• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.60-62
• /
• 2023

인공지능의 학습 작업은 연산량이 많아 고성능 연산 장치인 GPU(Graphics Processing Unit)를 필요로 하며, GPU 장치의 성능은 학습 작업의 실행 성능에 직접적으로 영향을 미치는 요소 중 하나로 작용한다. 인공지능 작업을 처리하기 위해 많이 사용되는 텐서플로의 경우 GPU를 사용해 연산을 수행할 때 기본적으로 거의 모든 GPU 메모리 영역을 단일 학습 작업이 점유하도록 GPU 메모리를 관리한다. 이 방법은 컴퓨팅 자원 중 확장성이 가장 낮은 GPU 메모리의 단편화를 방지하기 위해 사용되는 방법이지만, 하나의 학습 작업이 GPU를 점유하게 되면, 실제 GPU 메모리 사용량과 상관없이 다른 프로세스는 GPU를 사용할 수 없는 문제를 유발한다. 특히, 전이학습, 소규모 학습과 같이 상대적으로 작업 규모가 작은 경우에는 전체 GPU 메모리 용량 중 대부분의 영역이 낭비된다. 본 논문에서는 컨테이너 환경에서 텐서플로의 기본 GPU 메모리 사용 방식으로 인해 다수의 학습 작업을 동시 실행하는 것이 불가능한 문제를 확인하고 GPU 메모리 사용량을 제한한 경우와 하지 않은 경우에 실제 GPU 메모리 사용량과 학습 작업의 실행 시간에 대한 성능 비교를 통해 GPU 메모리의 단편화 방지가 성능에 유의미한 요소인지 검증한다.

• Journal of Korea Port Economic Association
• /
• v.27 no.1
• /
• pp.95-109
• /
• 2011

The aim of this study is to analyze the container terminal operation by considering the key factors that fluctuates the container traffic volume using the System Dynamics (SD) method. The target area of this study is the 'A' container terminal which is located in the Port of Incheon and the simulation period is from 2004 to 2020. As evaluation indexes for container terminal operation, three factors such as 'total sales', 'operating ratio of C/Y' and 'operating ratio of G/C' are selected, and as for the key factors of fluctuating container traffic volume, 'variation ratio of world trade', 'variation ratio of trade among three countries in North-East Asia' and 'variation ratio of won-dollar rate are used. As of 2020, the result of this study is that import-export container traffic volume increases almost 880,000TEU and total sales and operating ratio of G/C each reach 7.1 bilion won and 65 percent. No changes however in loadage and operating ratio of C/Y in 'A' container terminal are indicated. The reason is that capability of C/Y is exceeded. Therefore this study suggest that decision-makers of 'A' container terminal realize the importance of additional space of C/Y.

• Journal of Korea Port Economic Association
• /
• v.22 no.4
• /
• pp.81-98
• /
• 2006

Many studies on port tariff have been done over twenty years using publicly assessed data on tariff. Public data for tariff rates do not reflect, however, the port tariff in a real market, since the cargo handling charge, which is the important fraction of port tariff, is confidentially decided by the negotiations between a shipping company and a container terminal operator. In this paper, we collected the real price data of the port tariff on the world major sixteen container ports from a global shipping company and transformed it into the tariff per TEU(US$/TEU). The comparative analysis of port tariff was performed using the port tariff per TEU, and a panel regression analysis was done to identify the relations between the port tariff and demand variables: throughput, GDP and trade amount.

• Journal of Korea Port Economic Association
• /
• v.27 no.2
• /
• pp.195-215
• /
• 2011

The purpose of this article is to provide data to be utilized for marketing to attract future container cargo to the Korea-China Car Ferry Service and to estimate container traffic for the development of the International Passenger Terminal to be newly built in the Incheon Southern Port. An analysis of container traffic through the Passenger Terminal of Incheon Port showed that the container throughput of Incheon Port comprised 20% of traffic. The export cargo traffic in the International Passenger Terminal, 71.4%, or 111,698 TEU of 156,284 TEU in total, was brought from metropolitan areas. To increase the container traffic of the car ferry service currently concentrated on the metropolitan areas, more endeavor is required to strengthen marketing competence to attract container traffic to other regions apart from the metropolitan areas as well as provide general services.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.23 no.3
• /
• pp.283-288
• /
• 2010

Mobile harbor is characterized by the lightweight compact structure when compared to the conventional above-ground port container crane. A new concept RORI crane system, which was devised for mobile harbor to satisfy the compactness and light weightness, not only can load/unload containers with high speed on sea but can be completely folded at maneuvering mode. This study is concerned with the tip deflection of the horizontal boom of mobile harbor at container loading operation. Both the theoretical method utilizing the Castigliano's theorem and the numerical approach by finite element method are employed, and the reliability of the latter approach is verified through the comparison with the theoretical results. And then, the effect of the initial cable tension on the tip deflection is parametrically examined by the finite element analysis.

• Journal of Navigation and Port Research
• /
• v.45 no.3
• /
• pp.102-108
• /
• 2021

The purpose of this study is to develop a proactive methodology for disruption due to uncertainty in vessels' arrival time. As worldwide imports and exports increased rapidly, the importance of berth planning in container terminals has increased accordingly. Since the berth plan determines the capacity of the container terminal, it aims to maximize efficiency by minimizing the time and space gap between the vessels. In reality, several uncertainties disrupt the initial berth plan resulting in economic losses. In this study, we propose a robust berth plan for preventing disruption.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.11 no.10
• /
• pp.1999-2007
• /
• 2007

This article describes advanced clover algorithm for effective loading and unloading of containers using stackers position data in a yard. The job scheduling must rely on job assign of stackers and position data processing to dynamically allocate stackers, and maintain multiple job processing, all based on task requirements. A stacker tracking using GPS and GIS is an essential capability and is used as yard loading and unloading process improvement for yard management. After estimating position of stackers in a yard the raper describes advanced clover algorithm and other techniques for monitoring loading and unloading of individual containers as well as combinatorial stacker load balancing problems such as estimating load of each stackers. Results from simulations and experimental implementations have demonstrated that the suggested approaches are efficient in stacker management.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2018.10a
• /
• pp.581-584
• /
• 2018

최근 다양한 분야에서 대량의 데이터를 수집하여 처리하고 분석하는 빅데이터 기술이 활용되고 있다. 빅데이터 분석을 위해서는 데이터 시각화 기술이 필수적이다. 본 논문에서는 REST API를 사용하여 시계열 데이터베이스에 데이터를 질의하고, 응답받은 시계열 데이터를 다양한 형태의 차트로 시각화하는 마이크로서비스(Graphgen)를 설계하고 구현한다. 이 서비스는 데이터의 변동에 따라 실시간으로 시각화 객체를 갱신하며, 대용량 데이터 처리의 성능저하를 최소화하도록 개발된다. Graphgen은 InfluxDB와 OpenTSDB 시계열 데이터베이스와 Bokeh 시각화 라이브러리를 지원하며, 추후 서비스 확장이 용이하도록 개발된다. 또한 부하 분산과 통합 배포 관리를 위하여 컨테이너를 기반으로 개발된다.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.30 no.4D
• /
• pp.323-330
• /
• 2010

The time in port for vessels is one of the important factors for analyzing the operation status and the capacity of ports. In addition, the time in port for vessels can be directly used for estimating the greenhouse gas emissions resulted from vessels in port. However, it is unclear which variables can affect the time in port for vessels and what the marginal effect of each variable is. With these challenges in mind, the study analyzes the time in port for vessels arriving and departing port of Busan by using a parametric survival model. The results show that the log-logistic accelerated failure time model is appropriate to explain the time in port for 19,167 vessels arriving and departing port of Busan in 2008, in which the time in port is significantly affected by gross tonnage of vessels, service capacity of terminal, and vessel type. This study also shows that the greenhouse gas emission resulted from full-container vessels, which accounted for about 61% of all vessels with loading/unloading purpose arriving and departing port of Busan in 2008, is about "17 ton/vessel" in the boundary of port of Busan. However, the hotelling greenhouse gas emissions resulted from non-container vessels (3,774 vessels; 20%) are greater than those from the full-container vessels. Hence, it is necessary to take into account more efficient port management polices and technologies to reduce the service time of non-container vessels in port of Busan.

• Journal of Navigation and Port Research
• /
• v.41 no.3
• /
• pp.137-148
• /
• 2017

Port service level is a metric of competitiveness among ports for the operating/managing bodies such as the terminal operation company (TOC), Port Authority, or the government, and is used as an important indicator for shipping companies and freight haulers when selecting a port. Considering the importance of metrics, we developed software to objectively define and manage six important service indicators exclusive to container and bulk terminals including: berth occupancy rate, ship's waiting ratio, berth throughput, number of berths, average number of vessels waiting, and average waiting time. We computed the six service indicators utilizing berth 1 through berth 5 in the container terminals and berth 1 through berth 4 in the bulk terminals. The software model allows easy computation of expected ship's waiting ratio over berth occupancy rate, berth throughput, counts of berth, average number of vessels waiting and average waiting time. Further, the software allows prediction of yearly throughput by utilizing a ship's waiting ratio and other productivity indicators and making calculations based on arrival patterns of ship traffic. As a result, a TOC is able to make strategic decisions on the trade-offs in the optimal operating level of the facility with better predictors of the service factors (ship's waiting ratio) and productivity factors (yearly throughput). Successful implementation of the software would attract more shipping companies and shippers and maximize TOC profits.