nodes No. of sink node Packet size Buffer Size Feedback message Size of sequence number

Test Value 16 (4 x 4 grid network) 1 500 – 1500 Bytes Maximum 10 packets DSbACK 2 Bytes(E2E, HbH sequence number)

Fig. 4. An example of transmission control using DSbACK feedback messages

intermediate nodes can be reduced and energy consumption

B. Performance analysis The main purpose of the simulation is to measure the total

number of feedback messages requesting packet retransmission, transfer delay, and packet loss rate. Fig. 6 shows that hop-by-hop transfer delay can be reduced by the proposed method compared to PSFQ. The transfer delay is increased exponentially when the error rate exceeds 40%.

That means it can prevent a failure packet from being forwarded to the next upstream node along the routing path. And by using the proposed method, packet loss rate can be reduced compared to SenTCP like shown in Fig. 8. V. CONCLUSION Because sensor node may be deployed in harsh environmental conditions, unexpected node failure and packet loss are common in WSNs. Therefore, achieving reliable transmission between sensor nodes over multiple hops with wireless channel errors, collisions, or congestion is very important in WSNs. In this paper, we propose a hop-by-hop based congestion control protocol using DSbACK for ensuring as fewer packets dropping as possible and for reducing unnecessary retransmissions as well as deceasing transfer delay through a given WSN. For the hop based congestion control, it uses end-to-end sequence number and hop-to-hop sequence number in order to adjust packet transfer rate at a source node and to reduce the possibility of buffer overflow. Furthermore, the total amount of feedback messages and retransmitted packets can be decreased by the proposed loss recovery method. Finally, we evaluate the proposed protocol to compare the performance against PSFQ and SenTCP with a simulation performed on TinyOS and TOSSIM.

Fig. 6. Comparison of the degree of increase in transfer delay

REFERENCES
[1] [2] M. Weiser, “Some Computer Science Issues in Ubiquitous Computing,” Communications ACM, Vol. 36, No. 7, pp. 75-84, Mar. 1993. J. M. Kahn, R. H. Katz and K.S.J. Pister, “Next Century Challenges: Mobile Networking for Smart Dust,” In Proceedings of the 5th Annual ACM/IEEE International Conference on Mobile Computing and Networking, pp. 263-270, Aug. 1999. I. F. Akyildiz and S. Weilian, Y. Sankarasubramaniam and E. Cayirci, “A Survey on Sensor Networks,” IEEE Communications Magazine, Vol. 40, No. 8, pp. 102-114, Aug. 2002. C. Y. Wan, A. Campell and L. Krishnamurthy, “PSFQ: A Reliable Transport Protocol for Wireless Sensor Networks”, In Proceedings of the Int'l Workshop on Sensor Networks and Arch(WSNA), Sept. 2002. C. Y. Wan, S. B. Eisenman, and A. T. Campbell, “CODA: Congestion Detection and Avoidance in Sensor Networks,” In Proceedings of ACM Sensys’03, Los Angeles, USA, Nov. 2003. C. Wang, K. Sohraby and B. Li, “SenTCP: A Hop-by-hop Congestion Control Protocol for Wireless Sensor Networks,” In Proceedings of IEEE INFOCOM 2005 (Poster Paper), Miami, Florida, USA, Mar. 2005. TinyOS Homepage[Online]. Available: http://webs.cs.berkley.edu/tos/ P. Levis, N. Lee, M. Welsh and D. Culler, “TOSSIM: Accurate and Scalable Simulation of Entire TinyOS Application,” In Proceedings. of ACM Sensys, 2003, Nov. 2003. D. Gay, P. Levis, R. V. Behren, M. Welsh, E. Brewer and D. Culler, “The nesC Language: A Holistic Approach to Network Embedded Systems,” In Proceedings of Programming Language Design and Implementation, Jun. 2003.

Fig. 7. Comparison of the number of control messages

[3]

As shown in Fig. 7, the total number of feedback messages is considerably decreased in the case of the proposed DSbACK compared to using NACK and typical congestion control message. And the number of repair messages is decreasing while the number of hops from a source node to a sink node is increasing.

[4]

[5]

[6]

[7] [8]

[9]

Fig. 8. Comparison of packet loss rate on packet error probability