Abstract
With the advances of technologies, wireless connectivity becomes a critical component of networking infrastructure due to the favorable mobility and incessant reduction of hardware cost. Many creative wireless applications are feasible. On one hand, the scalable services allow numerous mobile users to receive the audiovisual streams on demand with short delays. On the other hand, the system should recognize the user positions to fulfill the pertinent location-aware contents as responses from the continuous queries without repetitive explicit specification from users. Both desirable features discern the very characteristics of wireless networks, and honestly give free rein to their applications. In this thesis, we propose two techniques. First is an efficient periodic broadcast scheme that preserves bandwidth efficiency when extensive segmentation on stream contents is necessary to initial service delay reduction. Our new design elegantly controls the effect of the packet overhead on the bandwidth efficiency by only fragmenting the very first segment. The latter segments are kept constant as the size of the full-sized packets. As a result, the proposed scheme can avoid the performance deterioration due to the induced overhead of segmentation, while accommodate the available server capacity to reduce the service latency. The performance study indicates our new scheme is constantly better than the existing competitors with all possible design options for the delivery of low bit-rate media streams. The second one is the location estimation problem. Our solution exploits the signal strengths of packets on wireless Ethernet networks to predict the location from the previously collected knowledge. To reduce the noise effects on signals, we use the quartiles to portrait strength distributions seen from the positions. We can deliver more reliable results than the techniques based on the average signal strengths, which are greatly biased by the strongest signals. Particularly, using quartiles allows us to answer the location query more accurately than the schemes that employ numerous strength data in each query. As a result, our solution scales gracefully in terms of computation cost, a favorable feature particularly suitable to the hand-held devices with limiting computing and storage resources.