• Proceedings of the Korean Information Science Society Conference
• /
• 2003.04d
• /
• pp.148-150
• /
• 2003

지리적으로 분산된 유휴 자원들을 통합하여, 하나의 메타 컴퓨팅 환경을 제공하기 위한 연구가 활발히 진행되고 있다. 대표적인 것으로 그리드를 들 수 있는데, 통신망에 연결된 슈퍼컴퓨터. 클러스터 워크스테이션, 대용량 저장장치 및 측정장치 등과 같은 자원을 공유하는 더 강력한 가상의 컴퓨팅 환경이다. 그리드의 세부분야 중 어카운팅 분야는 다른 분야와는 달리 아직 그 연구가 미미한 상태이고 그리드의 특성상 기존의 플랫폼에서 취해왔던 방식이 그대로 적용되기 어려운 점이 있다. 본 논문에서는 그리드 환경에 적합하도록 분산 어카운팅을 제공하고 각 사이트의 자율성을 보장할 수 있는 어카운팅 플랫폼을 설계하였다.

• The KIPS Transactions:PartA
• /
• v.12A no.2 s.92
• /
• pp.161-170
• /
• 2005

Grid makes a virtual high-performance computing resource by connecting geographically distributed heterogeneous resources. Building access control system is an important factor In the grid environment In this paper, we design and implement a grid access control system based on Globus Toolkit, which is one of the popular grid middleware. Especially, to guarantee the site autonomy for resource provider, we use several environment configuration files. Moreover, we design and implement PGAM to produce more detail and diverse information to ease the development of value added services.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2015.10a
• /
• pp.250-252
• /
• 2015

다양한 분야에서 활발히 연구되는 빅 데이터와 최근 이슈가 되고 있는 딥러닝(Deep-learning) 등은 컴퓨터공학 분야뿐만 아니라 다양한 분야와 접목하여 이에 대한 관심이 증가하고 있다. 대규모 클러스터를 통하여 빅데이터와 딥러닝 같은 계산 집약적인(computational-intensive) 작업을 빠르게 처리할 수 있다. 하지만 대규모 클러스터의 잦은 유휴상태는 클러스터의 활용률은 매우 낮아지게 한다. 본 논문에서는 작업 실행 시간 개선과 클러스터 활용 효율성을 향상시키는 HTCondor 클러스터를 위한 동적 자원 스케줄링 기법을 제안한다. 동적으로 자원 할당을 위해 가상머신으로 HTCondor 클러스터 환경을 구성하였으며, 가상머신의 관리를 위해 OpenStack을 사용하였다. OpenStack기반 HTCondor 클러스터 환경에서 HTCondor Python API와 OpenStack Python API를 사용하여 우리가 제안하는 동적 자원 스케줄링 기법을 구현하였으며, 실험을 통해 제안하는 기법의 성능 및 실현 가능성을 확인하였다.

• Journal of KIISE:Computing Practices and Letters
• /
• v.5 no.5
• /
• pp.583-594
• /
• 1999

With the advance of VLSI and network technologies, it has now become a common practice to deploy a various forms of distributed computing environments. A study shows that a lot of network-aware computers are in an idle state for considerable amount of time depending on the types of users and time frames of the day. If we can take the full advantage of those idle computers, we can obtain the enormous combined processing power without further costly investment. In this paper, we present a distributed batch-processing system, called the Java Distributed Batch-processing System (JDBS), which allows us to execute CPU-intensive, independent jobs across a pool of idle workstations on top of extant distributed computing environments. Since JDBS is implemented using a Java programming language, it not only extends the scope of machine types that can be joined to the pool, but makes it a lot easier to build an entire system. Besides, JDBS is scalable and fault-tolerant due to its multi-cluster organization and intelligent strategies. A graphical user interface is also provided to facilitate the registration and unregistration, job submission, and job monitoring.

• The KIPS Transactions:PartC
• /
• v.11C no.6 s.95
• /
• pp.807-814
• /
• 2004

This paper researches and evaluates a bandwidth reallocation mechanism for efficient DiffServ QoS support in MPLS networks by monitoring the network traffic status and reallocating unused bandwidth. While the Differentiated Services in MPLS Networks architecture provides QoS management through the RSVP resource reservation, this mechanism is based on a static provisioning of resource. But this approach can lead to waste bandwidth in some service classes or, leave some service classes' resource starved. This paper presents the bandwidth reallocation dynamically based on network traffic status for bandwidth usage maximization.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.14 no.5
• /
• pp.1146-1154
• /
• 2010

A virtual machine lease service in the cloud computing environment has been implemented using the open source cloud computing platform, OpenNebula. In addition, a web-based cloud user interface is developed for both convenient resource management and efficient service access. The present virtual machine allocation scheme adopted in OpenNebula has performance reduction problem because of not considering CPU allocation scheduler of the virtualization software. In order to address this problem we have considered both the priority of the idle CPU resources of the cluster and credit scheduler of Xen, which resulted in performance improvement of the OpenNebula virtual machine scheduler. The experimental results showed that the proposed allocation scheme provided more virtual machine creations and more CPU resource allocations for cloud service.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.30 no.12A
• /
• pp.1111-1119
• /
• 2005

A single fiber cut can potentially influence a total of 1.6 Tbps of traffic and is surprisingly frequent and results in significant loss of traffic. The survivability as we as the resource utilization of the connection in optical networks is important. Existing protection mechanisms to rapidly recover the failures allocate the backup path just SRLG-disjointed with working path. Those did not consider the unbalanced resource share of backup paths. The backup paths can be centralized on the specific link so that the idle resource is not used. As a result, as those do not efficiently use the resource the whole resource utilization is not so good. So we propose the mechanism to enhance the resources utilization as settling down the unbalanced resource share. Our purpose is to improve the link utilization as distributing the maximum link load.

• The Journal of the Korea Contents Association
• /
• v.15 no.7
• /
• pp.1-11
• /
• 2015

In recent high performance computing system, GPGPU has been widely used to process general-purpose applications as well as graphics applications, since GPU can provide optimized computational resources for massive parallel processing. Unfortunately, GPGPU doesn't exploit computational resources on GPU in executing general-purpose applications fully, because the applications cannot be optimized to GPU architecture. Therefore, we provide GPU research guideline to improve the performance of computing systems using GPGPU. To accomplish this, we analyze the negative factors on GPU performance. In this paper, in order to clearly classify the cause of the negative factors on GPU performance, GPU core status are defined into 5 status: fully active status, partial active status, idle status, memory stall status and GPU core stall status. All status except fully active status cause performance degradation. We evaluate the ratio of each GPU core status depending on the characteristics of benchmarks to find specific reasons which degrade the performance of GPU. According to our simulation results, partial active status, idle status, memory stall status and GPU core stall status are induced by computational resource underutilization problem, low parallelism, high memory requests, and structural hazard, respectively.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2013.10a
• /
• pp.455-458
• /
• 2013

The conventional design of the information system architecture based on the personal experience of information systems has been acted as a limit in progress utilizing appropriate resource allocation and performance improvements. Architecture design depending on personal experience makes differences in variance of a designer's experience, intellectual level in related tasks and surroundings, and architecture quality according to individual's propensity. After all these problems cause a waste of expensive hardware resources. At working place, post-monitoring tools are diversely developed and are running to find the bottleneck and the process problems in the information operation. However, there are no simulation tools or models that are used for expecting and counteracting the problems at early period of designing architecture. To solve these problems we will first develop a simulation model for designing information system architecture in a pilot form, and will verify validity. If an error rate is found in the permissible range, then it can be said that the simulation reflects the characteristic of information system architecture. After the model is developed in a level that can be used in various ways, more accurate performance computation will be able to do, getting out of the old way relying on calculations, and prevent the existence of idle resources and expense waste that comes from the wrong design of architecture.

• KIISE Transactions on Computing Practices
• /
• v.23 no.9
• /
• pp.572-577
• /
• 2017

Virtualization allows multiple virtual machines to share the resources of a physical machine in order to utilize idle resources. The purpose of virtualization is the efficient allocation of resources among virtual machines. However, the efficient allocation of resources is difficult because the workload characteristics of each virtual machine cannot be understood in the current virtualization environment. This causes performance interference among virtual machines, which leads to performance degradation of the virtual machine. Previous works have been carried out to develop a method of solving such performance interference. This paper introduces a representative method, a CPU scheduling method that guarantees I/O performance by using feedback control to solve performance interference. In addition, we compare and analyze a model-based feedback control method and a dynamic feedback control method.