| The proliferation of the Internet and World Wide Web, the development of wireless networks, the availability of asymmetric high-bandwidth links to the home, and the recent emergence of small-scale computing devices have fueled the development of applications with wide and varied data needs. The data-intensive environments in which these applications exist are characterized by very high data rates, the scale and variety of autonomous data sources, and the diverse needs of users and applications. Often, the resources (CPU, memory, bandwidth, etc.) in these environments are constrained and data management plays an important role in providing fast responsive systems through the effective scheduling of resources.; Databases have long benefited from the use of application-level knowledge to make decisions. Often, this application-level awareness comes from a priori knowledge of user interests. In these new networked environments there is evidence that more responsive systems can be delivered if users provide either some general indication of their interest or even better, a detailed query. We call this specification of user interest a "profile".; This dissertation focuses on two applications domains. In the first application domain, the process of synchronizing a database to improve the freshness of copies is studied. In the second application domain, we study the timely processing of large volumes of continuous high-rate data streams in the context of the Aurora stream processing engine. In both cases, profiles show promise in improving resource scheduling to provide effective data management. |
