Context-based file systems and spatial query applications

Abstract

This dissertation presents studies related to I/O techniques in data-intensive computation and advanced solutions of spatial queries. There is a lack of general mechanisms for integrating multiple file system techniques and, therefore, the dissertation first illustrates a framework for Context-Based File Systems (CBFSs), which simplifies the development of context-based file systems and applications. Unlike existing informed-based context-aware systems, the framework provides a unifying informed-based mechanism that abstracts context-specific solutions as views, thereby allowing applications to make view selections according to their behaviours. The framework can not only eliminate overheads induced by traditional context analysis, but also simplify the interactions between file systems and applications. In addition, offloading a portion of a program to an active storage is another way to improve I/O performance in a cluster by significantly reducing data traffic. In the offloading study, we design a general offloading framework or ORCA that enables programmers without I/O offloading experience to complete the offloading development.

In the second part of this dissertation, we propose two novel spatial queries, multi-criteria optimal location query and keyword-spatial query. In our approaches, Voronoi diagram techniques are utilized for efficiently answering the queries. Besides two intuitive approaches, we explore two advanced solutions, Real Region as Boundary (RRB) and Minimum Bounding Rectangle as Boundary (MBRB), which are based on our proposed Overlapping Voronoi Diagram (OVD) model. High complexity of Voronoi diagram overlap computation in RRB motivates us to reduce costs of the overlap operation by replacing real boundaries of Voronoi diagrams with their Minimum Bounding Rectangles (MBR). Moreover, we employ a filter-and-refine strategy in an evaluation system for the keyword-spatial query. The system is comprised of Keyword Constraint Filter (KCF), Keyword and Spatial Refinement (KSR), and a ranker. KCF calculates the keyword relevancy of spatial objects, and KSR refines intermediate results by considering both spatial and keyword constraints. The extensive experimental results show that the queries can be efficiently and effectively evaluated by the proposed solutions.