• Proceedings of the Korean Information Science Society Conference
• /
• 2004.04a
• /
• pp.13-15
• /
• 2004

CC-NUMA 시스템은 메모리를 분산시켜 트랜잭션을 지역화 함으로써 고성능 및 확장성을 꾀하는 구조이다. 그러나 CC-NUMA 시스템에서 여러 병렬 프로그램들이 수행될 경우, 각 프로그램의 부하 차이에 의하여 균등한 네트워크 활용율을 나타내지 못하며. 이중 링 CC-NUMA 시스템에서 이러한 불균등한 네트워크 부하로 인한 성능 감소가 발생한다. 본 논문은 이중 연결 구조중 하나를 건너뜀 연결을 갖도록 배치하여 균등한 네트워크 부하를 나타내도록 하며, 이중 링에 비하여 균등한 네트워크 부하를 나타냄을 시뮬레이션을 통하여 검증한다.

• Journal of KIISE:Computer Systems and Theory
• /
• v.29 no.9
• /
• pp.503-513
• /
• 2002

Since CC-NUMA architecture has to access remote memory, the interconnection network determines the performance of the CC-NUMA system. Bus which has been used as a popular interconnection network has many limits in a large-scale system because of the limited physical scalability and bandwidth. The dual ring interconnection network, composed of high-speed point-to-point links, is made to resolve the defects of the bus for the large-scale system. However, it also has a problem, in that the response latency is rapidly increased when many nodes are attached to the snooping based CC-NUMA system with the dual ring. In this paper, we propose a ring architecture with repeater nodes in order to overcome the problem of the dual ring on a snooping based CC-NUMA system, and design a repeater node adapted to this architecture. We will also analyze the effects of proposed architecture on the system performance and the response latency by using a probability-driven simulator.

• Journal of KIISE:Computer Systems and Theory
• /
• v.31 no.11
• /
• pp.643-657
• /
• 2004

The chordal ring based CC-NUMA system contains many links to transmit transactions between a local node and a remote node because of its structural characteristics. However, the inclination that the transactions concentrate on the ring link increases both the traffic of the ring link and the response time, which degrades the overall performance of the chordal ring based CC-NUMA system. In this paper we suggest a new remote cache replacement policy that considers both the number of total links and the number of ring links to traverse for the transactions. Our proposed replacement policy can balance data between the ring link and the chordal link properly because it reflects the characteristics of chordal ring based CC-NUMA system well.

• Journal of KIISE:Computer Systems and Theory
• /
• v.31 no.9
• /
• pp.487-494
• /
• 2004

The multiprocessor system suffers interconnection network contension while exploiting the program's parallelism. A CC-NUMA system based on point-to-point link ring structure is one of the scalable architectures that expand the system bandwidth the number of processors/nodes increases. The dual-ring system is a simple solution to enhance the system performance and scalability by duplicating the links. In ring-based systems, an unbalanced transaction among links makes a hot spot on the interconnection network. In this situation, total system performance and scalability are bound by the hot spot of the links In this paper, I propose a dual-link CC-NUMA system, which alleviates the concentration of transactions among the links. By the simulation results, the proposed system significantly outperforms the single-ring and bidirection dual-ring systems. In addition, the proposed system show better distribution of transactions among the links that achieves an extended scalability.

• Journal of KIISE:Computer Systems and Theory
• /
• v.29 no.2
• /
• pp.105-115
• /
• 2002

Since NUMA architecture has to access remote memory an interconnection network determines the performance of CC-NUMA system Bus which has been used as a popular interconnection network has many limits to build a large-scale system because of the limited physical scalabilty and bandwidth Dual ring interconnection network composed of high speed point-to-point links is made up for resolving the defects of the bus for large-scale system But it also has a problem that the response latency is rapidly increased when many node are attached to snooping based CC-NUMA system with dual ring In this paper we propose a chordal ring architecture in order to overcome the problem of the dual ring on snooping based CC-NUMA system and design and efficient link controller adopted to this architecture. We also analyze the effects of chordal ring architecture on the system performance and the response latency by using probability driven simulator.

• The KIPS Transactions:PartA
• /
• v.11A no.3
• /
• pp.199-206
• /
• 2004

Under the multiprogrammed situation, the performance of multiprocessor system is affected by the process allocation policy of the operating systems. The lowest communication cost can be achieved when the related processes positioned to the adjacent processors. While the effective allocation is quite difficult to the real situation, and the processing of the allocation policy consumes some computation time. The dual-ring CC-NUMA systems exhibit a quite performance difference according to the process a1location policy due to a lot of unbalanced memory transactions on the interconnection networks. In this paper, I propose a load balanced dual-link CC-NUMA system that does not requires the processes allocation policy. By the program-driven simulation results. the proposed system shows no remarkable difference according to the allocation policy while the dual-ring systems shows 10％ performance improvement by the process allocation. In addition, the proposed system outperforms the dual~ring systems about 1.5 times.

• Proceedings of the Korean Information Science Society Conference
• /
• 2002.10c
• /
• pp.697-699
• /
• 2002

CC-NUMA 구조에서는 원격 메모리에 대한 접근이 불가피한 구조적인 특성 때문에 사호 연결망이 성능을 좌우하는 큰 변수로 작용한다. 기존에 사용되는 버스는 대역폭의 한계와 물리적 확장성 때문에 대규모의 시스템에는 적합하지 않다. 이를 대체하는 고속의 지점간 링크를 도입한 이중 링 구조는 이러한 버스의 한계를 극복하고는 있지만 많은 노드를 거쳐야 하는 문제로 인해 응답 지연 시간이 증가하는 단점을 안고 있다. 본 논문에서는 요청과 응답 패킷의 지연 시간을 줄이는 방안으로 리피터 노드를 이용한 다중링을 제안한다. 제안된 시스템은 링과 링 사이의 구조가 대칭형을 이루고 있어 요청을 내보내는 링을 제외한 다른 링의 hop수는 똑같은 수치를 갖고 있으며, 이중 링에 비해 최대의 hop수와 최소의 hop수의 차가 적고 평균 hop수 또한 적어 좋은 성능을 보인다. 본 논문에서는 또한 이러한 구조를 유지하기 위한 리피터 노드의 구조를 제안하며 리피터 노드의 구조와 노드의 확장에 따른 다양한 성능을 확률 구동 시뮬레이터를 사용하여 평가를 수행한다.

• The Transactions of the Korea Information Processing Society
• /
• v.7 no.2
• /
• pp.488-496
• /
• 2000

this paper describes the design and implementation of node scheduling for MX Server that is CC-NUMA System COMSIX, the operating system of MX Server, is designed to suit for CC-NUMA Architecture. MX Server consists of up to 8 nodes, and each node is connected by SCI ring. This node scheduling scheme considers data locality for performance improvement of Oracle8i DBMS on the CC-NUMA architecture. For DBMS such as Oracle8i, a multi-threaded process may be run to tie on particular disk. We have developed a CG binding function that the multi-threaded process bound the node. Currently, We don't have an available CC-NUMA Platform. Instead of MX Server, we developed the Node scheduling scheme for multi-threaded process to suit server platform on the PC test-bed and tested completely.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2004.05a
• /
• pp.929-932
• /
• 2004

DRACO 구조는 이중 링 연결형태에 대비하여 노드간 접근 경로를 단축할 수 있는 구조로, CC-NUMA 시스템에 대하여 그 적용성과 프로토콜이 시험되어 그 효율성이 입증되었다. 본 논문은 이러한 DRACO 구조가 보다 많은 프로세서를 수용할 수 있는 클러스터 시스템에 대하여 동일한 경로 단축과 대역폭 확장을 적용할 수 있음에 기반하여, 이중 연결 링크를 가지고 있는 점대 점 연결 형태의 클러스터 시스템에 대한 적용성을 연구하였다. 노드 내의 네트워크 인터페이스 및 소프트웨어만을 이용한 DRACO 구조 적용은 많은 수의 노드를 필요로 하는 시스템에 적합하고, 하드웨어를 이용한 연결 경로를 이용할 경우 상대적으로 적은 수의 노드를 채용하고 고속의 링크 성능을 필요로 하는 시스템에 적합할 것이다.

• The KIPS Transactions:PartA
• /
• v.11A no.1
• /
• pp.49-56
• /
• 2004

The performance of the multiprocessor systems is limited by the several factors. The system performance is affected by the processor speed, memory delay, and interconnection network bandwidth/latency. By the evolution of semiconductor technology, off the shelf microprocessor speed breaks beyond GHz, and the processors can be scalable up to multiprocessor system by connecting through the interconnection networks. In this situation, the system performances are bound by the latencies and the bandwidth of the interconnection networks. SCI, Myrinet, and Gigabit Ethernet are widely adopted as a high-speed interconnection network links for the high performance cluster systems. Performance improvement of the interconnection network can be achieved by the bandwidth extension and the latency minimization. Speed up of the operation clock speed is a simple way to accomplish the bandwidth and latency betterment, while its physical distance makes the difficulties to attain the high frequency clock. Hence the system performance and scalability suffered from the interconnection network limitation. Duplicating the link of the interconnection network is one of the solutions to resolve the bottleneck of the scalable systems. Dual-ring SCI link structure is an example of the interconnection network improvement. In this paper, I propose a network topology and a transaction path algorism, which optimize the latency and the efficiency under the duplicated links. By the simulation results, the proposed structure shows 1.05 to 1.11 times better latency, and exhibits 1.42 to 2.1 times faster execution compared to the dual ring systems.