透過您的圖書館登入
IP:3.135.183.221
  • 學位論文

使用馬可夫決策過程來改善Android記憶體使用效率

Improving Android Memory Utilization using Markov Decision Processes

指導教授 : 楊正仁

摘要


在Android系統中,資源管理的問題對使用表現有重要影響。而在眾多的資源當中,記憶體管理是影響最直接的部分。根據我們的觀察,Android系統當中的記憶體管理機制是基於最近最少使用演算法(LRU) 去管理記憶體回收。此外,並使用垃圾回收機制(Garbage Collection) 去回收空閒頁的記憶體。然而,由於在有限的記憶體空間中有越來越多的應用程式會被執行,目前的記憶體管理機制會造成應用程式載入時間拉長。因此在本論文中,我們使用馬可夫決策過程(Markov Decision Processes) 提出了一個預測式的記憶體管理機制,來改善Android系統上記憶體使用效率。本論文研究主要達成以下兩項貢獻:第一,基於馬可夫決策過程的記憶體管理機制可以有效率的在Android系統上提升記憶體使用率。第二,能有效減少後續執行應用程式時所需要的載入時間。

並列摘要


In Android, resource management has significant impacts on system performance. Especially, memory management is the most crucial. According to our observations, current memory management in Android is based on the Least Recently Used (LRU) algorithm to claim for more free memory space. In addition, it uses a Garbage Collection (GC) mechanism to perform memory recycling. If numerous applications are executed in the limited memory space, currently Android may suffer from the lengthened loading problem. Therefore, in this research we propose a predictive memory management scheme using the Markov Decision Processes (MDP) model to improve the memory utilization. The proposed scheme has two following contributions: (1) the MDP-based memory management can efficiently improve the memory utilization in Android, and (2) it can effectively reduce the loading time while many applications are executed.

參考文獻


[2] Brian T. Bennett and Vicent J. Kruskal, “LRU Stack Processing,” IBM Journal of
age Time for the Pure Literal Rule,” SIAM Journal on Computing, vol. 18, no. 2, pp.
409–418, Apr. 1989.
Flash-Memory Storage Systems of Real-Time Embedded Systems,” ACM Transac-
“Markov Decision Process (MDP) Framework for Optimizing Software on Mobile

延伸閱讀