Research Of Memory Compression Based On Mobile System

With the development of mobile devices,the capacity of operating systems and applications is increasing.By contrast,the limited memory resources of mobile system gradually become the key of system development.Memory compression technology is widely used in smartphones to meet the increasing demand for memory capacity.In modern systems,memory is compressed page-by-page to avoid read amplification,but it compromises efficiency.Under high memory pressure,the conflict between fast page demand and inefficient memory compression results in a bad user experience.Based on this,this paper discovers the potential for compressing memory pages in batches and proposes Ant,an adaptive memory compression framework.The granularity potential of memory compression is fully exploited through the compression in batch.To the best of our knowledge,this is the first effort to realize large granularity memory compression in android smartphones.This dynamic granularity of memory compression fits well with application launch in mobile operating systems.By analyzing the launch phase of an application,this paper found that a large number of compression memory are accessed during the application hot launch phase and these memory pages have a good predictability.Using the feature of this application launch phase,you can avoid read amplification when compressing memory pages in batches.The contributions of this paper are listed as follows:·This paper analyzes the access delay model of mobile system and proposes that synchronous memory reclamation is the key factor affecting user experience under memory pressure.From the aspects of user experience,application behavior analysis,memory compression system and so on,it elaborates how the low efficiency of synchronous memory compression has a great impact on user experience.At the same time,this paper analyzes the efficiency of memory compression in commercial Android smartphones and observes that existing memory compression has considerable potential in granularity.·This paper shows that the memory of the Android system can also be compressed in large granularity,which has not been well concerned by either academia or smartphone manufacturers.So,this paper proposes Ant,an adaptive memory compression framework in Android smartphones.In Ant design,efficient memory compression under mobile operating systems can be achieved without modifying either application code or hardware infrastructure.·This paper evaluates dynamic granularity memory compression system from the perspective of data movement,and proposes a systematic data movement optimization scheme.Data movement issues are optimized from the compression/decompression process and the management of the compressed memory area.This effectively reduces memory and CPU overhead on Ant.In this paper,Ant is implemented on a real mobile operating system platform.The architecture implementation and data management technologies are introduced in detail in this paper.Finally,through experiments on real Android phones,the results show that the launch latency of applications is reduced by 26.1%in comparison with the state-ofthe-arts.At the same time,the user experience has been comprehensively improved.