• Proceedings of the KIPE Conference
• /
• 2020.08a
• /
• pp.278-279
• /
• 2020

최근 전기차 배터리는 일 주행거리 증대를 위해서 50~100 kWh 급으로 대용량화 되고 있다. 그와 동시에 충전시간 단축을 위해서 배터리 전압은 400V에서 800V로 높아지고 급속충전기 용량은 50kW에서 350kW 급으로 대용량화 되고 있다. 본 논문에서는 전기차 증가 추세에 따라 기존 주유소를 대체하는 MW급 전기차 급속충전소를 위한 에너지 저장 장치(Energy Storage System, ESS) 연계 지능형 반도체 변압기(Solid State Transformer, SST) 기술 기반의 전력변환 시스템 모델을 제안하고자 한다. 이 모델은 배전계통에 직접 연계가 가능하기 때문에 대용량 전기차 급속충전소뿐만 아니라 DC Grid 구축에 응용 가능하다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.11a
• /
• pp.383-385
• /
• 2022

텍스트 요약(Text Summarization)은 자연어 처리(NLP) 분야의 주요 작업 중 하나이다. 높은 정확성을 보이는 문서 요약 딥 러닝 모델을 만들기 위해서 대용량 학습 데이터가 필요한데, 대용량 데이터 전처리 과정에서 처리 시간, 메모리 관리 등과 같은 문제가 발생한다. 본 논문에서는 대규모 병렬처리 플랫폼 Apache Spark 를 사용해 추상 요약 딥 러닝 모델의 데이터 전처리 과정을 개선하는 방법을 제안한다. 실험 결과 제안한 방법이 기존 방법보다 데이터 전처리 시간이 개선된 결과를 보이고 있다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.33-34
• /
• 2023

성장하는 빅 데이터 시장과 빅 데이터 수의 기하급수적인 증가는 기존 컴퓨팅 환경에서 데이터 처리의 어려움을 야기한다. 특히 이미지 데이터 처리 속도는 데이터양이 많을수록 현저하게 느려진다. 이에 본 논문에서는 Apache Spark와 OpenCV를 활용한 분산 클러스터 컴퓨팅 환경의 대용량 이미지 머신러닝 시스템을 제안한다. 제안하는 시스템은 Apache Spark를 통해 분산 클러스터를 구성하며, OpenCV의 이미지 처리 알고리즘과 Spark MLlib의 머신러닝 알고리즘을 활용하여 작업을 수행한다. 제안하는 시스템을 통해 본 논문은 대용량 이미지 데이터 처리 및 머신러닝 작업 속도 향상 방법을 제시한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2007.11a
• /
• pp.579-582
• /
• 2007

최근 들어 더욱 UCC(User Created Contents)등과 같은 대용량 멀티미디어(multimedia) 서비스에 대한 요구가 나날이 증가되면서 부하분산에 중점을 둔 웹 클러스터링 시스템(Web Clustering System) 에서 기존의 작은 크기의 스트림 데이터(Stream Data)나 조금 더 다양한 데이터를 위한 읽기 성능을 대용량 데이터에 초점을 맞춘 방안으로 최적화 시키는 것이 중요시되고 있다. 본 논문에서는 대용량 멀티미디어를 중심으로 실제 서비스시 간과 되어질 수 있는 운영체제(Operating System, O/S)에서의 I/O 인식, 디스크 제어 프로그램에서의 I/O, 웹 클러스터의 부하분산정책의 파라미터(Parameter)를 개선함으로써 읽기성능 향상 방안을 제시한다.

• Journal of KIISE:Computer Systems and Theory
• /
• v.30 no.1
• /
• pp.33-49
• /
• 2003

Recently, there have been large storage demands for manipulating multimedia data. To solve the tremendous storage demands, one of the major researches is the SAN(Storage Area Network) that provides the local file requests directly from shared-disk storage and also eliminates the server bottlenecks to performance and availability. SAN also improve the network latency and bandwidth through new channel interface like FC(Fibre Channel). But to manipulate the efficient storage network like SAN, traditional local file system and distributed file system are not adaptable and also are lack of researches in terms of a metadata structure for large-scale inode object such as file and directory. In this paper, we describe the architecture and design issues of our shared-disk file system and provide the efficient bitmap for providing the well-formed block allocation in each host, extent-based semi flat structure for storing large-scale file data, and two-phase directory structure of using Extendible Hashing. Also we describe a detailed algorithm for implementing the file system's device driver in Linux Kernel and compare our file system with the general file system like EXT2 and shard disk file system like GFS in terms of file creation, directory creation and I/O rate.

• Journal of KIISE
• /
• v.42 no.3
• /
• pp.307-319
• /
• 2015

Current ontology studies use the Hadoop distributed storage framework to perform map-reduce algorithm-based reasoning for scalable ontologies. In this paper, however, we propose a novel approach for scalable Web Ontology Language (OWL) Horst Lite ontology reasoning, based on distributed cluster memories. Rule-based reasoning, which is frequently used for scalable ontologies, iteratively executes triple-format ontology rules, until the inferred data no longer exists. Therefore, when the scalable ontology reasoning is performed on computer hard drives, the ontology reasoner suffers from performance limitations. In order to overcome this drawback, we propose an approach that loads the ontologies into distributed cluster memories, using Spark (a memory-based distributed computing framework), which executes the ontology reasoning. In order to implement an appropriate OWL Horst Lite ontology reasoning system on Spark, our method divides the scalable ontologies into blocks, loads each block into the cluster nodes, and subsequently handles the data in the distributed memories. We used the Lehigh University Benchmark, which is used to evaluate ontology inference and search speed, to experimentally evaluate the methods suggested in this paper, which we applied to LUBM8000 (1.1 billion triples, 155 gigabytes). When compared with WebPIE, a representative mapreduce algorithm-based scalable ontology reasoner, the proposed approach showed a throughput improvement of 320% (62k/s) over WebPIE (19k/s).

• Journal of Korea Spatial Information System Society
• /
• v.10 no.2
• /
• pp.1-12
• /
• 2008

Automatic semiconductor manufacture system generating transaction from 50,000 to 500,000 per a second needs storage management system processing very large data at once. A lot of storage management systems are researched for storing very large data. Existing storage management system is typical DBMS on a disk. It is difficult that the DBMS on a disk processes the 500,000 number of insert transaction per a second. So, the DBMS on main memory appeared to use memory. But it is difficultthat very large data stores into the DBMS on a memory because of limited amount of memory. In this paper we propose storage management system using insert transaction of a block unit that can process insert transaction over 50,000 and store data on low storage cost. A transaction of a block unit can decrease cost for a log and index per each tuple as transforming a transaction of a tuple unit to a block unit. Besides, the proposed system come cost to decompress all block of data because the information of each field be loss. To solve the problems, the proposed system generates the index of each compressed block to prevent reducing speed for searching. The proposed system can store very large data generated in semiconductor system and reduce storage cost.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.49 no.2
• /
• pp.23-31
• /
• 2012

With advance of high bandwidth network infrastructure, the requirement is dramatically increasing to cooperate between the users who are far from each other and make use of bulk data. However, as the prominent data transmission protocol, it is well known that TCP suffers some degrees of inefficiency for bulk data transmission when RTT is relatively big. So, some works are on going to suggest a new transmission method to utilize the bandwidth in effective. UDT(UDP-based Data Transfer protocol) is one of these. It is a UDP based application level protocol which can guarantee reliability and stability. much like as TCP. In this paper, we present a design and implementation of UDT based bulk data transmission tool by applying parallel and compressive techniques. The implementation result is examined to measured its performance improvement on a real test-bed, and then compared with existing bulk data transmission tools. Experimental results show that proposed tool is more stable and shows greater performance than that of native UDT. Especially, the performances show 244% improvement in RTT 400ms without losses and 229% in RTT 250ms with 0.005% losses respectively.

• KIPS Transactions on Software and Data Engineering
• /
• v.3 no.10
• /
• pp.397-406
• /
• 2014

In order to answer the questions successfully on behalf of the human in DeepQA environments such as Jeopardy! of the American quiz show, the computer is required to have the capability of fast temporal and spatial reasoning on a large-scale commonsense knowledge base. In this paper, we present a scalable spatial reasoning algorithm for deriving efficiently new directional and topological relations using the MapReduce framework, one of well-known parallel distributed computing environments. The proposed reasoning algorithm assumes as input a large-scale spatial knowledge base including CSD-9 directional relations and RCC-8 topological relations. To infer new directional and topological relations from the given spatial knowledge base, it performs the cross-consistency checks as well as the path-consistency checks on the knowledge base. To maximize the parallelism of reasoning computations according to the principle of the MapReduce framework, we design the algorithm to partition effectively the large knowledge base into smaller ones and distribute them over multiple computing nodes at the map phase. And then, at the reduce phase, the algorithm infers the new knowledge from distributed spatial knowledge bases. Through experiments performed on the sample knowledge base with the MapReduce-based implementation of our algorithm, we proved the high performance of our large-scale spatial reasoner.

• The Journal of the Korea Contents Association
• /
• v.15 no.5
• /
• pp.570-581
• /
• 2015

In recent years, many institutions predict that cloud services and big data will be popular IT trends in the near future. A number of leading IT vendors are focusing on practical solutions and services for cloud and big data. In addition, cloud has the advantage of unrestricted in selecting resources for business model based on a variety of internet-based technologies which is the reason that provisioning and virtualization technologies for active resource expansion has been attracting attention as a leading technology above all the other technologies. Big data took data prediction model to another level by providing the base for the analysis of unstructured data that could not have been analyzed in the past. Since what cloud services and big data have in common is the services and analysis based on mass amount of data, efficient operation and designing of mass data has become a critical issue from the early stage of development. Thus, in this paper, I would like to establish data processing architecture based on technological requirements of mass data for cloud and big data services. Particularly, I would like to introduce requirements that must be met in order for distributed file system to engage in cloud computing, and efficient compression technology requirements of mass data for big data and cloud computing in terms of cost-saving, as well as technological requirements of open-source-based system such as Hadoop eco system distributed file system and memory database that are available in cloud computing.