• Annual Conference of KIPS
• /
• 2024.10a
• /
• pp.21-24
• /
• 2024

다양한 AI 서비스들의 확산으로 특수 목적형 고성능 시스템에 대한 관심이 증가하면서, PCIe 버스를 인터커넥션 네트워크로 사용하는 사례가 늘고 있다. 본 연구에서는 NTB를 활용한 PCIe 네트워크의 통신 성능을 TCP를 사용하는 방법과 NTB 모듈에 직접 접근하는 방법으로 나누어 측정하고 그 결과와 의미를 분석하였다. 두 가지 방법들이 성능에서 상이한 패턴을 보이기 때문에 PCIe 네트워크를 사용할 시스템의 성능 요구 사항에 따라 그에 부합하는 통신 방법을 선택해야 할 것으로 예상된다.

• Annual Conference on Human and Language Technology
• /
• 1997.10a
• /
• pp.15-22
• /
• 1997

인터넷과 정보 서비스 기술의 발달로 일반 대중에게 제공되는 정보의 양은 기하급수적으로 증가하고 있는 추세지만 사용자가 원하는 정보를 얻기는 더욱 어려워지고 있으며, 필요한 정보를 찾은 경우에도 그 양이 많기 때문에 전체적인 내용을 파악하는 데 많은 시간을 소비하게 된다. 이러한 문제를 해결하고자 본 연구에서는 통계적 모델을 사용하여 문서로부터 문장을 추출한 후 요약문을 작성하여 사용자에게 제시하는 시스템을 개발하였다. 문서 요약 시스템의 구축을 위하여 사용된 방법은 문서 집합으로부터 중요 문장을 추출한 후 이로부터 요약문에 나타날 수 있는 특성(feature)과 중요 단어를 학습하여 학습된 내용을 이용하여 요약문을 하는 방법이다. 시스템 개발 및 평가를 위해 사용된 문서는 정보 과학 분야의 논문 모음이며 이를 학습 데이터와 실험 데이터로 구분한 후 학습 데이터로부터 필요한 정보를 얻고 실험 데이터로 평가하였다.

• Journal of KIISE:Software and Applications
• /
• v.26 no.5
• /
• pp.665-673
• /
• 1999

본 논문에서는 웹 정보검색을 위한 신경망 기반 멀티 에이전트 시스템을 제안한다. 제안된 시스템에서 각 에이전트는 신경망 메카니즘을 이용하여 사용자의 관련도 피드백으로부터 환경을 학습하고 사용자가 원하는 정보를 제공하는 자원을 찾아내어 효율적으로 웹 정보를 검색한다. 먼저 신경망 기반 웹 정보 검색 에이전트를 제시하고 단일 에이전트 기법을 사용할 경우의 문제점을 분석한다. 이를 기반으로 하여 멀티 에이전트 웹 정보 검색 시스템을 정의하고 사용자로부터 정보 검색 지식을 습득하기위한 훈련절차를 기술하며 협동적 정보 검색에 대해 설명한다. 마지막으로 제안된 시스템의 성능을 정형적으로 분석하고 실험을 통하여 기존의 검색 서비스와 비교 평가한다.

• Journal of KIISE:Information Networking
• /
• v.34 no.5
• /
• pp.305-315
• /
• 2007

Broadband Convergence Network (BcN) is the network which unifies telephone network, the Internet and broadcasting networks. Threats to each network can bring serious problems in BcN environment since the whole network can be damaged by various types of attack. The purpose of this study is to suggest the prototype of intrusion-tolerant network design of BcN to guarantee the continuous operation of BcN services against malicious attacks. First, BcN service components, selected by analysis of service time and coverage importance, are classified into three groups by their type: server type, gateway type and hybrid type. Second, the necessity of applying intrusion tolerance on BcN services is deduced by possible attack scenarios on BcN. Finally, we suggest the intrusion-tolerant network design suitable to BcN, using hardware redundancy and secure policies. Also, we present that the suggested network design can increase the intrusion tolerance of BcN.

• Journal of KIISE:Information Networking
• /
• v.29 no.5
• /
• pp.553-561
• /
• 2002

We present a novel service discipline, called linear service discipline, to serve multiple QoS queues sharing a resource and analyze its properties. The linear server makes the output traffic and the queueing dynamics of individual queues as a linear function of its input traffic. In particular, if input traffic is Gaussian, the distributions of queue length and output traffic are also Gaussian with their mean and variance being a function of input mean and input power spectrum (equivalently, autocorrelation function of input). Important QoS measures including buffer overflow probability and queueing delay distribution are also expressed as a function of input mean and input power spectrum. This study explores a new direction for network-wide traffic management based on linear system theories by letting us view the queueing process at each node as a linear filter.

• Journal of KIISE:Information Networking
• /
• v.34 no.6
• /
• pp.525-532
• /
• 2007

Recently, broadband access networks have been competitively deployed all over the world with the development of various multimedia services for such a network. ETRI(Electronics and Telecommunications Research Institute) built the FTTH(Fiber-to-the-Home) service center to develop a next-generation equipment and a high-quality service/platform as well as to promote domestic industry in the field of optical communication. The FTTH service center has a variety of test equipments such as FTTH standard testbed for a BMT(Bench Mark Test), a network operation center, and IPTV service equipments. Also, ETRI deployed a new type of commercial optical access networks cooperating with ISP(Internet service provider). In this paper, we analyzed the quality degradation appeared in QoE(Quality of Experience), when IP data broadcasted from the FTTH service center is transmitted to subscribers through the commercial optical access network. Moreover, we proposed a noble solution to address this problem.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.1
• /
• pp.65-69
• /
• 2010

Moving Picture Expert Group(MPEG) is the organization foundedin 1998 to establish the standards for compressing and expressing the multimedia contents. The organization has established the technological standards such as MPEG-1, MPEG-2, MPEG-4 and MPEG-7. As the 3D video related standards, there is Multiview Profile which is included in the MPEG-2 video of 1996. However, as the MPEG-2 multiview profile is the standard for compressing the videos from two viewpoints on the object, it is not enough to meet the requirement of multiviewvideo technology. In addition, it does not have the technology on the viewpoint switching that it does not provide the services such as the user interaction. This paper proposes the structure in which the specific viewpoint can be described for video switching in addition to the current MPEG-4 system.

• Journal of KIISE:Software and Applications
• /
• v.35 no.12
• /
• pp.716-730
• /
• 2008

In Service-Oriented Architecture (SOA), service providers develop and deploy reusable services on the repositories, and service consumers utilize blackbox form of services through their interfaces. Services are also highly evolvable and often heterogeneous. Due to these characteristics of the service, it is hard to manage the faults if faults occur on the services. Autonomic Computing (AC) is a way of designing systems which can manage themselves without direct human intervention. Applying the key disciplines of AC to service management is appealing since key technical issues for service management can be effectively resolved by AC. In this paper, we present a theoretical model, Symptom-Cause-Actuator (SCA), to enable autonomous service fault management in SOA. We derive SCA model from our rigorous observation on how physicians treat patients. In this paper, we first define a five-phase computing model and meta-model of SCA. And, we define a schema of SCA profile, which contains instances of symptoms, causes, actuators and their dependency values in a machine readable form. Then, we present detailed algorithms for the five phases that are used to manage faults the services. To show the applicability of our approach, we demonstrate the result of our case study for the domain of 'Flight Ticket Management Services'.

• KIISE Transactions on Computing Practices
• /
• v.22 no.8
• /
• pp.387-392
• /
• 2016

Context-aware systems require sensor data collecting model and context representing model to provide user-demand services. Sensor data collecting model consists of sensor access information, sensor value, and definition of value types. Context representing model involves certain keywords to symbolize environmental information including the field from sensor data collecting model that is described in markup language such as XML. However, duplicated keywords could be assigned to different contextual information by service developers. As a result, the system may cause misunderstanding and misleading wrong situational information from unintended contextual information. In this paper, we propose a context classification model for collecting appropriate access information and defining the specification of context.

• KIISE Transactions on Computing Practices
• /
• v.22 no.6
• /
• pp.284-289
• /
• 2016

In the Hadoop-based Big Data analysis model, the data movement between the legacy system and the analysis system is difficult to avoid. To overcome this problem, a unified Big Data file system is introduced so that a unified platform can support the legacy service as well as the analysis service. However, major challenges in avoiding the performance degradation problem due to the interference of two services remain. In order to solve this problem, we first performed a real-life simulation and observed resource utilization, workload characteristics and I/O balanced level. Based on this analysis, two solutions were proposed both for the system level and for the technical level. In the system level, we divide I/O path into the legacy I/O path and the analysis I/O path. In the technical level, we introduce an aggressive prefetch method for analysis service which requires the sequential read. Also, we introduce experimental results that shows the outstanding performance gain comparing the previous system.