Earticle

LR-PON은 수동형 광가입자망 기술을 사용하여 서비스 영역이 100 km이상 확장된 광대역 가입자망이다. 서비스 영역이 확장된 LR-PON의 경우 OLT와 ONU들간 전송 지연시간이 큰 폭으로 증가하는 문제점으로 인해 효과적인 상향 대역할당에 어려움이 있다. 본 논문에서는 LR-PON망에서 동적인 상향 대역 할당을 효과적이며 공정하게 하기 위해서, 고정형 폴링 주기를 갖는 다중 스레드 기반의 새로운 동적대역할당 방안을 제안한다. 제시한 동적대역할당 방안의 우수성을 입증하기 위하여 상향 트래픽의 평균 지연시간을 기존의 DBA 방안들과 비교하였다. 특히, CoS 특성 분석을 위하여 다양한 트래픽 로드별로 지연 특성을 분석하였다.

Long-Reach PON (LR-PON) is a broadband access network using passive optical network (PON) technology which the reach is extended to 100 km or higher. A major challenge in LR-PON is that the propagation delay between OLT and ONUs is increased by a very significant amount. To effectively and fairly distribute the upstream bandwidth dynamically in LR-PON, we propose a new periodic and multi-thread polling based dynamic bandwidth allocation (DBA) algorithm. We compare the proposed algorithm with traditional DBAs and show its advantage on average packet delay. Numerical results are analyzed under varying offered loads.

With the recent progress in flexible manufacturing systems (FMS) in industry, increasing attention has been given to Automatic Guided Vehicle (AGV) systems. An AGV is a self-powered unit for transporting materials between stations without needing to be controlled by an operator. Such a system has several sensors to recognize the external state, and it is designed to travel between stations automatically without external assistance. To manage each device automatically in real time it requires a distributed controller with a main computer as the host, as well as a number of micro-controllers. In this study, an AGV system with dual motor drive was constructed. A Pentium 4 personal computer was set up as the main host for the distributed control, and this communicated with other micro-controllers in the management of the motor. The speed of each motor was also controlled by a micro-controller.

IP보안 카메라의 보급으로 원격에서 얼굴인식을 수행함에 있어 서버의 부하를 줄이기 위한 여러 가지 방법들이 구현되고 있다. 본 논문에서는 원격지에 있는 IP 보안 카메라 영상을 얼굴검출기능이 탑재된 DSP 보드를 통해 입력받아 얼굴검출을 수행 한 후 해당 얼굴영역 이미지를 서버로 전송하여 이를 얼굴인식 분산 처리를 통해 얼굴인식 기능을 수행한다. 결과적으로 전체적인 서버시스템 로드를 상당히 줄이는 성과와 실시간 얼굴 인식처리를 최대 256 대의 카메라를 연동하면서 수행할 수 있는 장점을 가지고 있다. 이를 수행할 수 있는 기술은 분산처리 서버기술을이용하여 한 서버 당 64채널 얼굴인식을 수행하며, 4개 분산처리 서버를 운영할 경우 250여개 카메라 채널을 통한얼굴검출 결과를 처리하는 성과를 가져올 수 있었다.

Various methods for reducing the load on the server have been implemented in performing face recognition remotely by the spread of IP security cameras. In this paper, IP surveillance cameras at remote sites are input through a DSP board equipped with face detection function, and then face detection is performed. Then, the facial region image is transmitted to the server, and the face recognition processing is performed through face recognition distributed processing . As a result, the overall server system load and significantly reduce processing and real-time face recognition has the advantage that you can perform while linked up to 256 cameras. The technology that can accomplish this is to perform 64-channel face recognition per server using distributed processing server technology and to process face search results through 250 camera channels when operating four distributed processing servers there was.

멀티 코어 프로세서의 보급 확산으로 최근에는 임베디드 시스템에서도 채택되고 있다. 따라서 일반적으로 대규모의 컴퓨팅과 메모리 접근을 필요로 하는 멀티미디어 응용은 멀티 코어 플랫폼 기반의 병렬화가 가능하다. 본 논문에서는 멀티 코어 CPU을 이용한효율적 색 공간 변환을 위한 스레드 수준 병렬 기법의 성능 향상을 검증하였다. 스레드 수준 병렬화 특히 멀티 코어 프로세서기반공유 메모리 컴퓨팅 시스템에서는 매우 유용한 병렬 처리 패러다임이 되고 있다. 본 구현에서 스레드 수준 병렬화는 각 스레드에다른 입력 픽셀을 할당하여 실행하였다. 성능 평가를 위해 직렬 및 병렬 구현들 사이의 처리 속도의 비교에 기초하여 대표적 멀티코어 프로세서에서 색 변환을 위한 성능 향상 정도를 평가하였다. 결과는 스레드 수준의 병렬 구현에 관계없이 다른 멀티 코어에서전반적으로 비슷한 성능 향상의 비율을 보여주었다.

With the increasing popularity of multi-core processors, they have been adopted even in embedded systems. Under thiscircumstance many multimedia applications can be parallelized on multi-core platforms because they usually require heavycomputations and extensive memory accesses. This paper proposes an efficient thread-level parallel implementation for colorspace conversion on multi-core CPU. Thread-level parallelism has been becoming very useful parallel processing paradigmespecially on shared memory computing systems. In this work, it is exploited by allocating different input pixels to eachthread for concurrent loop executions. For the performance evaluation, this paper evaluate the performace improvements forcolor conversion on multi-core processors based on the processing speed comparison between its serial implementation andparallel ones. The results shows that thread-level parallel implementations show the overall similar ratios of performanceimprovements regardless of different multi-cores.

With the growth of increasingly complex in wireless ad-hoc networks, performance modeling and evaluation play a crucial part in their design process to ensure successful deployment and exploitation in practice. Compared with the actual implementation (testbeds), simulation has become a valuable tool, which allows studying large scale systems that cannot be built practically. However, most simulators currently used for such networks are intended to evaluate individual routing protocol, which should be extracted from the source code and rewritten in accordance with the simulator’s requirements. Thus, in this paper, we design a new wireless ad-hoc network simulator, called Time Step-based Wireless Ad-hoc Network Simulator (TimSim). It is intended to facilitate the migration of simulation code to real devices via providing useful APIs with real device driver. We use a discrete event processor with time step-based feature as the simulating engine, and the data can be transmitted in bit-level. A unique feature of TimSim is its ability to support the simulation of multi-thread programming. Additionally, we import database access to store the simulation information. Extensive experiments have been conducted to evaluate the efficiency of TimSim. Furthermore, a case study on a real system has been presented to illustrate its practicability.

With the rapid development of information technology, the data scale expands unceasingly, produce a large amount of image data, a massive image data processing has also is an important technology. Traditional processing techniques, in dealing with such a large scale of image data, have been unable to meet the requirements. One of the important reason is that whether the c ++ or Java, there is memory leak problem, when the data is the relatively small size, these problems may be insignificant, but when the data is very large, will highlight these shortcomings, will lead to serious program crashes, the system outage, ultimately unable to achieve expected goal. In considering the data processing efficiency, stability, is proposed in this paper, based on multithreading, a dynamic self-healing capacity of mass image data processing technology, compared with the traditional technology can improve the efficiency and improve the stability. Finally, after the experimental results verify that the method is effective.

In order to meet the different data browsing requests for different users to the visual quality of remote sensing images in heterogeneous network especially under narrow bandwidth environments, an online remote sensing image progressive transmission model is constructed in which remote sensing image compression and decompression are synchronized with transmission. At the same time, a pipeline-based multi-threaded acceleration method has been proposed through solving the asynchronous problem between compression decompression and transmission to improve the efficiency of remote sensing progressive transmission. At last, an idea of retry broken downloads transmission interruption has been implemented to improve end-user interactive experience. Experimental results show that the whole processing speed has been improved nearly twice without reducing image transmission quality and the amount of data transmission was reduced evidently by using the proposed progressive transmission and real-time compression model.

Multi-thread is a mature and widely-used programming mode for multitasking applications developing, especially in Java. However java.lang.Thread is not designed for HPC parallel programming. Java multi-thread program is confined within single compute host. CommonJ Work Manager is a joint specification proposed by IBM and BEA. Work Manager is a higherlevel alternative to using Java Thread. Originally, Work Manager is used to solve the constraint of opening independent threads in J2EE container. Although there are some enterprise servers such as WebLogic and WebSphere which have supported Work Manager, they are not aim at large scale parallel applications. They do not have a grid-enabled and expandable architecture, and the multi-thread programs are still executed on single compute host. Currently, grid-enabled parallel programming model like MPICH-G2 is limited with C/C++ implementations. There are no official bindings for MPI with Java. In this paper, we present a grid-enabled multi-thread programming model: SymWorkManager to enable multithread programs to run on distributed environments. SymWorkManager is designed as a high performance computing (HPC) implementation of CommonJ Work Manger specification using Platform Symphony. It provides J2EE developers a simpler way to program parallel multi-thread applications under a HPC platform.

본 논문은 철판 표면에 발생하는 결함을 검출하는 방법 중 이미지 프로세싱을 이용한 철판결함 검출 시스템을 제안하였다. 기존 이미지 프로세싱은 특정 주파수대의 빛 에너지를 방출하는 광원과 그 광원에 응답하는 수광 소자를 이용하는 이미지 프로세싱 기법을 사용하였으나, 본 논문에서는 자연 광원을 이용하고 광원의 주파수 특성에 의존하지 않고 이미지 프로세싱과 멀티 스레드 기법을 사용하여 결함을 검출하였다. 이 검출방법의 장점은 기존 시스템의 단점이었던 검증 방법의 난해함으로 사용자가 쉽게 접근하지 못하는 단점을 보완하여, 철판의 표면을 실시간으로 모니터링 가능하도록 하였다. 또한 발광소자와 수광 소자의 주파수 매칭을 위해 각각의 하드웨어를 주문제작 함으로 많은 비용이 소모되는 것을 시중에서 구매 가능한 하드웨어를 기본으로 구성하여 비용을 절감하였다.

The purpose of this research is to propose a system to detect a strip defect on a iron plate using an image processing, one way of finding defects on an iron plate. An existing way of image processing is using a light source which release a light energy in a certain frequency and a light absorbing display which responds to the light source. This research attempts to detect defects by using a image processing and multi-Tread which handles an illumination, without depending on characteristics of light frequency. One of the advantages of this method is that it makes up for the weakness of the existing method which was too difficult for users to notice a defect. Also this method makes it possible to realize a real-time monitoring on a plate of iron. The other advantage of this method is that it reduces the price of hardwares on demand to match the frequency of light emitting display and light absorbing display because this method only needs a hardware which is easy to buy in any market.

대규모 병렬 워크로드를 GPGPU의 연산 유닛에 할당하기 위한 스케줄링으로 라운드 로빈 방식이 널리 사용되고 있다. 라운드 로빈은 작업을 각 연산 유닛에 순차적으로 할당하여 구현이 쉽다는 장점이 있으나, 클라우드와 같은 다중 워크로드 환경에서는 연산 유닛 간 부하 균형이 잘 이루어지지 않는 문제점이 발생한다. 본 논문에서는 이러한 문제를 해결하기 위해 새로운 스레드 블록 스케줄링을 제안한다. 제안하는 방식은 다양한 GPGPU 워크로드가 만들어낸 스레드 블록들을 그 작업량에 근거해 다중큐로 관리하고 각 연산 유닛의 잔여 자원을 가장 잘 활용할 수 있는 큐에서 스레드 블록을 선택하여 연산 유닛들의 자원 이용률을 극대화시키고 부하균형을 유도한다. 다양한 부하 환경에서의 시뮬레이션 실험을 통해 제안하는 방식이 라운드 로빈 대비 평균 24.8%의 성능개선 효과가 있음을 보인다.

Round-robin is widely used for the scheduling of large-scale parallel workloads in the computing units of GPGPU. Round-robin is easy to implement by sequentially allocating tasks to each computing unit, but the load balance between computing units is not well achieved in multi-workload environments like cloud. In this paper, we propose a new thread block scheduling policy to resolve this situation. The proposed policy manages thread blocks generated by various GPGPU workloads with multiple queues based on their computation loads and tries to maximize the resource utilization of each computing unit by selecting a thread block from the queue that can maximally utilize the remaining resources, thereby inducing load balance between computing units. Through simulation experiments under various load environments, we show that the proposed policy improves the GPGPU performance by 24.8% on average compared to Round-robin.

본 논문에서는 RTSP 프로토콜을 제어하기 위한 서버 시스템을 IOCP 기반의 Multi-Thread 기법을 이용하여 구현하는 방법을 소개한다. 다수의 클라이언트에 대한 응답을 Thread로 구성하는 부분에서 Multi-Threading을 이용함으로써 수행 속도를 높이고 Winsock2에서 제공하는 IOCP(T/O Completion Port)를 이용하여 견고하고 확장이 용이한 RTSP(Real Time Streaming Protocol) 스트리밍 서버를 개발하였다.

현재 개발되고 있는 Shader 프로세서는 처리 성능을 높이기 위하여 Multi-Core, Multi-Thread를 채택하고 있다. 또한 Shader 프로세서에서 각 수행 단계별 마다 IP를 따로 구현하지 않고 하나의 Core IP를 다양한 목적으로 사용할 수 있도록 설계하고 있다. 본 논문에서는 이러한 목적에 맞게 Shader-Core를 이용하여 연산이 가능하고, Multi-Core, Multi-Thread 기반에서 픽셀의 병렬처리가 가능하도록 고안된 Vector 기반의 Rasterization알고리즘을 제안한다. 이를 통하여 동일 조건의 기존 알고리즘에 비하여 약 2%의 연산량을 가지면서 각 픽셀이 독립적으로 연산이 가능하도록 하였다.

A Multi-Core/Multi-Thread architecture is adopted for the Shader processor to enhance the processing performance. The Shader processor is designed to utilize its processing core IP for multiple purposes, such as Vertex-Shading, Rasterization, Pixel-Shading, etc. In this paper, we propose a 'Rasterization based on Vector Algorithm' that makes parallel pixels processing possible with Multi-Core and Multi-Thread architecture on the Shader Core. The proposed algorithm takes only 2% operation counts of the Scan-Line Algorithm and processes pixels independently.

A novel a-Si TFT integrated gate driver circuit using multi-thread driving has been developed. The circuit consists of two independent shift registers alternating between the two modes, "wake" and "sleep". The degradation of the circuit is retarded because the bias stress is removed during the sleep mode. It has been successfully integrated in 14.1-in. XGA LCD Panel, showing enhanced stability.

Online game servers usually has been using the static thread pool system. But this system is not fit for huge online game server because the overhead is always up-and-down. Therefore, in this paper, we suggest the new algorithm for huge online game server. This algorithm is based on the prediction-based dynamic thread pool system. But it was developed for web servers and every 0.1 seconds the system prediction the needed numbers of threads and determine the thread pool size. Some experimental results show that the check time of 0.4 seconds is the best one for online game server and if the number of worker threads do not excess or lack to the given threshold then we do not predict and keep the current state. Otherwise we apply the prediction algorithm and change the number of threads. Some experimental results shows that this proposed algorithm reduce the overhead massively and make the performance of huge online game server improved in comparison to the static thread pool system.

Different cores typically share the last-level cache in a multi-core processor. Threads running on different cores may interfere with each other. Therefore, the multi-core worst-case execution time (WCET) analyzer must be able to safely and accurately estimate the worst-case inter-thread cache interference. This is not supported by current WCET analysis techniques that manly focus on single thread analysis. This paper presents a novel approach to analyze the worst-case cache interference and bounding the WCET for threads running on multi-core processors with shared L2 instruction caches. We propose to use an interference matrix to model inter-thread interference, on which basis we can calculate the worst-case inter-thread cache interference. Our experiments indicate that the proposed approach can give a worst-case bound less than 1%, as in benchmark fib-call, and an average 16.4% overestimate for threads running on a dual-core processor with shared-L2 cache. Our approach dramatically improves the accuracy of WCET overestimatation by on average 20.0% compared to work.

최근까지 도시시설물 관리를 위해 활용된 GIS 기술은 주로 2차원 기반의 GIS 기술이었다. 하지만 도시시설물은 3차원 공간상에 존재하며 이를 2차원으로 추상화시키면 정보의 손실이 일어나게 된다. 또한 도시 공간 내에 점차 시설물의 수가 증가하고 있고 대부분의 도시시설물이 지하공간에 위치하기 때문에 2차원 기반의 시스템에서 도시 내 모든 시설물의 정보를 통시에 파악하고 관리하는 것이 어려워지고 있다 본 논문에서는 멀티 텍스처 기법을 이용한 다차원 도시시설물 관리 시스템을 제안한다. 제안된 시스템은 3차원 상에서 도시시설물을 가시화하여 데이터의 손실을 줄이고 정보의 가독성을 높이며, 2차원 벡터 데이터의 3차원 라스터 변환을 통한 가시화 기법을 개발하여 다양한 2차원 GIS 의 공간분석 결과를 3차원 데이터와 동시에 활용할 수 있도록 한다. 또한 3차원 데이터를 처리하는 경우에 시스템의 속도 및 성능이 저하되어 활용 효율성이 떨어지는 문제를 해결하기 위하여 다중 스레드 프로세스를 적용하였다. 본 논문에서 제안한 기술은 기존에 텍스처를 통해 보여주던 센서의 상태정보를 3차원 영상, 시설물 데이터와 함께 시각정보로 표출함으로써 직관적인 시설물 모니터링이 가능해졌다는데 의의가 있다.

Recently, 2D GIS technology is applied for urban facility management. However, urban facilities are located in 3D space and the information loss is occurring during data abstraction from 3D urban facility to 2D object. Also, the number of urban facilities is increasing steadily and most of urban facilities are located in underground space in the city. Therefore 2D urban facility management system has a limitation on visualization and management for a large number of urban facilities. In this paper, a multi-dimensions urban facility management system based on multi-texture technology is proposed. The proposed system reduces the information loss and improves the readability of information by visualizing urban facilities on 3D virtual space. A multi-texturing technology is applied for integrating of 2D vector data and 3D raster data, and a multi-thread technology is used for improving speed and performance of the system. The proposed technology can be used as a guideline for urban facility monitoring as providing visual information of a facility status with 3D image and facility data.

맡은 사용자 영역의 프로그램들이 확장성과 성능 향상을 목적으로 다중 프로세스 모델에서 다중 스레드 모델로 바뀌고 있다. 웹 서버도 폭발적으로 증가하는 사용자 요청을 수용할 수 있도록 확장성과 성능 향상을 위해서 다중 스레드 모델을 적용하였다. 본 논문에서는 다중 프로세스 모델을 적용한 웹 서버와 다중 프로세스/다중 스레드 모델을 적용한 웹 서버에 대한 성능을 여러 운영체제 환경에서 평가하였다. 실험 결과 전체적으로 다중 프로세스 모델이 다중 스레드 모델보다 더 좋은 성능과 안정성을 보였다.