We have conducted some simulations to verify our results. Unless otherwise indicated, the simulation environment contains 400 sensor nodes randomly de-ployed in 200 × 200 field, each with a transmission range of 25. The packet arrival rate is 1/50 per node and each packet has a random priority between 0 and 1000. The BA is within 10 hops from the sink. All results are from the average of 50 test runs. Fig. 5.1 shows a snapshot of priority distribution in a network with the sink at (0, 0) after applying DSMS.
We first compares DSMS using a mesh-like communication graph with a tree structure. To construct a tree structure, we reduce the communication graph into a short-path spanning tree rooted at the sink. Each node only allow to exchange packets with its parent or children. The results are shown in Fig. 5.2. Using communication graph can collect much higher priority packets than using tree
x coordinate
Figure 5.1: A snapshot of priority distribution.
structure.
Fig. 5.3 compares the average priorities of the packets collected by mobile mules under various stop-by intervals of mules and collected sizes (in terms of numbers nodes) of mules. We compare DSMS against Greedy Forward (GF), where a node always tries to send its packets to any node closer to the sink until the latter has no storage space. OPT represents the ideal solution if global op-timization is possible. Fig. 5.3(a) shows that as the stop-by interval increases, the average priority also increases. Fig. 5.3(b) shows that the average priority decreases slightly as the collecting sizes increases. However, the impact is in-significant.
Fig. 5.4 shows the effect of the BA size. In Fig. 5.4(a), we vary the BA size but fix the stop-by interval such that 1/3 of the data in the BA can be collected.
In terms of the average priority of collected packets, DSMS outperforms GF and
550
Numer of collecting nodes Communication graph
Tree structure
Figure 5.2: DSMS using communication graph and tree structure.
is close to OPT when the hop count is larger than 5. Fig. 5.4(b) shows that as the BA size gets larger, the data loss will decrease since the storage space is larger.
On the other hand, traffic overhead will decrease first and then increase as the BA size gets larger. Each packet transmission counts for one. When the BA size is very small (say, 4 hops), packets have to travel long to reach the BA. For example, a packet with a small priority travel long before being dropped. That causes the traffic overhead keep on decreasing before the hop count reaches 7. However, as the hop count is larger than 7, there are more exchanges in BA, causing the overhead to increase.
Fig. 5.5 shows our DSMS overheads. Fig. 5.5(a) compares the overhead by varying the number of nodes and the packet arrival rate. So DSMS gets packets with higher priorities at the cost of more packet exchanges. The overhead of DSMS is about a constant higher than to that of GF. Fig. 5.5(b) shows the number
0
Number of collecting nodes DSMS
OPT GF
(b)
Figure 5.3: Comparison of average priorities of packets collected by mules by varying (a) stop-by interval of mules and (b) collected size of mules.
0
BA size (hop count) DSMS
Transmission overhead (packet) Packet loss
BA size (hop count) Number of Transmission
Data Loss
(b)
Figure 5.4: Effect of BA size on (a) average priority and (b) traffic overhead and packet loss.
of packet transmissions incurred when a new packet with a random priority is inserted into a stablized network. The transmission increases while the number of nodes increases, but the effect is insignificant.
0
Number of packet transmissions
Numer of nodes DSMS: Arrival Rate 1/10 DSMS: Arrival Rate 1/50 DSMS: Arrival Rate 1/100 GF: Arrival Rate 1/50
(a)
Number of packet transmissions
Numer of nodes Average case of random-priority
Worst case of random-priority
(b)
Figure 5.5: Traffic overheads (a) when packets arrive at arrival rates and (b) when one packet arrives at a random node in a stablized network.
Chapter 6
Implementation
We implemented DSMS in real hardware platform. Our implementation in-cludes a grid WSN and a mule. Fig. 6.1 shows our implementation structure. The WSN has 4 × 4 sensor nodes and a sink. The communication graph is shown in Fig. 6.2. The electric train performs in the role of mule, comes the sink to collect sensory data.
Our sensor hardware platform includes a low power, low cost wireless micro-controller, JN5139 [2] which is implemented our DSMS. We use a light sensor to generate sensory packets with priority ranging from 0 to 9 and display it on a 7-segment display. Nodes will exchange their packets according to the DSMS exchanging rules. When the mule comes to the sink, it will transmit a COL-LECT DATA message to the sink. After collecting data, the mule will transmit an ACK message to the sink and display the number of collecting data on the display.
Figure 6.1: DSMS implementation structure.
Our implementation result shows that the DSMS can be easily used in a real sensor platform. Fig. 6.3 shows that after applying the exchange rules, all nodes will be in-order. Because our DSMS protocol is simple and only needs local information, it is suitable for distributed WSN environment.
g c
h f
k l
d b
j
a
e
i
m
n o p q
Figure 6.2: DSMS implementation communication graph.
Figure 6.3: DSMS implementation result.
Chapter 7
Conclusions
We have proposed a distributed storage management strategy (DSMS) for data buffering in an isolated WSN. DSMS can reduce data loss while keep higher-priority packets closer to the sink area. Properties of DSMS are proved and its efficiency is verified by simulations. In the future, we will implement our protocol on sensor platforms and develop a storage system.
Bibliography
[1] Design and construction of a wildfire instrumentation system using net-worked sensors. http://firebug.sourceforge.net/.
[2] Jennic - wireless microcontroller. http://www.jennic.com/.
[3] Mica wireless measurement system. http://www.xbow.com/.
[4] Terrestrial ecology observing systems. http://research.cens.ucla.edu/.
[5] G. Anastasi, M. Conti, E. Monaldi, and A. Passarella. An adaptive data-transfer protocol for sensor networks with data mules. In Proc. of IEEE International Symposium on a World of Wireless Mobile and Multimedia Networks (WoWMoM), 2007.
[6] A. Chakrabarti, A. Sabharwal, and B. Aazhang. Using predictable observer mobility for power efficient design of sensor network. In Proc. of Int’l Sym-posium on Information Processing in Sensor Networks (IPSN), 2003.
[7] Y. Chen, W. Zhao, M. Ammar, and E. Zegura. Hybrid routing in clustered dtns with message ferrying. In Proc. of ACM/SIGMOBILE Workshop on Mobile Opportunistic Networking(MobiOpp), 2007.
[8] C.-Y. Chong and S. P. Kumer. Sensor networks: evolution, opportunities, and challenges. Proceedings of the IEEE, 91(8):1247 – 1256, 2003.
[9] C.-Y. Lin, W.-C. Peng, and Y-C. Tseng. Efficient in-network moving object tracking in wireless sensor networks. IEEE Trans. on Mobile Computing, 5(8):1044 – 1056, 2006.
[10] L. Luo, C. Huang, T. Abdelzaher, , and J. Stankovic. Envirostore: A cooper-ative storage system for disconnected operation in sensor networks. In Proc.
of IEEE INFOCOM, 2007.
[11] R. Shah, S. Roy, S. Jain, and W. Brunette. Data mules: Modeling a three-tier architecture for sparse sensor networks. In Proc. of IEEE Workshop on Sensor Network Protocols and Applications (SNPA), 2003.
[12] A. Terzis, A. Anandarajah, K. Moore, and I.-J. Wang. Slip surface localiza-tion in wireless sensor networks for landslide prediclocaliza-tion. In Proc. of Int’l Symposium on Information Processing in Sensor Networks (IPSN), 2006.
[13] H. C. Thomas, E. L. Charles, L. R. Ronald, and S. Clifford. Introduction to Algorithms. The MIT Press, 2001.
[14] I. Vasilescu, K. Kotay, D. Rus, M. Dunbabin, and P. Corke. Data collection, storage, and retrieval with an underwater sensor network. In Proc. of ACM Int’l Conference on Embedded Networked Sensor Systems (SenSys), 2005.
[15] W. Zhao, M. Ammar, and E. Zegura. A message ferrying approach for data delivery in sparse mobile ad hoc networks. In Proc. of ACM Int’l Symposium on Mobile Ad Hoc Networking and Computing (MobiHoc), 2004.
[16] W. Zhao and M. H. Ammar. Message ferrying: Proactive routing in highly-partitioned wireless ad hoc networks. In Proc. of the 9th IEEE Workshop on Future Trends in Distributed Computing Systems(FTDCS), 2003.