Please use this identifier to cite or link to this item: http://hdl.handle.net/10397/83513
Title: Data warehousing support for mobile environment
Authors: Lee, Chi-keung Ken
Degree: M.Phil.
Issue Date: 1999
Abstract: Over the last decade, the rapid advances in wireless communication technology and enhancement of portable computers have led to an evolution of mobile database applications. Typically, a mobile environment is built upon a cellular network which is a combination of a back-bone wired network and a set of small wireless cells. In the wired network, the database servers and the base stations are fixed in their locations. In the wireless cell, the mobile clients dynamically make connections to the base stations. To access information, a client queries one or more database servers via any connected base station. The relationship among these three entities could be roughly viewed as a three-tier client/server architecture. The base station works as a middle-tier to serve the other two entities. In this thesis, we generalize such an architecture into a framework of Mobile Warehousing System (MoWS) and investigate issues for improving the performance of client query processing in this new environment. Very often, the population of mobile clients is huge compared with that of database servers in the mobile environment. A server being accessed by numerous clients will be overloaded. The query processing performance could not be guaranteed. In order to enhance the performance, a common approach is to replicate data in distributed hosts, from where clients can access information. In our research, we propose to maintain useful information pertaining to a subset of databases, which is of common interest to a handful of clients, in a form of materialized database view in the base stations. We intend to equip with the base stations ability to help answering client queries instead of merely directing the requests to the server; thus the server loading can be relieved to a large degree. We term each base station, which serves as a data repository for clients, mobile data warehouse. One of the most important issues regarding data replication is data consistency maintenance. The replicated data becomes stale when the source is updated. In the scope of MoWS, we study a pull-based view update scheme for a mobile data warehouse to request differential changes from the source database servers. In order to demonstrate the capability of addressing the view update anomaly problem due to server autonomy and asynchronous database updates, the correctness and complexity of our pull-based update scheme are studied. In addition, we address the client query processing limitation over a narrow bandwidth and unreliable wireless channel. Accessing remote database server over a wireless channel will make it suffer from a lot of communication overhead. The narrow bandwidth amplifies data transfer latency while the unrelability causes a client occasionally disconnected. To cater for this problem, client caching is recommended. Most conventional caching schemes, such as page-based or record-based, are unable to assist clients to determine if there is sufficient cached data to answer their queries, thus forcing them to contact the server for possibly missing data. In this thesis, we suggest the use of a semantic caching scheme in which every query result associated with a semantic description is cached in a mobile client as a data block. By reasoning with the specification of an initiated query and the semantic description of the cached data block, a client becomes intelligent to assert whether the cache can contribute to answering the query completely and deduce what is missing from the cache. The main drawback of our scheme is an introduction of dynamic cache granularity that complicates the cache manipulation. We propose several cache management techniques for our semantic caching scheme. Shortly concluded, MoWS is a hierarchical data replication framework in which information in database servers is replicated in the base stations and the mobile clients as materialized view and cache respectively. The strength of this design is that it enables mobile clients to answer their queries with little dependency on the wireless channel. In addition, mobile data warehouses are able to serve a mass of mobile clients, thus sharing the server loading. To quantify the performance of MoWS, we implement a prototype. We conduct a series of experiments based on the prototype, along side with the appropriate quantitative analysis. From the results, we identified scenarios where our approach is beneficial, resulting in shorter response time, better cache hit, smaller transmission cost and lower storage overhead. These results demonstrate the effectiveness of our proposed schemes and the suitability of MoWS.
Subjects: Mobile computing
Data warehousing
Hong Kong Polytechnic University -- Dissertations
Pages: vii, 113 leaves : ill. ; 30 cm
Appears in Collections:Thesis

Show full item record

Page views

88
Last Week
0
Last month
Citations as of Mar 24, 2024

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.