Scoop: a hybrid, adaptive storage policy for sensor networks

One problem with existing store-and-query sensor networks is that they fail to take data and query rates or network topology information into account. This leads to expensive (and avoidable) communication overhead that reduces the lifespan of battery-powered sensor networks. Scoop reduces this overhead (up to a factor of four in our experiments) by dynamically creating and adapting an in-network storage policy based on statistics it collects about data, queries, and network conditions. Whereas existing in-network storage techniques are often at the extreme ends of the storage policy spectrum (e.g., store all data externally on a basestation, or store all data locally), Scoop’s storage policy allows it to adapt between these extremes depending on the situation. The intuition behind our scheme is that data should be stored close to where it is needed, if it is needed. If query rates are high relative to the rate of data production, Scoop adapts its storage policy to store data closer to the source of queries. If query rates are low, the policy adapts to store data closer to its source. We have built a complete implementation of Scoop for TinyOS motes [3] and evaluated its performance on a 62-node testbed and in the TinyOS simulator, TOSSIM. Our results show that Scoop not only provides substantial performance benefits over alternative approaches on a range of data sets, but is also able to efficiently adapt to changes in the distribution and rates of data and queries.