• 한국전산유체공학회:학술대회논문집
• /
• 2001.10a
• /
• pp.182-189
• /
• 2001

A project has been initiated at KISTI Supercomputing Center at year 1999 to build a Tflops cluster in order to support newly emerging demands for larger scale parallel computing resources. As a supportive substitute of traditional supercompters, the cluster systems are expected to play important role in supporting diversified computational needs. In this paper, KISTI Supercomputing Center's working status of building Tflops cluster system and future plans.

• The Journal of Korean Association of Computer Education
• /
• v.11 no.3
• /
• pp.81-89
• /
• 2008

The failure occurrence of resources in the grid computing is higher than in a tradition parallel computing. Since the failure of resources affects job execution fatally, fault tolerance service is essential in computational grids. And grid services are often expected to meet some minimum levels of quality of service (QoS) for desirable operation. However Globus toolkit does not provide fault tolerance service that supports fault detection service and management service and satisfies QoS requirement. Thus this paper proposes fault tolerance service to satisfy QoS requirement in computational grids. In order to provide fault tolerance service and satisfy QoS requirements, we expand the definition of failure, such as process failure, processor failure, and network failure. And we propose resource scheduling service, fault detection service and fault management service and show implement and experiment results.

• Proceedings of the Korean Information Science Society Conference
• /
• 2007.06c
• /
• pp.336-340
• /
• 2007

We envisage that grid computing environments allow us to implement distributed data mining services, that is, those applications which analyze large sets of geographically distributed databases and information using the computational power and resources of a grid environment. This paper describes an experimental framework towards such a distributed data mining approach, including design considerations and a prototype implementation. Based on the "Knowledge Grid" architecture suggested by Cannataro et al., we identify four major components - user node, broker node, data node, and computation node - and define their individual roles. For implementing the prototype, we have investigated methods for utilizing distributed resources within a grid computing environment, e.g., communication and coordination among the various resources available.

• Journal of Information Processing Systems
• /
• v.5 no.2
• /
• pp.87-96
• /
• 2009

Grid computing is a new technology which involves efforts to create a huge source of processing power by connecting computational resources throughout the world. The key issue of such environments is their resource allocation and the appropriate job scheduling strategy. Several approaches to scheduling in these environments have been proposed to date. Market driven scheduling as a decentralized solution for such complicated environments has introduced new challenges. In this paper the bidding problem with regard to resources in the reverse auction resource allocation model has been investigated and the new bidding strategies have been proposed and investigated.

• Annual Conference of KIPS
• /
• 2002.04b
• /
• pp.1333-1336
• /
• 2002

웹 서비스(Web Service)와 그리드 컴퓨팅(Grid Computing)은 서로 다른 목적을 위해 개발되고 있다. 하지만 두 개의 기술은 "동적이며 여러 기관에서 참여하는 가상 조직을 통한 문제 해결과 자원 공유" 와 같은 동일한 문제를 다루고 있다. Computational 웹 서비스(CWS: Computational Web Service)는 그리드 기반 기술로 웹 서비스를 사용하는 새로운 아이디어이다. 현재의 CWS 는 워커와 CWS 사이의 중개 역할을 하는 풀 서비스(Pool Service)를 각자 고유하게 하나씩 소유하고 있다. 그러나 풀 서비스가 특정 CWS 에 소속되어 있다면 자원 사용 측면에서 비효율적이다. 본 논문에서는 CWS 에 독립적이며 일반적인 풀 웹 서비스(PWS: Pool Web Service)와 이를 위한 스케줄링 방법을 제안한다.

• Journal of KIISE:Information Networking
• /
• v.32 no.4
• /
• pp.471-482
• /
• 2005

The high-speed network permits Grid computing to handle large problem of management areas and share various computational resources. As there are many resources and changes of them in Grid computing, the resources should be detected effectively and matched correctly with tasks to provide high performance. In this paper, we propose a mechanism that maximizes the performance of Grid computing systems. According to a priority, grade and site of heterogeneous resources, we assign tasks to those resources. Initially, a volunteer's priority and ranking are determined by static information like as CPU speed, RAM size, storage size and network bandwidth. And then, the rank of resources is decided by considering dynamic information such as correctness, response time, and error rate. We find that overall Grid system performance is improved and high correctness using resource reallocation mechanism is achieved.

• 한국전산유체공학회:학술대회논문집
• /
• 1998.05a
• /
• pp.82-89
• /
• 1998

There are many moving-body problems to be solved, the solution of which necessary for proper design of flight vehicles in aerospace industry. Since a body moves relative to other bodies in the category of these problems, difficulty arises regarding both generation of computational grid around the body in motion and conservation of flow properties in the moving grid system. A few example could be store separation from the aircraft and relative vibration of multiple bodies in the high-speed flow passage. In this paper we report on the progress made in computing moving-body aerodynamics related with sabot separation characteristics. Conservative overlapping grid together with cell-merging-unmerging technique is used to solve the Euler equations for a body in high-speed motion. Carbuncle errors has to be removed before we obtain physically adequate solution. Two-dimensional application is reported here.

• 유체기계공업학회:학술대회논문집
• /
• 2006.08a
• /
• pp.33-36
• /
• 2006

As the computing environment is rapidly improved, the interests of CFD are gradually focused on large-scale computation over complex geometry. Keeping pace with the trend, essential computational tools to obtain solutions of complex aerospace flow analysis and design problems are examined. An accurate and efficient flow analysis and design codes for large-scale aerospace problem are presented in this work. With regard to original numerical schemes for flow analysis, high-fidelity flux schemes such as RoeM, AUSMPW+ and higher order interpolation schemes such as MLP (Multi-dimensional Limiting Process) are presented. Concerning the grid representation method, a general-purpose basis code which can handle multi-block system and overset grid system simultaneously is constructed. In respect to design optimization, the importance of turbulent sensitivity is investigated. And design tools to predict highly turbulent flows and its sensitivity accurately by fully differentiating turbulent transport equations are presented. Especially, a new sensitivity analysis treatment and geometric representation method to resolve the basic flow characteristics are presented. Exploiting these tools, the capability of the proposed approach to handle complex aerospace simulation and design problems is tested by computing several flow analysis and design problems.

• Proceedings of the Korea Contents Association Conference
• /
• 2007.11a
• /
• pp.221-225
• /
• 2007

PC grid computing is a paradigm of distributed computing that are collection the idle resource of numerous PCs to perform large-scale. As this way, created high performance resources use a large-scale computational application, also this paradigm studied new measurement in place of application using super computer. This paper suggests ORRS(Open Resource Registration System) that selects a adequate resource what application client want. This system register descriptions of resource that provide group, such as organization, party, team, PCs idle resource in open archive. Also, this system provide PC grid computing environment which is select suitable resource actively what application client want.

• Journal of the Korea Society for Simulation
• /
• v.15 no.4
• /
• pp.129-136
• /
• 2006

Grid computing has been appeared for solving large-scaled data which are not solved by a single computer. Grid computing is a new generation platform which connects geographically distributed heterogeneous resources. However, gathering heterogeneous distributed resources produces many difficult problems. Especially. to assure resource reliability is one of the most critical problems. So, we propose a grid resource scheduling model using grid resource reliability measurement. We evaluate resource reliability based on resource status data and apply it to the grid scheduling model in DEVSJAVA modeling and simulation. This paper evaluates parameters such as resource utilization, job loss and average turn-around time and estimates experiment results of our model in comparison with those of existing scheduling models such as a random scheduling model and a round-robin scheduling model. These experiment results showed that the resource reliability measurement scheduling model provides efficient resource allocation and stable Job processing in comparison with a random scheduling model and a round-robin scheduling model.