Simulation of Gradient-Based Data Dissemination in Wireless Sensor Networks

Simulation of Gradient-Based Data Dissemination in Wireless Sensor Networks

Jun Li, Mani Srivastava, Peter Reiher

Computer Science Department Electrical Engineering Department

University of California, Los Angeles

1. Communication paradigm of sensor network

Sensor network has many characteristics different from conventional computation and communication models. More important, a user making queries of an environment via sensor network mainly cares that his query is forwarded to the sensor that has the data with specific attributes, not the identity or address of the sensor nodes at all. The address based packet routing does not fit well of the data-centric communication paradigm of sensor networks[1].

Gradient-based data dissemination for sensor networks has been proposed as the communication model for sensor networks[1]. The sensor network is composed of very large number of sensors distributed arbitrarily to collectively sense some information. It is a two-phase strategy. The first phase is interest propagation phase, in which the interest on particular data is propagated to reach sensor nodes that have data satisfying the interest. Meanwhile the gradients on some links are established. The interest initiator is called sink node. The second phase is the data propagation phase, in which the data matching the interest propagates back to the sink node(s) according to the gradient value of links from node to node.

2. Gradient establishment and utilization

The unsolved problem for the gradient-based data dissemination in sensor network is that how to establish the gradient value and how to utilize them for data dissemination. We invented three gradient establishment strategies and found them interesting – direction-based, distance-based and gravity-based. The direction-based scheme aims to forward data at each intermediate node with least deviation from the straight line toward the sink. The distance-based scheme aims to make data flow to the neighbor nearest to the sink. And in the gravity-based scheme the data flows towards the sink as if the sink node is exerting a gravitation pull. In each strategy the data flows along the link(s) with greatest gradient(s).

3. Simulation of gradient-based dissemination

PARSEC[2] language is used to simulate the data dissemination in wireless sensor networks. Each different gradient-based scheme is modeled. Moreover, since each sensor is energy-sensitive, each gradient establishment with energy as a factor is also simulated.

Sensors are divided into partitions and each partition is an entity. The interest or data propagation from one sensor to another has to be done by message passing from partition to partition. This is important for scalability – letting each sensor be an entity is just too resource consuming. There are also other messages defined for simulation purposes. Each sensor also maintains a structure including topology, transmission radius, energy amount, interest information, statistical information, gradient table, and so on.

The evaluation criteria for comparing each different scheme include data availability (whether the data flow can hit sinks), latency and the energy consumption. We made some assumptions on sensor networks. We assume constant power for reception and transmission (ratio 1:3), constant size of interest and data (128 bytes and 1024 bytes), and constant transmission rate (1MB/s).

Different topologies are used for simulation. The result shows that the distance-based strategy consumes the least amount of energy, and the direction-based the most. The reason is that the former usually goes through the least number of hops to reach the sink. But in a mesh topology the result for different strategies are basically same.

Furthermore, with energy as an additional factor in calculation gradient, if a neighbor has less energy, the possibility of data flowing to this neighbor is also lower, i.e. lower gradient value, preventing that neighbor from burning out. The energy consumption is fairer this way and the whole sensor network is expected to have longer lifetime.

4. Gradient aggregation and its simulation

At each sensor node, instead of maintaining a gradient value toward its neighbor for each sink, the gradient values for multiple sinks can be aggregated and replaced by just one gradient value. The data dissemination based on the aggregated gradient is also simulated. Whereas the data availability might be compromised and the energy consumption for interest propagation phase is still the same, the energy consumption for data propagation phases can be greatly decreased.

5. Conclusion

Gradient-based data dissemination is a promising communication paradigm for wireless sensor networks. Our simulation using PARSEC language shows many interesting results. Three different strategies on establishing and utilizing gradients are compared in simulation, with or without taking energy consumption as a factor. Aggregation over multiple sinks when calculating gradients is also simulated to compare the energy consumption.

References

[1]Estrin, D., Govindan, R., Heidemann, J., and Kumar, S. “Next Century Challenges: Scalable Coordination in Sensor Networks”, ACM MobiCom 99, August 1999, Seattle, WA.

[2]UCLA Parallel Computing Laboratory. PARSEC – Parallel Simulation Environment for Complex Systems.

[3]Broch, J., Maltz, D., Johnson, D., Hu, Y. Jetcheva, J., “A performance comparison of multi-hop wireless ad hoc network routing protocols” Proceedings of the Mobicom’98 Conference, Dallas, Texas, pp. 85-97, October 1998.

[4]Kulik, J., Rabiner, W., Balakrishnan, H., “Adaptive Protocols for Information Dissemination in Wireless Sensor Networks”.Proc. 5th ACM/IEEE MobicomConference, Seattle, WA, August 1999. Also MIT-LCS-TR-772 (Feb. 1999).

[5]Bagrodia, R. and Takai M. “Position Paper on Validation of Network Simulation Models”, DARPA/NIST Network Simulation Validation Workshop, May 1999.

[6]Zeng, X., Bagrodia, R., Gerla M., “GloMoSim: a Library for Parallel Simulation of Large-scale Wireless Networks”, Proceedings of the 12th Workshop on Parallel and Distributed Simulations -- PADS '98, May 26-29, 1998 in Banff, Alberta, Canada.

[7]Liu, W., Chiang, C., Wu, H., Gerla, M., Jha, V., Bagrodia, R., “Parallel Simulation Environment for Mobile Wireless Networks”, Proceedings of the 1996 Winter Simulation Conference - WSC '96, Coronado, CA, 1996, pp. 605-612.

[8]Bagrodia, R., "Parallel Languages for Discrete-Event Simulation Models", IEEE Computational Science and Engineering, Apr-Jun 1998, pp. 27-38.

[9]Bagrodia, R., Meyer, R., Takai, M., Chen, Y., Zeng, X., Martin, J., Park, B., Song, H., “Parsec: A Parallel Simulation Environment for Complex Systems", Computer, Vol. 31(10), October 1998, pp. 77-85.