Embedded Java Virtual Machine Research Articles

Real-time Embedded Java Virtual Machine for Application Development in Wireless Sensor Network

The application development in wireless sensor network (WSN) is complicated due to the diverse software and hardware platforms. One way to solve this problem is to apply the embedded Java virtual machine (EJVM) on the WSN nodes. With the EJVM, the users can program the applications by the popular and high-level abstract Java language. However, several challenges exist for applying the EJVM on the WSN nodes, such as the memory resources on the WSN nodes are constrained, the multitasking programming and real-time response are required by many WSN applications. Currently, these challenges cannot be addressed well by most EJVMs. And in order to address these challenges, a real-time memory-efficient Java operating system HEROS is developed. By means of HEROS, the multitasking real-time Java applications can be developed even on the memory-constrained WSN nodes. Nevertheless, the execution efficiency of the Java bytecode is not high, making the EJVM not suitable to be used on the tight energy-constrained WSN nodes. To solve this problem, a new mid-layer software REMID, which is designed to have some similar functionalities as the EJVM, is also developed. Different from the EJVM, REMID is designed to be both memory efficient and energy efficient, thus it can substitute the EJVM to be used even on the severe energy-constrained WSN nodes. The final evaluation works prove that a user-friendly WSN application development environment can be provided to the users by means of the EJVM, HEROS and REMID.

안드로이드 달빅과 자바 가상머신의 성능비교

본 논문에서는 안드로이드에 탑재되는 가상머신인 달빅(Dalvik) 가상머신의 성능을 벤치마크를 구현하여 측정하고 기존 임베디드 자바 가상머신과의 성능을 비교한다. 달빅 가상머신의 성능 측정을 위하여 가상머신 성능 측정에 널리 사용되는 SPECJVM 벤치마크를 사용하였다. 달빅과의 성능 비교를 위하여 임베디드 자바 가상머신인 SUN Java 임베디드 버전을 비교 대상으로 선정하여 동일한 벤치마크를 사용하여 실행시간을 비교하였다. 성능 측정은 스마트폰 개발 하드웨어인 오드로이드(Odroid)에 가상머신과 벤치마크 프로그램들을 포팅하여 이루어 졌다. 또한, 달빅 가상머신의 최근 버전에서 구현되지 않은 적시(Just-In-Time) 컴파일 기능을 달빅에 추가하여 성능 개선 정도를 분석하였다. 분석 결과 안드로이드 달빅 가상머신은 SUN 가상머신 성능의 15% 수준을 보였으며, 적시 컴파일 기법을 적용한 달빅은 63% 수준에 도달함을 보였다. In this paper we analyzed performance of Andriod's Davik virtual machine(VM) using standard benchmark and compared the result with the embedded Java virtual machine. We used a well known benchmark suit named SPECJVM for the measurement. For the fair comparison, Sun Java embedded JVM is ported and the same benchmark is ported on it. The Odriod smartphone hardware platform is used as the target hardware. We have added a Just-In-Time compiler to Dalvik, which is not supported in the recent Android release, and measured performance improvement. The experiment result show that Dalvik achieved 15% and Dalvik with JIT shows 63% of the Sun's JVM performance.

Architecture Independent Characterization of Embedded Java Workloads

This paper presents architecture independent characterization of embedded Java workloads based on the industry standard GrinderBench benchmark which includes different classes of real world embedded Java applications. This work is based on a custom built embedded Java virtual machine (JVM) simulator specifically designed for embedded JVM modeling and embodies domain specific details such as thread scheduling, algorithms used for native CLDC APIs and runtime data structures optimized for use in embedded systems. The results presented include dynamic execution characteristics, dynamic bytecode instruction mix, application and API workload distribution, object allocation statistics, instruction-set coverage, memory usage statistics and method code and stack frame characteristics.

CVM 환경에서 임베디드 가비지 컬렉터의 성능 평가

가비지 컬렉션은 자바 가상 머신의 핵심적인 기능으로서 개발자들이 겪는 메모리 관리의 어려움을 줄여준다. 본 논문에서는 임베디드 자바 가상 머신을 위한 가비지 컬렉터인 GenGC와 GenRGC의 성능을 평가하고 분석한다. 성능 평가를 위하여 썬 마이크로시스템즈사에서 개발한 실제 임베디드 자바 가상 머신인 CVM을 플랫폼으로 사용하며, SpecJVM98을 벤치마크 프로그램 집합으로 사용한다. 첫째, GenGC와 GenRGC의 성능을 비교하기 위하여 힙의 크기 및 각 영역의 크기를 변화시키면서 가비지 컬렉션 수행 시간 및 지연 시간을 비교한다. 둘째, GenRGC의 성능을 보다 세밀하게 분석하기 위하여 힙의 구성 요소 중 블록과 프레임의 크기를 변화시키면서 가비지 컬렉션 수행 시간 및 지연 시간을 측정하여 분석한다. 셋째, GenRGC를 사용하기 위하여 필요한 저장 공간의 크기를 분석하고, GenRGC가 제한된 메모리를 가지는 임베디드 환경에서 적합하다는 것을 보인다. CVM은 가장 대표적인 임베디드 자바 가상 머신이므로, 이와 같은 성능 연구는 실제 응용 환경에서 가비지 컬렉터의 성능을 보다 정확히 예측할 수 있다는 점에서 큰 의미를 갖는다. Garbage collection in the Java virtual machine is a core function that relieves application programmers of difficulties related to memory management. In this paper, we evaluate and analyze the performance of GenGC and GenRGC, garbage collectors for embedded Java virtual machines. For performance evaluation, we employ CVM, a real embedded Java virtual machine developed by Sun Microsystems, Inc., as a platform and also use a widely-used SpecJVM98 as a set of benchmark programs. To compare the performance of GenGC and GenRGC, we first evaluate the time of garbage collection and the delay time caused by garbage collection. Second, for more detailed performance analysis of GenRGC, we evaluate the time of garbage collection and the delay time caused by garbage collection while changing the sizes of a block and a frame. Third, we analyze the size of storage space required for performing GenRGC, and show GenRGC to be suitable for embedded environment with a limited mont of memory. Since CVM is the most representative one of embedded Java virtual machines, this performance study is quite meaningful in that we can predict the performance of garbage collectors in real application environments more accurately.

Java client ahead-of-time compiler for embedded systems

The performance of embedded Java virtual machine can be improved by ahead-of-time compiler (AOTC), which translates bytecode into machine code in the server and installs the machine code on the client device. Although AOTC is more advantageous than just-in-time compiler (JITC) due to its absence of the translation overhead, AOTC cannot be applicable to dynamically downloaded classes at runtime. This paper proposes client-AOTC (c-AOTC) which performs AOTC on the client device using the JITC module installed on the device, complementing the server-AOTC. The machine code of a method translated by JITC is cached on a persistent memory of the device, and when the method is invoked again in a later run of the program, the machine code is loaded and executed directly without the translation and interpretation overhead. One of major issues in c-AOTC is relocation because some of the addresses used by the cached machine code are not correct when the machine code is loaded and used in a different run; those addresses should be corrected before they are used. Constant pool resolution and inlining complicate the relocation problem and we propose our solutions. We developed a c-AOTC on Sun's CDC VM reference implementation (RI) and our evaluation results indicate that c-AOTC can improve the performance significantly, as much as an average of 12%. We also experiment with the issue of reducing the number of c-AOTC methods to be cached when the persistent space is tight, with a graceful degradation of performance.

Energy consumption analysis for two embedded Java virtual machines

In this paper we present a general framework for estimating the energy consumption of an embedded Java virtual machine (JVM). We have designed a number of experiments to find the constant overhead and establish an energy consumption cost for individual Java opcodes for two JVMs. The results show that there is a basic constant overhead for every Java program, and that a subset of Java opcodes have an almost constant energy cost. We also show that memory access is a crucial energy consumption component.

EVM에서의 자바 동적 메모리 관리기 및 쓰레기 수집기의 구현 및 성능 분석

자바 언어는 객체지향성, 안전성, 유연성으로 인하여 현재 가장 널리 쓰이는 프로그래밍 언어의 하나가 되었으며, 자바 가상머신이 제공해주는 메모리 관리기 및 가비지 컬렉터로 인하여 프로그래머는 메모리 관리에 관한 많은 고민이 줄어들었다. 본 논문에서는 임베디드용 자바 가상 머신에서 구현된 메모리 관리기 및 가비지 컬렉터에 대해서 제안한다. 메모리 관리기는 힘을 다양한 크기의 셀로 분할한 후 통일한 셀의 집단인 블록 단위로 관리하여 빠른 메모리 할당과 해제가 가능하다. 가비지 컬렉션 방법으로는 3-색상 기반 표기-쓸어담기 가비지 컬렉터를 기반 알고리즘으로 채택하였으며 멀티쓰레드를 지원하기 위해 새로운 4-색상 기반 표기-쓸어 담기 가비지 컬렉터를 제안한다. 제안하는 기법은 메모리 단편화가 발생하지만 객체 생성수가 많아짐에 따라 단편화율이 거의 일정함을 실험을 통해 보인다. Java has been established as one of the most widely-used languages owing to its support of object-oriented concepts, safety, and flexibility. Garbage collection in the Java virtual machine is a core component that relieves application programmers of difficulties related to memory management. In this paper, we propose a memory manager and a garbage collector that is implemented on a embedded java virtual machine. The memory manager divide a heap into various-sized cells and manages it as blocks of same-sized cells. So it is possible to allocate and free memory fast. We adopted the 3-color based Mark & Sweep garbage collector as our base algorithm and we propose 4-color based Mark & Sweep garbage collector for supporting multi-threaded program. The proposed garbage collector occurs memory fragmentation but we show through a experiment that the fragmentation ratio is almost fixed according as we create objects continuously.

Accelerating embedded Java for mobile devices

With the proliferation of wireless devices, networks, and systems, the deployment of efficient embedded Java virtual machines is becoming a challenging and important research area. Accordingly, a plethora of acceleration techniques have been proposed. In this article we present a new acceleration technology that we developed for embedded Java virtual machines. Acceleration is achieved by the integration of a new selective dynamic compiler, which we called Armed E-Bunny, into the J2ME/CLDC (Java 2 Micro-Edition for Connected Limited Device Configuration) kilobyte virtual machine (KVM). The modified KVM is ported on a handheld PDA that is powered with embedded Linux. Experimental results demonstrate that we accomplished an important speedup (more than 360 percent) with respect to Sun's latest version of KVM. This experimentation was carried out using standard J2ME benchmarks.

에너지 관점에서 임베디드 자바가상기계의 메모리 접근 형태

임베디드 시스템에서 일반적으로 메모리가 가장 많은 에너지를 소비하는 것으로 알려지고 있다. 임베디드 자바가상기계의 경우도 마찬가지이며, 따라서 보다 에너지 효율이 높은 자바가상기계의 개발을 위해서는 자바 메모리의 에너지 사용을 최적화 하는 것이 무엇보다 중요하다. 본 논문에서는 자바 프로그램 실행 시 수많은 바이트코드들이 어떻게 논리적 메모리를 접근하는지 분석하였다. 이런 접근 형태 분석은 자바 메모리의 설계 및 구현 기술을 선택하는데 큰 통찰력을 제공해 준다 힙, 오퍼랜드 스택, 지역변수배열 등 세 가지 논리적 데이터 공간에 대해 각각 메모리 접근을 분석하였으며, 분석 결과 오퍼랜드 스택이 가장 빈번하게, 또한 균일하게 사용되었으며 힙이 가장 드물게, 그리고 불균일하게 사용되었음을 알 수 있었다. 힙과 지역변수배열은 읽기 위주로 사용되었으며, 오퍼랜드 스택은 읽기와 쓰기 비율이 크게 다르지 않았다. Several researchers have pointed out that the energy consumption in memory takes a dominant fraction on the energy budget of a whole embedded system. This applies to the embedded Java virtual machine tn, and to develop a more energy-efficient JVM it is absolutely necessary to optimize the energy usage in Jana memory. In this paper we have analyzed the logical memory access pattern in JVM as it executes numerous number of bytecode instructions while running a Java program. The access pattern gives us an insight how to design and select a suitable memory technology for Java memory. We present the memory access pattern for the three logical data spaces of JVM: heap, operand stack, and local variable array. The result saws that operand stack is accessed most frequently and uniformly, whereas heap used least frequently and non-uniformly among the three. Both heap and local variable array are accessed mostly in read-only fashion, but no remarkable difference is found between read and write operations for operand stack usage.

E-Bunny: A Dynamic Compiler for Embedded Java Virtual Machines.

A new acceleration technology for Java embedded virtual machines is presented in this paper. Based on the selective dynamic compilation technique, this technology addresses the J2ME/CLDC (Java 2 Micro Edition for Connected Limited Device Configuration) platform. The primary objective of our work is to come up with an ecient, lightweight and low-footprint accelerated embedded Java Virtual Machine. This is achieved by the means of integrating a selective dynamic compiler that we called E-Bunny into the J2ME/CLDC virtual machine KVM. This paper presents the motivations, the architecture, the design and the implementation issues of E-Bunny and how we addressed them. Experimental results on the performance of our modified KVM demonstrate that we accomplished a speedup of 400% with respect to the Sun’s latest version of KVM. This experimentation was carried on using standard J2ME benchmarks.

USING MEMORY COMPRESSION FOR ENERGY REDUCTION IN AN EMBEDDED JAVA SYSTEM

The Java programming language is being increasingly used for application development for mobile and embedded devices. Limited energy and memory resources are important constraints for such systems. Compression is an useful and widely employed mechanism to reduce the memory requirements of the system. As the leakage energy of a memory system increases with its size and because of the increasing contribution of leakage to overall system energy, compression also has a significant effect on reducing energy consumption. However, storing compressed data/instructions has a performance and energy overhead associated with decompression at runtime. The underlying compression algorithm, the corresponding implementation of the decompression and the ability to reuse decompressed information critically impact this overhead. In this paper, we explore the influence of compression on overall memory energy using a commercial embedded Java virtual machine (JVM) and a customized compression algorithm. Our results show that compression is effective in reducing energy even when considering the runtime decompression overheads for most applications. Further, we show a mechanism that selectively compresses portions of the memory to enhance energy savings. Finally, a scheme for clustering the code and data to improve the reuse of the decompressed data is presented.