Earticle

Massive multiple-input multiple-output (MIMO) systems have become an essential feature in the fifth-generation (5G) wireless communication networks. Due to the commercialization of 5G mobile communication devices, the number of multi-antenna devices have increased over the years. Therefore, a precoding scheme based on multi-antenna users is an inevitable feature that needs to be analyzed in massive MIMO systems. The asymptotic ergodic sum-rate analysis of block diagonalization (BD) based multiuser MIMO systems is performed in this study. Based on the random matrix theory results derived from the massive base station antenna characteristics, the singular values of the block diagonalized channel are obtained asymptotically, and thereby, the closed-form expression for the BD based sum-rate is derived. The asymptotic sum-rate performance gap between the BD and the zero-forcing beamforming is then analyzed. Moreover, from the derived asymptotic sum-rate, the optimal number of users needed to maximize the ergodic sum-rate is proposed. Simulation results show that the derived asymptotic analyses are closely aligned to the sum-rate obtained by Monte Carlo simulations. In addition, the simulation results prove that the maximum ergodic sum-rate can be achieved by using the proposed number of users.

의학 분야에서 인터넷 활용의 증가로 대용량 의학 데이터를 효율적으로 전송할 수 있는 기술이 요구되고 있다. 광 인터넷에서 모든 OBS 노드들은 하드웨어 컴포넌트인 광 지연 라인들을 가지고 있다. 이것들은 몇 가지 최적 트래픽 조건을 이용해 계산되기 때문에 트랙픽 조건이 변하면 광 지연 라인들도 변해야 한다는 것을 의미한다. 이에 본 논문에 서는 트랙픽 조건이 변하더라도 기존에 설치된 광 지연 라인을 이용하는 서비스 차등화 알고리즘을 제안한다. 트래픽 조건이 변할 때, 새로운 알고리즘은 데이터 버스트의 길이를 결정하기 위해서 사용되는 스레쉬홀드 값을 동적으로 계 산한다. 그러므로 트래픽 조건이 변할지라도 제안된 알고리즘은 광 지연 라인의 대체 없이도 클래스들 사이에서 서비스 차등화를 달성할 수 있다. 본 알고리즘을 이용하면 손실에 민감한 대용량 의학 데이터를 효율적으로 전송할 수 있다.

As increasing the growth of the Internet in medical area, a new technology to transmit effectively massive medical data is required. In optical internet, all OBS nodes have fiber delay lines, hardware components. These components are calculated under some optimal traffic conditions, and this means that if the conditions change, then the components should be altered. Therefore, in this article a new service differentiation algorithm using the previously installed components is proposed, which is used although the conditions vary. When traffic conditions change, the algorithm dynamically recalculates the threshold value used to decide the length of data bursts. By doing so, irrelevant to changes, the algorithm can maintain the service differentiation between classes without replacing any fiber delay lines. With the algorithm, loss sensitive medical data can be transferred well.

데이터센터는 다양한 센서로부터 수많은 종류의 데이터를 수집하고 있으며 이러한 데이터는 데이터센터 모니터링시스템의 분석을 통해 센터의 효율 관리와 개선에 활용되고 있다. 최근에는 데이터센터의 대형화 추세에 따라 센서데이터 양이 폭발적으로 증가하고 모니터링 시스템은 대량의 실시간 스트리밍 데이터에 대한 분석 및 처리 방안이필요하게 되었다. 따라서 본 논문에서는 데이터 발생시간인 이벤트 시간을 기반으로 한 대량의 스트리밍 센서데이터 처리 방안과 데이터양의 증가에 따른 처리 확장성(scalability) 제공 방안을 제시하고자 한다. 이벤트 시간 기반스트리밍 처리를 위해서는 데이터 발생시간과 처리시간 사이의 시간 지연 문제가 해결되어야 하는데 이를 위해 필터링(filtering), 추가시간(slack time), 윈도우 더블링(window doubling), 조정시간(coordinate time) 등을 이용한 처리 방안을 제시한다. 또한 처리 확장성 제공 방안으로 분산 스트리밍 시스템을 활용한 모니터링 시스템 구축 방안을 제시하고 실험적 시스템 구현을 통해 그 효율과 성능을 분석한다.

A data center collects a very large volume of data from various kinds of sensors that is used by the data center monitoring system for the center’s efficiency management and improvement. As the recent trend of large size data center leads to the explosion of sensor data, the monitoring system requires streaming data processing to process and to analyze large volume of real-time data. This paper proposes an event time based real-time streaming processing model for large sensor data. The event time is the time when the data is generated at a sensor. This paper also presents a method to provide scalability of streaming data processing. For the event time based real-time streaming processing, a latency problem between data generation time and processing time should be resolved. We provide a solution based on filtering, slack time, window doubling and coordinate time methods. In order to provide the scalability, this paper builds an experimental monitoring system based on distributed streaming systems and shows an analysis of its performance and efficiency.

본 논문은 대규모 지형을 GPU를 이용한 실시간 가시화 및 고속 편집 방법에 대해 제안한다. 대규모 지형은 64× 64 평면 메시를 GPU 셰이더(Shader)의 테셀레이션 과정에서 LOD(Levels of Detail)를 적용하여 테셀레이션 수행 후 변위 매핑을 통해 지형의 높낮이를 결정하여 생성하였다. 이후 노멀맵으로부터 지형을 구성하는 정점들의 법선 정보를 각 정점별로 계산한 후 조명 효과를 적용하여 가시화하였다. 또한 데이터의 병렬처리에 최적화된 GPGPU(General-Purpose computing on GPU)를 이용하여 지형의 전체적인 노이즈 효과 및 가우시안 블러, 평균 필터를 적용한 부드러운 지형 편집에 대해 수행하였다. 마지막으로 본 논문에서는 대규모 지형의 가시화 성능과 지형 편집 시 CPU와 GPU를 이용해 수행한 성능을 측정하였으며 GPU에서 수행하였을 시 고속의 성능을 보였다.

In this paper, we propose a GPU based technique for massive terrain data visualization and editing while preserving details. The 64× 64 plane mesh tessellate with LOD(Levels of Detail) in GPU Tessellation stages for creation of massive terrain. After tessellation, the terrain height is calculated by displacement mapping. Normal vector of each vertex is calculated from normal map for natural visualization by applying light effect. And we perform noise, Gaussian blur and mean editing at the whole of massive terrain editing using GPGPU (General-Purpose computing on GPU) that is optimized for data parallel processing. The evaluation of performance measure of massive terrain visualization and editing between CPU and GPU shows that the GPU outperforms as compared to CPU.

본 논문에서는 세계 최대 규모의 생의학 분야 서지 데이터베이스인 MEDLINE 전체를 링크드 데이터로 변환·구축하는 효율적인 방안을 제시한다. 이를 위해서 우선 MEDLINE 레코드 구조를 세부적으로 분석하여 적합한 RDF 스키마를 도출하고 각 레코드를 도출된 스키마에 유효한 RDF 파일로 변환하는 과정을 거친다. 본 논문에서는 변환된 레코드 단위의 모든 RDF 파일을 병합하여 이를 단일 RDF 트리플 저장소에 저장할 때 주어 URI 중복 확인 절차를 효율화하는 이중 일괄 등록 방법을 적용한다. 이 방법을 통해서 RDF 파일 단위로 링크드 데이터를 순차적으로 구축하는 방법과 비교했을 때 주어 URI 중복 제거를 위한 RDF 트리플 저장소 접근 횟수가 26,597,850회에서 2,400회로 감소하는 결과를 가져왔다. 따라서 본 연구의 결과는 대용량 서지 레코드 집합을 링크드 데이터로 변환하는 과정에서의 비효율성을 제거하고 신속성과 시의성을 확보할 수 있는 중대한 계기를 제공할 것으로 기대한다.

In this paper, we propose an effective method to convert and construct the MEDLINE, the world's largest biomedical bibliographic database, into linked data. To do this, we first derive the appropriate RDF schema by analyzing the MEDLINE record structure in detail, and convert each record into a valid RDF file in the derived schema. We apply the dual batch registration method to streamline the subject URI duplication checking procedure when merging all RDF files in the converted record unit and storing it in a single RDF triple storage. By applying this method, the number of RDF triple storage accesses for the subject URI duplication is reduced from 26,597,850 to 2,400, compared with the sequential configuration of linked data in units of RDF files. Therefore, it is expected that the result of this study will provide an important opportunity to eliminate the inefficiency in converting large volume bibliographic record sets into linked data, and to secure promptness and timeliness.

In the high-speed backbone network, with the increasing speed of network link, the number of network flows increase rapidly. Meanwhile, with restrictions on hardware computing and storage resources, so, how to identify and measure large flows timely and accurately in massive data become a hot issue in high speed network flow measurement area. In this paper, we propose a new algorithm based on double hash algorithm to realize large flow frequent items identification, according to the defect of MF algorithm which produces false positive easily and frequent updates to bring the huge pressure to the system. The complexity and false positive rate of the algorithm was analyzed. The effect of large flow frequent items statistical accuracy and discard rate for parameter configuration was analyzed through simulation. The theoretical analysis and the simulation result indicate that compare to MF algorithm, our algorithm can identify large flow frequent items more accurately, and satisfies the need of actual measurement.

The scheme for the full-text search has drawn much attention due to its popular use in web document searches and enterprises’ document searches. The full-text search leads to a large size of index files and thus may consume massive computing resources for its processing. In this paper, we present both the system architecture of a full-text search engines with a huge volume of indexed data and its multi-level cache scheme. The presented system architecture and cache scheme were implemented in a commercial search engine, which has capacity enough to process more than 5-milion queries per day and index about 70-milion web documents crawled in Korea. In economic respect, the proposed cache scheme is very crucial for our full-text search engine.

Due to the development of IT, distribution of smart phone, and an increase of use of SNS, various types of contents are being produced and consumed in Internet. Therefore, information searching technology has become important due to a sharp rise in data. However, information searching technology requires much of background knowledge and hence has been recognized as what was difficult to access to. Issues with previous search engine were how many of qualified personnel with background knowledge along with huge amount of development expenses were required. Therefore, search engines have been recognized as what was exclusively possessed by leading IT companies or specialized organizations. This study is intended to suggest a search engine with an index structure for making it convenient to effectively search information by distributed crawling massive amount of websites and web-documents in the distributed environment. Search engine suggested in this study has been realized by Hadoop structure for supporting the distributed processing.

In order to lower the classification cost and improve the performance of the classifier, this paper proposes the approach of the dynamic cost-sensitive ensemble classification based on extreme learning machine for imbalanced massive data streams (DCECIMDS). Firstly, this paper gives the method of concept drifts detection by extracting the attributive characters of imbalanced massive data streams. If the change of attributive characters exceeds threshold value, the concept drift occurs. Secondly, we give Cost-sensitive extreme learning machine algorithm, and the optimal cost function is defined by the dynamic cost matrix. Build the cost-sensitive classifiers model for imbalanced massive data streams under MapReduce, and the data streams are processed in parallel. At last, the weighted cost-sensitive ensemble classifier is constructed, and the dynamic cost-sensitive ensemble classification based on extreme learning machine classification is given. The experiments demonstrate that the proposed ensemble classifier under the MapReduce framework can reduce the average misclassification cost and can make the classification results more reliable. DCECIMDS has high performance by comparing to the other classification algorithms for imbalanced data streams and can effectively deal with the concept drift.

The spatial data set has much useful information, but the amount of volume is massive and the type is complex. It makes hard to analyze the spatial data. There are software tools for general data. Hadoop is one of the tools to process the big data. Hadoop can be used to analyze the large amount of spatial data. This paper proposed a data analysis technique for massive spatial data using Hadoop. We extend the grid based clustering algorithm to use Hadoop. The grid based clustering algorithm makes clusters with cells. Each cell has a number that counts contained objects. Only the cells who had the sufficient population can be join in clusters. The other cells ignored as noise. This paper proposed to enhance performance using Hadoop. In order to evaluate the enhancement of performance, the execution time is measured and compared. As the result, the proposed algorithm is 1.8 times faster than the original grid based clustering algorithm.

In general, people read texts and decide by themselves to measure levels of understanda-bility and readability, which takes a lot of time and efforts. We believe visualizing readability gives intuitive impact on how difficult the texts will be before examining the texts further. Text visualization aims to provide structural characteristics of text contents in an efficient way. By using massive text data, such as books or documents, this study suggests readability meas-urement factors and formulas for the suggested methods that visualize texts by extracting a key factor ‘length’ for readability. In addition to the proposed methods, this study verifies effectiveness of visualization through the test of the case studies. The paper also includes case study findings that readers can have readability information not from independent texts, but from the comparison of previous texts, and therefore it becomes easier to accommodate diffi-cult level of new books.

Log data is a valuable resource for failure prediction and troubleshooting in large-scale systems. However, with the rapid growth of the system scale and the popularity of various applications in productional environments, the volume of logs emerged per day becomes huge, posing serious challenges for storage and analysis. To solve these problems, we propose an online log filtering mechanism to eliminate the redundant and noisy log records through event filtering and instance filtering, aiming to minimize the log size without losing important information required for the fault diagnosis. Our proposed log filtering is evaluated on a real log data derived from a productional cloud computing system, observing that over 76% of the storage space are saved without losing important information.

Top-k similarity join has been used in a wide range of applications that require calculating the most top-k similar pairs of data records in a given database. However, the time performance will be a challenging problem, as an increasing trend of applications that need to process massive data. Obviously, finding the top-k pairs in such vast amounts of data with traditional methods is awkward. In this paper, we propose the RDD-based algorithm to perform the top-k similarity join for massive multidimensional data over a large cluster built with commodity machines using Spark. The RDD- based algorithm consists of four steps, which loads a set of multidimensional records stored in HDFS and finally output an ordered list of top-k closest pairs into HDFS. Firstly, we develop an efficient distance function based on LSH(Locality Sensitive Hashing) to improve the efficiency in pairwise similarity comparison. Secondly, to minimize the amount of data during the RDD running- time, we split conceptually all pairs of LSH signatures into partitions. Moreover, we exploit a serial computation strategy to calculate all top-k closest pairs in parallel. Finally, all the local top-k pairs sorted by their Hamming distances will contribute to the global top-k pairs. In this paper, the performance evaluation between Spark and Hadoop confirms the effectiveness and scalability of our RDD-based algorithm.

In the field of science and technology, a lot of research has recently been done for collecting and analyzing information, for example, examination of global research trend, detection of emerging signals and searching for leading researchers from science and technology literature [1]. Since the science and technology literature information collected for the analysis has been produced for the purpose of each information source, the information is differently constructed and expressed [1, 2]. Therefore, it is necessary to integrate and manage the different information with the same structure and expression format [3]. To this end in this study, examination is made of methods of standardizing the data processing process and data format, and of implementing interoperability of data in different format. Examination is also made of a method of semi-automation through machine learning for even more automated data management. It is expected this study will contribute to improving integration of heterogeneous databases and efficient and easy management of contents in different fields and domains.

The concurrent data transmission from massive machine type communications (MTC) devices within a short time period makes the traffic flow of MTC more bursty. The classic Markovian traffic models under the assumption of Poisson arrival, whose inter- arrival time (IAT) is a negative exponentially distributed random variable with infinite support range, are no longer suitable for this situation. Beta distribution, a distribution with finite support range, is then proposed by 3GPP to account for this and used to serve as the IAT of MTC. The researches on the traffic modeling based on Beta distribution is still in its very initial step and is rarely reported in state of the art literatures. This article takes the initiatives to discuss some prime issues for MTC traffic modeling under the assumption of Beta arrival. By the usage of renewal theory and Volterra integral equation of the second kind with difference kernel, we present the methodology to deduce the access intensity of MTC arrival process, which is defined as the mean number of renewals by time t. Numerical results show that in the context of MTC, there is a sharp increase in the number of access requests from MTC devices which directly causes the well-known congestion problem of MTC. And the ratio of β and α of Beta( α , β ) distribution to some extent can be used as a metric to represent the burstiness of Beta distribution or that of MTC. These works together provide researchers

The real-world data process of large spatio-temporal data collection presents a very difficult technical problem. Firstly, the given process is very expensive, requiring a lot of various high-technology software instruments and modern hardware infrastructure (sensors, servers, GPS infrastructure etc.) installations; secondly, this process sometimes cannot show special traffic patterns, which we may characterize as patterned traffic trajectories. The Arena simulation framework introduced in this paper uses our suggested random linear interpolation algorithm and spatio-temporal prediction algorithm, which are applicable to visualize, handle and predict movement data with various time resolutions.

Random forests are known to be robust for missing and erroneous data as well as irrelevant features. Moreover, even though the forests have many trees, they can utilize the fast building property of decision trees, so they do not require much computing time. In this paper an efficient procedure that utilizes random forests to predict the cylinder bands in rotogravure printing is shown. Even though several research results have been published already to find better prediction accuracy based on other methods, a new and very good result has been found with the suggested method having appropriate parameters of random forests.

Many researchers strive to research on ways on how to improve the performance of RFID system and many papers were written to solve one of the major drawbacks of potent technology related with data management. As RFID system captures billions of data, problems arising from dirty data and large volume of data causes uproar in the RFID community those researchers are finding ways on how to address this issue. Especially, effective data management is important to manage large volume of data. Data reduction techniques in attempts to address the issues on data are also presented in this paper. This paper introduces readers to a new data reduction algorithm that might be an alternative to reduce data in RFID Systems. A process on how to extract data from the reduced database is also presented. Performance study is conducted to analyze the new data reduction algorithm. Our performance analysis shows the utility and feasibility of our categorization reduction algorithms.