On the Content Predictability of Cooperative Image Caching in Ad Hoc Networks

Emerging applications of mobile data management, such as content-based image retrieval in ad hoc networks, require the awareness of content distribution to enact and optimize the communication and search tasks. Caching is a widely used approach in mobile environments to improve system performance and keep track of replications. Within the scope of mobile and ubiquitous computing infrastructure, traditional caching techniques are not effective for large-size data such as images due to the limitations of bandwidth, storage, and power. The functionality of caching techniques relies on exact match, making them unsuitable for imprecise and similarity-based queries. In addition, the description of cached contents is defined based on the query context instead of data content, which fails to exploit the semantic locality of cached data and makes the traditional caching techniques inefficient in utilizing cache storage. In this paper, we present a semantic-aware caching scheme (SAIC) for image databases in ad hoc networks. The proposed scheme is designed based on several novel ideas: 1) multi-level partitioning of semantic space, 2) constraint-based representation of image semantics, 3) non-flooding query resolution, and 4) adaptive cache consistency maintenance. Our combination of theoretical analysis and simulation show the efficiency of the proposed scheme, and evaluate its performance in comparison against two ad hoc caching schemes as advanced in the literature.