Hybrid PUSH-PULL for Data Diffusion in Sensor Networks without Location Information

Xu Cheng, Feng Wang, Jiangchuan Liu

Abstract: PUSH and PULL are two common data dissemination algorithms for data-centric sensor networks. The two algorithms work well with only a few sources or a few sinks, respectively; however, when there are many sources and many sinks, both of them become inefficient. In this paper, we propose a novel location-oblivious hybrid PUSH-PULL data diffusion (LOHD) algorithm, which suits a wide range of networks and source/sink settings. Different from existing hybrid approaches, LOHD does not rely on any location information; it adaptively selects an ultra-node through a well-controlled flooding and the ultra-node maintains the gradients from sources to sinks. It then incorporates enhanced PUSH and PULL to distribute messages along the gradients instead of flooding. We model and analyze the algorithms and perform extensive simulations. The results show that LOHD remarkably outperforms both PUSH and PULL, particularly when the number of sources and sinks increases. We also show that the overhead well resists to such increase, suggesting LOHD is highly scalable.

External IDs:dblp:conf/icc/ChengWL08