• Journal of KIISE
• /
• v.43 no.1
• /
• pp.61-70
• /
• 2016

In recent years, there has been a need for techniques for large-scale ontology inference in order to infer new knowledge from existing knowledge at a high speed, and for a diversity of semantic services. With the recent advances in distributed computing, developments of ontology inference engines have mostly been studied based on Hadoop or Spark frameworks on large clusters. Parallel programming techniques using GPGPU, which utilizes many cores when compared with CPU, is also used for ontology inference. In this paper, by combining the advantages of both techniques, we propose a new method for reasoning large RDFS ontology data using a Spark in-memory framework and inferencing distributed data at a high speed using GPGPU. Using GPGPU, ontology reasoning over high-capacity data can be performed as a low cost with higher efficiency over conventional inference methods. In addition, we show that GPGPU can reduce the data workload on each node through the Spark cluster. In order to evaluate our approach, we used LUBM ranging from 10 to 120. Our experimental results showed that our proposed reasoning engine performs 7 times faster than a conventional approach which uses a Spark in-memory inference engine.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2012.04a
• /
• pp.1118-1120
• /
• 2012

컴퓨터 메모리의 용량이 커지고 기술이 발전하며 메모리와 저장장치의 데이터 처리속도 차이는 나날이 커지고 있다. 이를 보완하고자 데이터 처리를 가급적 메모리에서 해결하여 처리속도를 높이고자 하는 연구가 많이 있다. 그 중 MapReduce에 대한 연구는 현재 주목이 되고 있는 분야이다. MapReduce는 빅데이터를 클러스터 환경에서 처리하기에 대중적인 프로그래밍 모델이다. 본 논문은 MapReduce 기반의 Hadoop을 SSD를 적용하여 실행속도를 증진시키려 한다. 전통적인 MapReduce 모델은 데이터를 정렬하는데에 I/O가 크게 발생하는데, MapRedce가 사용하는 병합정렬의 I/O 병목현상을 개선하고자 SSD를 사용하였다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2016.10a
• /
• pp.442-444
• /
• 2016

과거에 비해 비약적으로 생산되는 공간 데이터에 대한 처리를 위한 공간 연산은 빠른 처리 응답성을 요구하는 경우가 많다. 그래서 최근 하둡(Hadoop)과 같은 빅데이터 처리 시스템을 이용하여 처리하고자 하는 시도가 많다. 한편, 공간 조인은 데이터 분할(Partitioning)과 공간 색인의 이용 여부, 여과 단계와 정제 단계를 거치는 등 그 복잡도가 강한 공간 연산이다. 그래서 빅데이터 처리 시스템을 이용한 공간 조인의 처리 방식은 매우 다양하다. 그러나 지금까지 이러한 공간 조인의 처리 방식에 다른 리소스 활용에 대한 비교는 거의 없다. 이 논문에서는 다양한 공간 연산의 수행 방법에 따른 빅데이터 시스템 클러스터에서 데이터 전송 방식을 고찰하고 데이터 전송에 따른 네트워크 리소스의 효율적인 사용 방안을 제안하고자 한다. 구체적으로 단일할당과 다중할당 색인 기법의 비교, 파티셔닝 방법의 비교, 맵리듀스 시스템의 태스크 할당 방법에 따른 비교를 통해 다양한 연산 유형에 따른 공간 조인의 처리 방안 선정에 고려 요소를 제시하고자 한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2013.11a
• /
• pp.740-743
• /
• 2013

클라우드 컴퓨팅 환경에서는 사용자의 데이터를 수많은 분산서버를 이용하여 데이터를 암호화하여 저장한다. 구글, 야후 등 글로벌 인터넷 서비스 업체들은 인터넷 서비스 플랫폼의 중요성을 인식하고 자체 연구 개발을 수행, 저가 상용 노드를 기반으로 한 대규모 클러스터 기반의 클라우드 컴퓨팅 플랫폼 기술을 개발 활용하고 있다. 이와 같이 분산 컴퓨팅 환경에서 다양한 데이터 서비스가 가능해지면서 대용량 데이터의 분산관리가 주요 이슈로 떠오르고 있다. 한편, 대용량 데이터의 다양한 이용 형태로부터 악의적인 공격자나 내부 사용자에 의한 보안 취약성 및 프라이버시 침해가 발생할 수 있다. 특히, 하둡에서 데이터 블록의 권한 제어를 위해 사용하는 블록 접근 토큰에도 다양한 보안 취약점이 발생한다. 이러한 보안 취약점을 보완하기 위해 본 논문에서는 비밀분산 기반의 블록 접근 토큰 관리 기법을 제안한다.

• KIPS Transactions on Software and Data Engineering
• /
• v.5 no.5
• /
• pp.231-242
• /
• 2016

In order to meet rapid changes of Web information, it is necessary to extend the current Web technologies to represent both the valid time and location of each fact and knowledge, and reason their relationships. Until recently, many researches on qualitative temporal reasoning have been conducted in laboratory-scale, dealing with small knowledge bases. However, in this paper, we propose the design and implementation of a parallel qualitative temporal reasoner, MRQUTER, which can make reasoning over Web-scale large knowledge bases. This parallel temporal reasoner was built on a Hadoop cluster system using the MapReduce parallel programming framework. It decomposes the entire qualitative temporal reasoning process into several MapReduce jobs such as the encoding and decoding job, the inverse and equal reasoning job, the transitive reasoning job, the refining job, and applies some optimization techniques into each component reasoning job implemented with a pair of Map and Reduce functions. Through experiments using large benchmarking temporal knowledge bases, MRQUTER shows high reasoning performance and scalability.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.13 no.1
• /
• pp.207-212
• /
• 2018

As a data analysis tool which is becoming popular in the Big Data era, R is rapidly expanding its user range by providing powerful statistical analysis and data visualization functions. Major advantage of R is its functional scalability based on open source, but its scale scalability is limited, resulting in performance degrades in large data processing. RHadoop, one of the extension packages to complement it, can improve data analysis performance as it supports Hadoop platform-based distributed processing of programs written in R. In this paper, we evaluate the validity of RHadoop by evaluating the performance improvement of RHadoop in real medical big data analysis. Performance evaluation of the analysis of the medical history information, which is provided by National Health Insurance Service, using R and RHadoop shows that RHadoop cluster composed of 8 data nodes can improve performance up to 8 times compared with R.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.21 no.9
• /
• pp.1731-1737
• /
• 2017

K Nearest Neighbor Join (KNN Join) is a simple yet effective method in machine learning. It is widely used in small dataset of the past time. As the number of data increases, it is infeasible to run this model on an actual application by a single machine due to memory and time restrictions. Nowadays a popular batch process model called MapReduce which can run on a cluster with a large number of computers is widely used for large-scale data processing. Hadoop is a framework to implement MapReduce, but its performance can be further improved by a new framework named Spark. In the present study, we will provide a KNN Join implement based on Spark. With the advantage of its in-memory calculation capability, it will be faster and more effective than Hadoop. In our experiments, we study the influence of different factors on running time and demonstrate robustness and efficiency of our approach.

• Journal of the Korean Data and Information Science Society
• /
• v.28 no.4
• /
• pp.755-768
• /
• 2017

As Big Data becomes the core of the fourth industrial revolution, big data-based processing and analysis capabilities are expected to influence the company's future competitiveness. Comparative studies of RHadoop and RHIPE that integrate R and Hadoop environment, have not been discussed by many researchers although RHadoop and RHIPE have been discussed separately. In this paper, we constructed big data platforms such as RHadoop and RHIPE applicable to large scale data and implemented the machine learning algorithms such as multiple regression and logistic regression based on MapReduce framework. We conducted a study on performance and scalability with those implementations for various sample sizes of actual data and simulated data. The experiments demonstrated that our RHadoop and RHIPE can scale well and efficiently process large data sets on commodity hardware. We showed RHIPE is faster than RHadoop in almost all the data generally.

• KIISE Transactions on Computing Practices
• /
• v.21 no.10
• /
• pp.651-657
• /
• 2015

The amount of web traffic has increased as a result of the rapid growth of the use of web-based applications. In order to obtain valuable information from web logs, we need to develop systems that can support interactive, flexible, and efficient ways to analyze and handle large amounts of data. In this paper, we present CERES, a log-based, interactive web analytics system for backbone networks. Since CERES focuses on analyzing web log records generated from backbone networks, it is possible to perform a web analysis from the perspective of a network. CERES is designed for deployment in a server cluster using the Hadoop Distributed File System (HDFS) as the underlying storage. We transform and store web log records from backbone networks into relations and then allow users to use a SQL-like language to analyze web log records in a flexible and interactive manner. In particular, we use the data cube technique to enable the efficient statistical analysis of web log. The system provides users a web-based, multi-modal user interface.

• KIISE Transactions on Computing Practices
• /
• v.21 no.12
• /
• pp.774-779
• /
• 2015

In this paper, we present the design and implementation of a large-scale qualitative spatial reasoner using Apache Spark, an in-memory high speed cluster computing environment, which is effective for sequencing and iterating component reasoning jobs. The proposed reasoner can not only check the integrity of a large-scale spatial knowledge base representing topological and directional relationships between spatial objects, but also expand the given knowledge base by deriving new facts in highly efficient ways. In general, qualitative reasoning on topological and directional relationships between spatial objects includes a number of composition operations on every possible pair of disjunctive relations. The proposed reasoner enhances computational efficiency by determining the minimal set of disjunctive relations for spatial reasoning and then reducing the size of the composition table to include only that set. Additionally, in order to improve performance, the proposed reasoner is designed to minimize disk I/Os during distributed reasoning jobs, which are performed on a Hadoop cluster system. In experiments with both artificial and real spatial knowledge bases, the proposed Spark-based spatial reasoner showed higher performance than the existing MapReduce-based one.