• Proceedings of the IEEK Conference
• /
• 2003.07a
• /
• pp.374-377
• /
• 2003

In this paper, a performance evaluation model of the Fat-Tree network with the multiple-buffered crossbar switches is proposed and examined. Buffered switch technique is well known to solve the data collision problem in the switch network The proposed evaluation model is developed by investigating the transfer patterns of data packets in a switch with output-buffers. Steady state probability concept is used to simplify the analyzing processes. Two important parameters of the network performance, throughput and delay, are then evaluated. To validate the proposed analysis model, the simulation is carried out on the various sizes of Fat-tree networks that use the multiple a$\times$b buffered crossbar switches. It is observed that both analysis and simulation results are match closely.

• Journal of KIISE:Information Networking
• /
• v.30 no.4
• /
• pp.520-534
• /
• 2003

In this paper, a performance evaluation model of the Fat-tree Network with the multiple-buffered crossbar switches is proposed and examined. Buffered switch technique is well known to solve the data collision problem of the switch network. The proposed evaluation model is developed by investigating the transfer patterns of data packets in a switch with output-buffers. Two important parameters of the network performance, throughput and delay, are then evaluated. The proposed model takes simple and primitive switch networks, i.e., no flow control and drop packet, to demonstrate analysis procedures clearly. It, however, can not only be applied to any other complicate modern switch networks that have intelligent flow control but also estimate the performance of any size networks with multiple-buffered switches. To validate the proposed analysis model, the simulation is carried out on the various sizes of Fat-tree networks that uses the multiple buffered crossbar switches. Less than 2% differences between analysis and simulation results are observed.

• Proceedings of the KIEE Conference
• /
• 2000.11d
• /
• pp.775-777
• /
• 2000

본 논문에서는 buffer를 장착한 양 방향성 $a{\times}b$ switch들로 구성된 fat-tree network의 성능 분석 기법을 제안하고, 분석 모형의 타당성을 검증하였다. 제안한 분석 기법은 먼저 스위치 내부의 데이터 이동 패턴을 확률식으로 표현하고. 나아가서 buffer를 장착한 $a{\times}b$ switch의 buffer 크기에 따른 정상상태 throughput을 간단한 수식으로 구할 수 있도록 하였다. 이를 토대로 buffer를 장착한 $a{\times}b$ switch로 구성된 fat-tree network의 성능을 분석하고, 제안한 분석모형의 실효성 입증을 위하여 simulation을 시행한 후 결과를 비교 분석하였다.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.13 no.4
• /
• pp.9-15
• /
• 2013

Most of recent studies on data center networks are based on the assumption of symmetrical multi-rooted tree topologies such as a Fat-Tree. While packet scatter schemes are very effective for such a network topology, it is known that various failures can result in an asymmetric topology which degrades TCP performance. In this paper, we reexamine the effects of link failures on packet scatter schemes in Fat-Trees. Our simulation results show that in case of a single link failure in a large-scale Fat-Tree, packet reordering does not occur enough to degrade TCP performance. This implies that we do not necessarily need a complex scheme to make packet schemes robust to link failures.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.12 no.6
• /
• pp.215-221
• /
• 2012

To address the bottleneck problem in data center networks, there have been several proposals for network architectures providing high path-diversity. In devising new schemes to utilize multiple paths, one must consider the effects on TCP performance because packet reordering can make TCP perform poorly. Therefore most schemes prevent packet reordering by sending packets through one of multiple available paths. In this study we show that packet reordering does not occur severely enough to have a significant impact on TCP performance when scattering packets through all available paths between a pair of hosts in Fat-Tree. Simulation results imply that it is possible to find a low-cost solution to the TCP performance problem for Fat-Tree-like topologies.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.13 no.2
• /
• pp.53-58
• /
• 2013

A Fat-Tree topology has multiple paths between any pair of hosts. The delay for the multiple paths with an equal number of hops depends mainly on the queuing delay. However, most of the existing load-balancing schemes do not sufficiently exploit the characteristics of Fat-Tree. In most schemes load-balancing is performed at a flow level. Packet-level load-balancing schemes usually require the availability of special transport layer protocols to address packet reordering. In this paper, we propose a new packet-level load-balancing scheme which can enhance network utilization while minimizing packet reordering in Fat-Trees. Simulation results show that the proposed scheme provides as high TCP throughput as a randomized flow-level Valiant load balancing scheme for a best case.

• Proceedings of the Korean Information Science Society Conference
• /
• 2003.04d
• /
• pp.256-258
• /
• 2003

본 논문에서는, a$\times$b 출력 버퍼 스위치로 구성된 fat-tree 망의 성능 예측 모형을 제안하고, 스위치에 장착된 버퍼의 개수 증가에 따를 성능 향상 추이를 분석하였다. 제안한 성능 예측 모형은 먼저 네트웍 내부 임의 스위치 입력 단에 유입되는 데이터 패킷이 스위치 내부에서 전송되는 유형을 확률적으로 분석하여 수립되었다. 성능분석 모형은 스위치에 장착된 버퍼의 개수와 무관하게 버퍼를 장착한 a$\times$b 스위치의 성능, 네트웍 정상상태 처리율(Steady state Throughput, ST)과 네트웍 지연시간(Network Delay)을 간단한 확률식으로 구할 수 있다. 제안한 수학적 성능 분석 연구의 실효성 검증을 위하여 병행된 시뮬레이션 결과는 상호 미세한 오차 범위 내에서 오형의 예측 데이터와 일치하는 결과를 보여 분석 모형의 타당성을 입증하였다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.21 no.9
• /
• pp.1655-1665
• /
• 2017

Recently data centers are being constructed actively by many cloud service providers, enterprises, research institutes, etc. Generally, they are built on tree topology using ECMP data forwarding scheme for load balancing. In this paper, we examine data center network topologies like tree topology and fat-tree topology, and load balancing technologies like MLAG and ECMP. Then, we propose a scheduling algorithm to efficiently transmit particular files stored on the hosts in the data center to the destination node outside the data center, where fat-tree topology and OpenFlow protocol between infrastructure layer and control layer are used. We run performance analysis by numerical method, and compare the analysis results with those of ECMP. Through the performance comparison, we show the outperformance of the proposed algorithm in terms of throughput and file transfer completion time.

• Journal of KIISE:Information Networking
• /
• v.34 no.3
• /
• pp.203-217
• /
• 2007

In this paper, a performance evaluation model of Networks with the multiple-buffered crossbar switches is proposed and examined. Buffered switch technique is well known to solve the data collision problem of the switch networks. The characteristic of a network with crossbar switches is determined by both the connection pattern of the switches and the limitation of data flow in a each switch. In this thesis, the evaluation models of three different networks : Multistage interconnection network, Fat-tree network, and other ordinary communication network are developed. The proposed evaluation model is developed by investigating the transfer patterns of data packets in a switch with output-buffers. Two important parameters of the network performance, throughput and delay, are evaluated. The proposed model takes simple and primitive switch networks, i.e., no flow control and drop packet, to demonstrate analysis procedures clearly. It, however, can not only be applied to any other complicate modern switch networks that have intelligent flow control but also estimate the performance of any size networks with multiple-buffered switches. To validate the proposed analysis model, the simulation is carried out on the various sizes of networks that uses the multiple buffered crossbar switches. It is shown that both the analysis and the simulation results match closely. It is also observed that the increasing rate of Normalized Throughput is reduced and the Network Delay is getting bigger as the buffer size increased.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.10 no.10
• /
• pp.4787-4807
• /
• 2016

Although the dense interconnection datacenter networks (DCNs) (e.g. FatTree) provide multiple paths and high bisection bandwidth for each server pair, the single-path TCP (SPT) and ECMP which are widely used currently neither achieve high bandwidth utilization nor have good load balancing. Due to only one available transmission path, SPT cannot make full use of all available bandwidth, while ECMP's random hashing results in many collisions. In this paper, we present OFPT, an OpenFlow based Parallel Transport framework, which integrates precise routing and scheduling for better load balancing and higher network throughput. By adopting OpenFlow based centralized control mechanism, OFPT computes the optimal path and bandwidth provision for each flow according to the global network view. To guarantee high throughput, OFPT dynamically schedules flows with Seamless Flow Migration Mechanism (SFMM), which can avoid packet loss in flow rerouting. Finally, we test OFPT on Mininet and implement it in a real testbed. The experimental results show that the average network throughput in OFPT is up to 97.5% of bisection bandwidth, which is higher than ECMP by 36%. Besides, OFPT decreases the average flow completion time (AFCT) and achieves better scalability.