- 學位論文
在普適計算環境中以品質服務為基礎之即時型服務系統
Time-Constraint Service System in Ubiquitous Computing Environments Based on Quality of Service
摘要
在行動網路的環境中，網路連接的效能將影響網路的服務品質，為提升系統的服務範圍以及處理多變化的資料樣本，即時無線廣播方法是一項可靠的資料傳輸機制，本研究針對即時無線廣播系統提出一個分析與實作模型，正如本研究實驗結果所示，傳統的即時運算演算法應用於無線廣播環境時，其即時性資料的處理，效能表現已無法滿足傳統的預測，因此，本研究提供一個以服務品質為基礎的排程演算法—在多重廣播頻道上動態調整演算法，來處理即時性的資料並服務行動使用者的需求。 為突顯本研究的貢獻，在效能評比方面，本研究引用傳統主從式即時運算演算法作為效能評量標準，評比項目包括網路存取延遲時間、使用者等待時間、系統擴展性與系統執行負載率，除此之外，更重要的是最佳化即時性資料的截止率，從一系列的實驗結果得知，本研究所提出的演算法與機制，在各項效能評估中均優於傳統的演算機制，由此證明，本研究的可行性與貢獻度。 在未來研究工作上，除了著手改進本研究的演算機制，以降低其執行時間複雜度，提高實用性與執行效率，在本研究的研究過程中，也觸發一些相關的研究方向，包括即時演算法的精進、即時傳輸交易的錯誤控制、即時性熱門資料的篩選與更新、可變動大小的即時資料傳輸、行動裝置快取資料管理與多重節點即時資料通訊。
並列摘要
Network connectivity affects the quality of service (QoS) in a mobile network. Real-time broadcasting is a promising data dissemination method to improve system scalability and deal with dynamic data access pattern. This study presents an analysis model and a simulation model for real-time broadcasting systems. As this study demonstrates, traditional strategies like EDF (Earliest Deadline First) and LSF (Least Slack First) in the non-mobile real-time environment do not perform efficiently in a mobile broadcasting environment. Therefore, this study proposes an efficient scheduling algorithm with guaranteed QoS, called dynamic adjustment scheduling (DAS), which is designed for timely delivery of data to mobile clients. This study also compares DAS with traditional client/server based real-time scheduling strategies and mobile non-real-time broadcast strategies. The proposed approach generally outperforms existing real-time strategies with different deadline distributions. A series of simulation experiments evaluates the performance of the proposed scheme. The results demonstrate that this algorithm outperforms other algorithms for performance metrics such as miss rate, waiting time, and stretch. Results also show that the overhead of this algorithm is low compared with other scheduling algorithms. In the future, we plan to improve the DAS strategy by reducing its scheduling time complexity. Other topics for future research include the investigation of real-time scheduling algorithms that can handle transmission errors, update access patterns, unfixed page sizes, client cache management schemes and multi-hop communication.