Shintaro Hamanaka

Improving choice of processes to terminate in Android OS

Android operating system, which is the most popular platform for smartphones and tablet computers, has an original memory managing system, which is call “low memory killer”. In case of lacking of memory, Android operating system terminates processes until enough memory is available. It selects targets in order of pre-defined priority and consuming memory size regardless of its re-launching time, its frequency of use, and whether it is recently used. Thus, this termination sometimes harms user convenience. In this paper, we propose two new policies on selection of processes to terminate. One is based on LRU and the other is based on application re-launching time. Then, we show our evaluation results and demonstrate that the proposed methods can improve process launching time.

Object lifetime trend of modern Android applications for GC performance improvement

Android operating system has become one of the most popular smartphone platforms. A large number of applications are developed for Android Runtime (ART). Garbage Collection (GC) is an essential function of ART for Java-based applications. GC suspends all the application threads, called Stop The World (STW), and sacrifices applications performance. For reducing pauses to check heap, generational GC was proposed. It separates objects into two groups, the objects which probably die in the next GC and the others. Then, it tries to reduce search space and time to check. With this GC, accurate estimation of object lifetime is important. In this paper, we explore trend of object lifetimes of modern Android application in order to improve estimation of lifetime of ART GC. For this purpose, we have constructed our application behavior monitoring system, which is called ART monitor. This is implemented by modifying Android and enable observation of creations and destruction of objects in ART. We have investigated the relation among object lifetime, object sizes, and object names in recent Android applications which are popular in Google Play Store. We than found that small object probably dies with short lifetime and objects of some classes always survive long time. Based on these findings, we discuss a method for improving ART GC performance.

Identifying battery-draining applications by monitoring behavior in screen-off state in Android

Power consumption is one of the most important issues of smartphones. Android OS, a popular smartphone operating system, has a function with which an application can be invoked in screen-off state without users operation. Some applications frequently work in screen-off state, and consume battery. In this paper, we propose a method for identifying applications that largely drain battery in Screen-off state in Android devices. We introduce the standard method of Android for estimating power consumption of each application, and show that it cannot always estimate consumption correctly. Then, we propose a method for identifying heavily battery-draining application by monitoring setting and invoking alarm, which is a common method for executing an application in screen-off state. Our experimental results demonstrate that our method can identify battery-draining applications more correctly than the standard method of Android operating system.

Accelerated application monitoring environment of Android

Huge amount of applications are developed and running in smartphones. Monitoring and analyzing application behaviors are important. Dynamic analyses with actual application execution require long time. Thus, reducing monitoring time is an important issue. In our work, we focus on Android operating system, which is a popular operating system for smartphones, and propose a method for constructing an environment that enables an accelerated application monitoring with which application can be monitored in shorter time than the actual time. Android operating system is based on Linux kernel and processes in a system are provided time from the kernel. It is expected that speed of time flows that applications in a system recognize can be accelerated by modifying the kernel. In this paper, we assume some conditions for simplifying as the first phases of this work, and propose a method for providing accelerated time for application. Then, we evaluated our method by monitoring practical applications, and demonstrate that our method can suitably accelerate speed of recognized time flow.

Application power consumption estimation considering software dependency in Android

Android operating system has become one of the most popular smartphone platforms. One report stated that the most important issue of smartphones was its power consumption. Android has a function with which an application can be invoked in screen-off state without users operation. Some applications frequently work in screen-off state, heavily and consume battery. For saving power consumption in the state, accurate estimation of power consumption of each application is important. However, estimating power consumption cannot be easily achieved because of its dependency on device. That is, applications power consumption varies on installed application and type of hardware module in the device, which can be called software and hardware dependency, respectively. In this paper, we discuss estimation of power consumption in screen-off state considering software dependency. First, we explain software dependency of power consumption. Second, we propose a method, which takes account of software dependency, for estimating power consumption due to GPS. The proposed method monitors GPS utilization individually. Third, we evaluate our method with a benchmark and practical applications using GPS. We then demonstrate that our method can estimate power consumption of each application and suitably predict consumption after uninstalling an application without uninstallation.

An accelerated application monitoring environment with accelerated servers

Android operating system has become one of the most popular smartphone platforms. A large number of applications are developed for the operating system. Monitoring application behaviors with practical execution requires severely long time. Thus, reducing monitoring time is an important issue. In this paper, we propose a method for decreasing this time by accelerating the speed of time flow in Linux kernels in smartphones and server computers. The acceleration is achieved by modifying the management implementation in the kernel. We evaluate our method using application systems which are composed of Android smartphone and Linux server. Then, we demonstrate that the proposed method can shorten monitoring twice suitably.

Accelerated test for applications with client application and server software

Application analyses is important for smartphone markets and application developers. Dynamic analysis involving actual execution of application requires severely long time. Reducing monitoring time is an important issue. In this paper, we focus on Android applications and propose a method for reducing dynamic application monitoring time by constructing an accelerated environment wherein time flows faster than real time. The accelerated environment supports both of client-side applications and server-side services, and is constructed by modifying time managing implementation of Android kernel. We then present evaluation of our proposed system by practically applying the method to benchmark and practical applications. The experimental results demonstrate that our method can achieves accelerated monitoring with shorter time than real time monitoring.

Application state aware GC selection optimization in Android

Android operating system has a function, called LowMemoryKiller, which forcibly terminates application processes when size of available memory is less than the threshold. On reusing the same application again, re-creation of a process is required and takes longer time. ART (Android Runtime environment) has several GC (Garbage Collection) implementations, and choice of GC has effect on size of processes and behavior of LowMemoryKiller. In this paper, we investigate performance of GC implementations and propose a method for choosing GC implementation depending on application size and state. Then, we show our experimental results and demonstrate that our method reduces the number of process terminations cause by LowMemoryKiller.