• Journal of Advanced Navigation Technology
• /
• v.18 no.4
• /
• pp.401-408
• /
• 2014

Hadoop is composed with MapReduce programming model for distributed processing and HDFS distributed file system. Hadoop is suitable framework for big data processing, but processing of mass small files have many problems. The processing of mass small file in hadoop have problems to created one mapper per one file, and it have problems to needed many memory for store of meta information of file. This paper have comparison evaluation processing method of mass small file with various method in hadoop platform. The processing of general compression format is inadequate because of processing by one mapper regardless of data size. The processing of sequence and hadoop archive file is removed memory problem of namenode by compress and combine of small file. Hadoop archive file is faster then sequence file about combine time of small file. The processing using CombineFileInputFormat class is needed not combine of small file, and it have similar speed big data processing method.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2014.05a
• /
• pp.479-482
• /
• 2014

Nowadays open-source hadoop systems have been using widely to efficiently manage a fast-growing big data. Hadoop systems consist of distributed file processing system called HDFS (Hadoop Distributed File System) and distributed parallel processing system called MapReduce. The MapReduce reads and processes big data from HDFS and then processed results are written in HDFS again by the MapReduce. Such a processing method has different system structure respectively according to hadoop version. Therefore, this paper shows analysis results for performance of hadoop systems. For this, we devise a way which monitors hadoop systems and measure occurrence frequency of processes, threads, and variables generated in hadoop system itself using the devised way. So, by using the measured results as analysis indicator, we help the indicator predict inner performance of hadoop systems.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.12 no.2
• /
• pp.367-374
• /
• 2017

This paper discusses management problems of Hadoop distributed node, which is a platform for big data processing, and proposes a novel technique for enabling flexible node management of Hadoop Distributed File System. Hadoop cannot configure Hadoop cluster dynamically because it judges temporarily unavailable nodes as a failure. Delayed block replication scheme proposed in this paper delays the removal of unavailable node as much as possible so as to be easily rejoined. Experimental results show that the proposed scheme increases flexibility of node management with little impact on distributed processing performance when the cluster size changes.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.23 no.5
• /
• pp.803-813
• /
• 2013

As the developments of smart devices and social network services, the amount of data has been exploding. The world is facing Big data era. For these reasons, the Big data processing technology which is a new technology that can handle such data has attracted much attention. One of the most representative technologies is Hadoop. Hadoop Distributed File System(HDFS) designed to run on commercial Linux server is an open source framework and can store many terabytes of data. The initial version of Hadoop did not consider security because it only focused on efficient Big data processing. As the number of users rapidly increases, a lot of sensitive data including personal information were stored on HDFS. So Hadoop announced a new version that introduces Kerberos and token system in 2009. However, this system is vulnerable to the replay attack, impersonation attack and other attacks. In this paper, we analyze these vulnerabilities of HDFS security and propose a new protocol which complements these vulnerabilities and maintains the performance of Hadoop.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2013.11a
• /
• pp.15-17
• /
• 2013

HDFS(Hadoop Distribute File System)는 대용량 파일 처리를 목적으로 설계 되었으며 현재 이상적인 분산 파일 시스템으로 각광 받고 있다. 이러한 HDFS는 기존 분산파일 시스템과 많은 유사성을 가지고 있으나, Fault Tolerance를 제공하고, 데이터 엑세스 패턴을 스트리밍 방식으로 지원하여 대용량 파일을 효율적으로 저장할 수 있다는 차별성을 가지고 있다. 하지만 실제 HDFS 데이터 집합에는 Small file이 차지하는 비중이 상당히 높으며, 이러한 다수의 Small file 은 데이터 처리에 있어 높은 비용을 초래할 뿐 아니라 Master Node 의 파일 처리 및 메모리 성능에 악영향을 미친다. 따라서 본 논문에서는 HDFS에서 Small file 이 미치는 영향을 분석하고 이러한 문제점을 해결 할 수 있는 로컬 인덱스 파일기반의 파일 병합 기법을 제안한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2013.05a
• /
• pp.476-477
• /
• 2013

Gadget-2 is a scientific simulation code has been used for many different types of simulations like, Colliding Galaxies, Cluster Formation and the popular Millennium Simulation. The code is parallelized with Message Passing Interface (MPI) and is written in C language. There is also a Java adaptation of the original code written using MPJ Express called Java Gadget. Java Gadget writes a lot of checkpoint data which may or may not use the HDF-5 file format. Since, HDF-5 is MPI-IO compliant, we can use our MPJ-IO library to perform parallel reading and writing of the checkpoint files and improve I/O performance. Additionally, to add reliability to the code execution, we propose the usage of Hadoop Distributed File System (HDFS) for writing the intermediate (checkpoint files) and final data (output files). The current code writes and reads the input, output and checkpoint files sequentially which can easily become bottleneck for large scale simulations. In this paper, we propose Sim-Hadoop, a framework to leverage HDFS and MPJ-IO for improving the I/O performance of Java Gadget code.

• Journal of the Korea Society of Computer and Information
• /
• v.20 no.8
• /
• pp.77-83
• /
• 2015

The storage and retrieval of multimedia data is becoming increasingly important in many application areas including record management, video(CCTV) management and Internet of Things (IoT). In these applications, the files containing multimedia that need to be stored and managed is tremendous and constantly scaling. In this paper, we propose a technique to retrieve a very large number of files, in multimedia format, using the Hadoop Framework. Our strategy is based on the management of metadata that describes the characteristic of files that are stored in Hadoop Distributed File System (HDFS). The metadata schema is represented in Hbase and looked up using SQL On Hadoop (Hive, Tajo). Both the Hbase, Hive and Tajo are part of the Hadoop Ecosystem. Preliminary experiment on multimedia data files stored in HDFS shows the viability of the proposed strategy.

• Journal of Korea Multimedia Society
• /
• v.18 no.9
• /
• pp.1098-1104
• /
• 2015

With the widespread use of smartphones and IoT (Internet of Things), data are being generated on a large scale, and there is increased for the analysis of such data. Hence, distributed processing systems have gained much attention. Hadoop, which is a distributed processing system, saves the metadata of stored files in name nodes; in this case, the main problems are as follows: the memory becomes insufficient; load occurs because of massive small files; scheduling and file processing time increases because of the increased number of small files. In this paper, we propose a solution to address the increase in processing time because of massive small files, and thus improve the processing performance, using the Reuse JVM method provided by Hadoop. Through environment setting, the Reuse JVM method modifies the JVM produced conventionally for every task, so that multiple tasks are reused sequentially in one JVM. As a final outcome, the Reuse JVM method showed the best processing performance when used together with CombineFileInputFormat.

• Journal of the Korea Society of Computer and Information
• /
• v.15 no.5
• /
• pp.49-56
• /
• 2010

In this paper, we implement a Hadoop based cluster for cloud computing and evaluate the performance of this cluster based on application characteristics by executing RandomTextWriter, WordCount, and PI applications. A RandomTextWriter creates given amount of random words and stores them in the HDFS(Hadoop Distributed File System). A WordCount reads an input file and determines the frequency of a given word per block unit. PI application induces PI value using the Monte Carlo law. During simulation, we investigate the effect of data block size and the number of replications on the execution time of applications. Through simulation, we have confirmed that the execution time of RandomTextWriter was proportional to the number of replications. However, the execution time of WordCount and PI were not affected by the number of replications. Moreover, the execution time of WordCount was optimum when the block size was 64~256MB. Therefore, these results show that the performance of cloud computing system can be enhanced by using a scheduling scheme that considers application's characteristics.

• Journal of Internet Computing and Services
• /
• v.13 no.5
• /
• pp.55-64
• /
• 2012

Nowadays, as Cloud computing becomes popular, a need for a DFS(distributed file system) is increased. But, in the current Cloud computing environments, there is no DFS framework that is sufficient to protect sensitive private information from attackers. Therefore, we designed and proposed a secure scheme for distributed file systems. The scheme provides confidentiality and availability for a distributed file system using a secret sharing method. In this paper, we measured the speed of encryption and decryption for our proposed method, and compared them with that of SEED algorithm which is the most popular algorithm in this field. This comparison showed the computational efficiency of our method. Moreover, the proposed secure read/write model is independent of Hadoop DFS structure so that our modified algorithm can be easily adapted for use in the HDFS. Finally, the proposed model is evaluated theoretically using performance measurement method for distributed secret sharing model.