Similarity search in sensor networks using semantic-based caching

Abstract

Sensor networks build temporary wireless connections in environments where the stationary infrastructures are either destroyed or too expensive to construct. Most of the previous research in sensor networks focuses on routing protocols that adapt to the dynamic network topologies, and not much work has been done on data accessing. One important data accessing application is similarity search, which provides the foundation of content-based retrieval. Many traditional similarity search algorithms are based on centralized or flooding mechanisms, which are not effective in wireless sensor network environments due to the multiple limitations such as bandwidth and power. In this paper we tackle the problem of similarity search by using semantic-based caching to reflect the data content distribution in the network. The basic idea is analyzing the cached results of earlier queries and trying to resolve the later queries within a small collection of content-related mobile nodes. Based on a Hilbert space-filling curve, the data points in a multi-dimensional semantic space are described as a linear representation. These data points are further cached to facilitate query processing. Through extensive simulations, we show that our method can perform similarity search with improved performance in terms of search cost and response time.

Introduction

The information retrieval research has been facing new challenges raised by the emerging large-scale distributed networking systems, such as peer-to-peer networks, sensor networks, and mesh networks. Among all these networking systems, sensor network is the most sophisticated and general form due to its dynamic network topology, limited system resources, and infrastructure-free nature. Conceptually, a sensor network is a collection of cooperative mobile nodes that communicate with each other without the intervention of centralized access points. In a sensor network, the conventional information retrieval schemes (Beckmann et al., 1990, Gionis et al., 2001, Ren and Dunham, 2000), based on either centralized or flooding search mechanisms, cannot guarantee satisfactory performance due to the restrictions of bandwidth, power, and connectivity. The recently proposed improvements have achieved better performance in several experimental systems (Tang et al., 2003, Hara, 2001); however, the existing search schemes seem to ignore the data content distribution among the mobile nodes. The search process is either blind, irrelevant to the queries, or based on static features of data sources. As the consequence, the search process is either inefficient or involved with high system overhead.

Similarity search, as a fundamental topic in information retrieval, has attracted considerable research attention in the past decade (Beyer et al., 1994). The problem of similarity search can be generally described as finding the top-k most relevant data objects of a given query according to the predefined search criteria and semantic distance metrics. Performing similarity search in sensor networks is a challenging task due to the lack of infrastructure: consider a commonly used similarity search—content-based multimedia data retrieval. When a query is issued, due to the mobility and infrastructure-free nature, the contents of data sources are unknown at the requesting node, and therefore flooding has to be employed for the query resolution. However, flooding implies large consumption of system resources—storage, bandwidth, and energy. In addition, due to the share-medium nature of most sensor networks, flooding may also cause serious redundancy, contention, and collision of rebroadcast messages in the shared wireless channels, which is known as the Broadcast Storm (Ni et al., 1999). Moreover, considering the sheer sizes of multimedia data, the performance deterioration is more drastic.

Caching frequently accessed data objects is an efficient technique for improving system performance in mobile environments. Average data access latency can be reduced as similarity queries may be issued on the same data objects cached in the local storage, thereby avoiding the need of repeated blind search in the network. Moreover, analysis of the earlier queries and their results may help deriving the content distribution knowledge of the sensor network, which provides the heuristic navigation to the similarity search for later queries. Based on the content distribution knowledge, the similarity search can be performed as follows: for a query submitted to the sensor network, the mobile nodes in the network can be divided into two categories: the nodes containing data related to the query (relevant nodes) and the nodes that do not contain related data (irrelevant nodes). To improve the performance, one should avoid query processing by the irrelevant nodes and try to forward the query only to the relevant nodes. By doing this, the wireless network traffic is reduced, and the system performance is improved as well.

This paper is intended to provide a solution to the efficient similarity search in sensor networks using adaptive semantic-based caching. We defined the Hilbert curve based representation of data contents of mobile nodes, and proposed a novel caching scheme—Ad-hoc Semantic Caching (ASC)—to facilitate the process of similarity search in sensor networks. Simulation results show that the proposed scheme can significantly reduce the search cost in terms of query delay and message complexity.

The remaining part of this paper is divided into four sections: Section 2 reviews the related work in similarity search and caching schemes in sensor networks. Section 3 gives the concepts of Hilbert-curve-based representation, providing the theoretical foundation for caching. Section 4 explains the semantic-based caching scheme. Section 5 evaluates the proposed caching scheme using experimental analysis. Section 6 draws the paper into conclusions.