Managing Web Data Repositories under Quality Contracts
Jie Xu (Pitt/CS)
PhD Proposal Defense
Wednesday, December 19th, 2007
12:00pm - SENSQ 6106 - Eli Lilly
Abstract
Since the proliferation of the Web in the mid-1990’s, web portals have become pervasive due to their great usability and the huge amount of information they provide to end-users. While using web portals, users expect quick answers (QoS requirement) and fresh data returned by the query (QoD requirement). Given the inherent resource constraints and the size of theWeb today, it is hard to optimize on both quality requirements (i.e., query response time and data freshness) at the same time.
In this thesis, we argue that it is beneficial to allow users to specify their preferences, and the system then can optimize with the objective of satisfying users’ preferences. We propose to use Quality Contracts (QC) framework to empower users to express their preferences over multiple quality specifications and multiple queries. Under the QC framework, we try to resolve three critical problems. (1) How do we selectively refresh a web portal’s local data repository, with the objective of maximizing the data repository’s freshness? (2) How dowe determine which data objects should be selected and maintained (i.e., copied and kept fresh) in our local data repository, with the objective of maximizing users’ satisfaction? (3) In the presence of multiple replicas for each data object, which replicas should we choose to answer queries, and which replicas should we use to refresh the web portal’s data repository? We propose to develop a suite of algorithms to address the three problems. We will evaluate the proposed framework and algorithms by an experimental evaluation with real web data traces and synthetic data.
Dissertation Adviser
Prof. Alexandros Labrinidis, Department of Computer Science
Committee Members
Prof. Panos K. Chrysanthis, Department of Computer Science
Prof. Ahmed Amer, Department of Computer Science
Dr. Vladimir I. Zadorozhny, School of Information Sciences





