• Proceedings of the Korean Information Science Society Conference
• /
• 2002.10c
• /
• pp.355-357
• /
• 2002

최근 인터넷이 발달하면서 인터넷 상의 다양한 컴퓨터들을 연결함으로써 이기종 클러스터 환경 구축이 용이해졌다. 이러한 이기종 클러스터 환경에서 알고리즘의 이식성을 높이기 위해서는 네트워크의 특성 및 노드의 이질성에 따른 부하 불균형에 효과적으로 적응할 수 있어야 한다 본 논문에서는 이기종 클러스터 환경에서 Message Passing 방식을 이용한 고성능 클러스터 컴퓨팅 작업 시 최적의 효율을 얻을 수 있는 Enhanced-WF 알고리즘을 제시한다 Enhanced-WF 알고리즘은 부하공유를 위하여 Weighted Factoring 알고리즘을 기반으로 적응할당정책을 적용하는 동시에 네트워크 통신시간과 계산시간을 겹치게 한다. Enhanced-WF 알고리즘의 성능을 측정하기 위해 이기종 PC클러스터 환경에서 PVM을 이용한 행렬곱셈 프로그램을 이용하였다. 그 결과, Enhanced-WF 알고리즘이 이기종 클러스터 환경에서 Send, GSS, Weighted Factoring 알고리즘과 같은 기존의 부하공유 알고리즘보다 효과적임을 보였다.

• Proceedings of the IEEK Conference
• /
• 2005.11a
• /
• pp.1091-1094
• /
• 2005

저가의 개인용 컴퓨터를 고속 네트워크 장비와 시스템 독립적인 통신 라이브러리를 이용하여 연결함으로서 병렬처리 프로그램을 수행할 수 있는 클러스터 시스템을 구축할 수 있다. 클러스터 시스템은 이를 쉽게 구성하는 개별 시스템의 교체 및 추가로 인해서 이기종화 된다. 본 논문에서는 이기종 시스템으로 구성된 클러스터 시스템의 성능 향상을 위해서 MPI Work Packet Manager (WPM) 라이브러리를 제안한다. 실험 결과에 의하면 이기종 시스템으로 구성된 클러스터 시스템에서 WPM 라이브러리를 적용한 병렬처리 프로그램이 이기종 클러스터 시스템의 성능을 적절히 활용하고 있음을 알 수 있다.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.46 no.6
• /
• pp.56-62
• /
• 2009

Inexpensive high performance computer systems combined with high speed networks and machine independent communication libraries have made cluster computing a viable option for parallel applications. In a heterogeneous cluster environment, efficient resource management is critically important since the computing power of the individual computer system is a significant performance factor when executing applications in parallel. This paper presents a dynamic task manager, called DWM (dynamic work manager). It makes a heterogeneous cluster system fully utilize the different computing power of its individual computer system. We measure the performance of DWM in a heterogeneous cluster environment with several kernel-level benchmark programs and their programming complexity quantitatively. From the experiments, we found that DWM provides competitive performance with a notable reduction in programming effort.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.44 no.4 s.316
• /
• pp.21-27
• /
• 2007

Recently cluster system built from personal computers and network devices are easily and economically constructed. Rapid technological change discloses new processors on the market yielding cluster systems heterogeneity. A parallel system in heterogeneous environment needs work managers for utilizing the full power of the heterogeneous cluster system. In this paper, we compare MPI and Cilk in a heterogeneous cluster system in terms of performance and code complexity. Experimental results show that Cilk is better than MPI with small sizes of data transfers while MPI outperforms Cilk with big sizes of data transfers. Also, We find that Cilk requires less programming efforts to write a parallel program.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2000.10b
• /
• pp.981-984
• /
• 2000

지난 10년간 인터넷은 정보화 시대의 새로운 패러다임으로 사회 전반을 변화시켜 왔다. 네트워크 요청이 크게 증가하고 대량의 정보처리가 요구되면서 이 기간 동안 고성능 컴퓨팅 환경에 대한 연구가 활발하게 진행되었다. 낮은 가격과 높은 가용성을 유지하면서 고성능 컴퓨팅 환경을 제공하려는 노력은 클러스터링 기술을 발전시켰지만, 기존 클러스터 시스템들은 운영체제 및 하드웨어에 종속적이기 때문에 높은 확장성을 제공하지 못하였다. 또한 여러 대의 시스템이 연결되는 클러스터 시스템은 각각의 운영체제를 가지고 있지만, 시스템 전체 자원을 효율적으로 이용하기 위해서는 하나의 시스템 이미지를 가지고 작업을 처리해주는 단일시스템이미지(SSI) 제공이 필요하다. 기존의 SSI 제공 기술을 JVM(Java Virtual Machine) 계층을 기준으로 살펴보면, JVM 상위계층에서 구현된 SSI는 불완전하고, JVM 하위계층에서 구현된 SSI는 Java 자체의 성능 향상 기능을 활용하지 못하는 단점을 가진다. 이러한 단점들을 해결한 JVM 계층에서의 SSI가 구현되고 현재 연구가 진행중이지만, 이 역시 동기종 클러스터 시스템에 국한되어 있다. 본 논문에서는 클러스터 시스템의 SSI 구현 방법들을 각 구현 계층별로 비교하여 살펴보고, JVM을 기반으로 SSI를 제공함으로써 이기종 클러스터 시스템에까지 SSI를 확장, 제공하는 방안을 모색하고자 한다.

• The KIPS Transactions:PartA
• /
• v.10A no.4
• /
• pp.331-338
• /
• 2003

Cluster is the cost-effective parallel processing environment, and consists of the off-the-shelf computers connected by the computer net works. The characteristics of cluster are the node heterogeneity, the variety of node load, and the variety of network load. Because these characteristics influence the performance of parallel program executions, the load sharing for cluster is important, and by using the proper load sharing strategy, we can reduce the execution time of parallel programs. In this paper, we propose modified GSS algorithm, αGSS. In the proposed load sharing algorithms α GSS, the size of tasks are decided using the BogoMIPS of node. From the result of out experiments, we conclude that the proposed αGSS algorithm is effective in the heterogeneous cluster.

• The KIPS Transactions:PartA
• /
• v.10A no.3
• /
• pp.225-230
• /
• 2003

In this paper, we develope algorithms that distribute the load on the heterogeneously clustered web system, The response time based on the concurrent user is examined for the suggested algorithms. Simulation experience shows that the response time using the dynamically weighted methods seems to have a good results compare to that with the fixed weighted methods. And, also the effectiveness of clustered system becomes better as long as the number of concurrent user increases.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2002.11a
• /
• pp.405-408
• /
• 2002

Cluster is the cost-effective parallel processing environment, and consists of the off-the-shelf computers connected by the computer networks. The characteristics of cluster are the node heterogeneity, the variety of node load, and the variety of network load. Because these characteristics influence the performance of parallel program executions, the load sharing for cluster is important, and by using the proper load sharing strategy, we can reduce the execution time of parallel programs. In this paper, we propose modified GSS algorithm, aGSS. In the proposed load sharing algorithms aGSS, the size of tasks are decided using the BogoMTPS of node. From the result of out experiments, we conclude that the proposed aGSS algorithm is effective in the heterogeneous cluster.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.38A no.6
• /
• pp.520-527
• /
• 2013

In this paper, a joint cell grouping and user association technique proposed for a semi-dynamic grouped network MIMO in a clustered heterogeneous cellular network (HCN). With the conventional macro BSs, small cells are being overlaid to increase a spectral efficiency per area and these small cells are expected to be concentrated to support exponentially increasing data traffic in hot spot areas. The main culprits of performance degradation in the clustered HCN are interference and load imbalance. The proposed scheme jointly handles them to maximize a proportional-fair metric. It is shown that the proposed technique can largely improve user average rate and proportional fairness among users than any other conventional schemes in the clustered HCN.

• Journal of KIISE:Computing Practices and Letters
• /
• v.8 no.2
• /
• pp.183-194
• /
• 2002

The clustering environment with heterogeneous workstations provides the cost effectiveness and usability for executing applications in parallel. Load balancing is considered a necessary feature for a cluster of heterogeneous workstations to minimize the turnaround time. Previously, static load balancing that assigns a predetermined weight for the processing capability of each workstation, or dynamic approaches which execute a benchmark program to get relative processing capability of each workstation were proposed. The execution of the benchmark program, which has nothing to do with the application being executed, consumes the computation time and the overall turnaround time is delayed. In this paper, we present efficient methods for task distribution and task migration, based on the relative load index. We designed and implemented a load balancing system for the clustering environment with heterogeneous workstations. Turnaround times of our methods and the round-robin approach, as well as the load balancing method using a benchmark program, were compared. The experimental results show that our methods outperform all the other methods that we compared.