在行動環境中應用於位置相關查詢之快取替換與資料廣播策略

Abstract

在行動計算環境中，位置相關查詢 (Location Dependent Query) 是一項重要的個人化服務。行動設備快取的效能和行動支援站對頻寬的利用，在行動通訊的環境中是兩項經常被討論的重要議題。當這些議題被套用在位置相關查詢的主題時，行動設備和行動支援站都需要更有彈性和高效能的策略，以處理這些接踵而來具有時間性，空間性及差異性的查詢請求。在這篇論文中，我們在行動設備中充分的運用使用者記錄檔來增進快取替換的效能，並根據位置相關查詢的特點來安排資料的廣播順序，以降低資料查詢的讀取時間。實驗結果顯示，我們所提出的快取替換策略，在最好的情況下，快取命中率比以最近最少讀取及以物件距離為基礎的兩種策略至多約高20%。同時，我們所提出的資料廣播策略，在資料平均讀取時間方面，也比以優先權為基礎及以格子區域為基礎的兩種廣播策略約低30% ~ 40%。

Location Dependent Query (LDQ) is an important personal service in mobile environments. Cache efficiency of mobile devices (MDs) and bandwidth utilization of a mobile support station (MSS) are two important design issues often discussed in mobile communications. To address these issues for LDQs, both MDs and MSS need more flexible and efficient strategies to deal with subsequent data requests which are temporal, spatial and different. In this thesis, we fully utilize user profiles of MDs to enhance efficiency of cache replacement and according to the characteristics of LDQs to arrange a data broadcasting order to reduce latencies of data requests. Evaluation shows that the proposed strategy for cache replacement outperforms the LRU and distance-based strategies by as much as 20% in average cache hit ratio. In addition, the proposed strategy for data broadcasting outperforms the priority-based or the grid-based strategies from 30% to 40% in reducing average latency.