摘要
针对低内存进程管理LMK(low memory killer)算法中内存阈值的设置主要依靠经验值,无法充分利用系统内存的问题,提出了一种基于统计分析和预测的低内存进程管理算法(low memory killer algorithm based on statistical analysis and predicting,LMK-SAP)。通过统计进程的内存使用,利用条件概率预测下一时间段内存增量,获得各类型进程的内存需求总量预测值,以此动态调整不同等级的内存阈值,最终根据内存阈值选择需要关闭的进程。实验结果表明,与LMK算法相比,LMK-SAP算法在内存中驻留的应用数目提高了56%,能够有效加快应用间的切换速度。
The memory threshold is set mainly relying on the empirical value in LMK （low memory killer）, which cannot make full use of system memory. To solve such problem, a low memory killer algorithm based on statistical analysis and predicting （LMK-SAP） is proposed. Firstly this algorithm records the use of processes memory and employs conditional probability to pre dict the memory increment in the next time window. Then the total predictive value of system memory for various types of processes that used to dynamically adjust the memory threshold is achieved. Finally the process to be closed is selected. The ex periment results show that this algorithm resides 56 ~ more applications in memory than LMK and effectively reduces the switch time between applications.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第1期107-111,118,共6页
Computer Engineering and Design
基金
国家科技支撑项目课题基金项目(2011BAH08B01)
国家863高技术研究发展计划基金项目(2011AA01A102)
中国科学院战略性先导科技专项课题基金项目(XDA06030500)
关键词
嵌入式操作系统
进程管理
低内存
概率
统计分析
embedded operating system
process management
low memory
probability
statistical analysis
