In a mobile computing environment, the limitation of wireless bandwidth and battery power on clients makes system design different from the conventional stationary client-server paradigm. To overcome these resource constraints, techniques such as information broadcast to a large population, caching of frequently accessed data, and point-to-point communication for pull-based data requests can be used together to reduce data access time. Broadcasting index along with the data can reduce power consumption while maintaining an acceptable access time. Allocating channels between broadcast and on-demand modes can significantly reduce data access time with the limited bandwidth. Due to mobility and possibility of client disconnection, maintaining cached data consistency becomes a challenging...[ Read more ]

In a mobile computing environment, the limitation of wireless bandwidth and battery power on clients makes system design different from the conventional stationary client-server paradigm. To overcome these resource constraints, techniques such as information broadcast to a large population, caching of frequently accessed data, and point-to-point communication for pull-based data requests can be used together to reduce data access time. Broadcasting index along with the data can reduce power consumption while maintaining an acceptable access time. Allocating channels between broadcast and on-demand modes can significantly reduce data access time with the limited bandwidth. Due to mobility and possibility of client disconnection, maintaining cached data consistency becomes a challenging research issue. Periodically broadcasting invalidation reports to the clients is an efficient method to keep cache consistency. In previous studies, some of the research issues have not been addressed or were studied under certain constraints on the mobile systems. In this thesis, all the above techniques were investigated and several solutions were proposed.

In the first part of this thesis, adaptive invalidation report methods were proposed. For cache invalidation, the server broadcasts different invalidation reports according to client needs so that the data access/update patterns and client disconnection time have less influence on the performance of cache invalidation strategies. Second, index techniques based on index trees and signatures were investigated. Unlike previous studies, two important data organization factors, namely clustering and scheduling, were covered. The cost models for the index tree and signature methods under various data organizations were proposed. Queries on single attribute and multiple attributes were explored in the study. Moreover, a hybrid indexing method combining the strengths of the signature and the index tree techniques was proposed. Third, since broadcast channels and on-demand channels incur different access overhead for different system work-loads, the cost models for dynamic channel allocation between broadcast and on-demand modes were derived. Based on that, channel adaptation algorithms for optimizing overall access time were developed. Finally, the performance of a dynamic data delivery system was studied. A distinguishing feature of the proposed model is that data are disseminated through data broadcasting, data caching, and pull-based delivering according to data access patterns. A dynamic data access pattern collection mechanism was proposed. Additionally, cache management policies and indexing schemes for data broadcast were explored. The proposed methods were evaluated using simulation models and were found to yield consistently better performance when compared with existing methods.