Database Support for M-Commerce and L-Commerce

Introduction

M-commerce (mobile commerce) applications have evolved out of e-commerce (electronic commerce) applications, riding on recent advancement in wireless communication technologies. Exploiting the most unique aspect inherent in m-commerce, namely, the mobility of customers, l-commerce (location-dependent m-commerce) applications have played an increasingly important role in the class of m-commerce applications. All e-commerce, m-commerce, and l-commerce applications rely on the provision of information retrieval and processing capability. L-commerce applications further dictate the maintenance of customer and service location information. Various database systems are deployed as the information source and repository for these applications, backed by efficient indexing mechanisms, both on regular data and location-specific data.

Bean (2003) gave a good report on supporting Web-based e-commerce with XML, which could be easily extended to m-commerce. An m-commerce framework, based on JINI/XML and a workflow engine, was defined by Shih and Shim (2002). Customers can receive m-commerce services through the use of mobile devices such as pocket PCs, PDAs, or even smart phones. These mobile devices together with their users are often modeled as mobile clients. There are three types of entities central to m-commerce and l-commerce applications: mobile device, wireless communication, and database. In this article, we focus our discussion on mobile-client enabled database servers, often referred to as mobile databases. Mobile databases maintain information for the underlying m-commerce and l-commerce applications in which mobile devices serve as the hardware platform interfacing with customers, connected through wireless communication.

Location is a special kind of composite data ranging from a single point, a line, a poly-line, to a shape defining an area or a building. In general, locations are modeled as spatial objects. The location of a static point of interest, such as a shop, is maintained in a database supporting spatial features and operations, often a spatial database (Güting, 1994). The location of a moving object, like a mobile customer, needs to be maintained in a moving object database (Wolfson, Sistla, Xu, Zhou, & Chamberlain, 1999), a database that supports efficient retrieval and update of object locations. To enable l-commerce, both spatial databases and moving object databases need to support location-specific query processing from mobile clients and location updates they generated.

The two major types of data access requirements for a mobile database are data dissemination and dedicated data access. Data dissemination is preferred, since it can serve a large client population in utilizing the high bandwidth downlink channel to broadcast information of common interest, such as stock quotations, traffic conditions, or special events. On the other hand, dedicated data access is conveyed through uplink channels with limited bandwidth. To disseminate database items effectively, the selected set of hot database items can be scheduled as a broadcast disk (Acharya, Alonso, Franklin, & Zdonik, 1995). Proper indexes can be built to facilitate access to broadcast database items (Imielinski & Badrinath, 1994). Redundancy can be included in data (Leong & Si, 1995) and index (Tan & Ooi, 1998) to combat the unreliability of wireless communication.

For dedicated data access, queries and updates to databases are transmitted from the client to the server. L-commerce services involve processing of location-dependent queries (Madria, Bhargava, Pitoura, & Kumar, 2000). The high frequency of updates to the location of moving objects calls for special indexing technique. The call-to-mobility ratio serves as a good indicator on the tradeoff of indexing mechanisms. The moving object databases should enable efficient execution of queries such as k-nearest neighbor, reversed nearest neighbor (Benetis, Jensen, Karĉiauskas, & Ŝaltenis, 2006), and nearest surrounder search (Lee, Lee, & Leong, 2006). In addition, they should support continuous queries (Prabhakar, Xia, Kalashnikov, Aref, & Hambrusch, 2002), such as continuous k-nearest neighbor, being executed continuously and returning location-dependent results (Lee, Leong, Zhou, & Si, 2005). Reversing the role of query and data, it is equally important to process data streams effectively (Babu & Widom, 2001) such as incoming sensor data streams (Mokbel, Xiong, Hammad, & Aref, 2005) for traffic monitoring in navigational applications.