4th International ICST Conference on Communications and Networking in China

Research Article

A load balancing scheme for two-stage switches with minimum buffers for scalability

  • @INPROCEEDINGS{10.1109/CHINACOM.2009.5339764,
        author={Hui  Li  and Liang-Min  Lin and Jia-Qing  Huang and Hui-Yao  AN and Peng   Yi  and Bin-Qiang Wang},
        title={A load balancing scheme for two-stage switches with minimum buffers for scalability},
        proceedings={4th International ICST Conference on Communications and Networking in China},
        publisher={IEEE},
        proceedings_a={CHINACOM},
        year={2009},
        month={11},
        keywords={Load balancing; Self-routing; Large Scale Switch},
        doi={10.1109/CHINACOM.2009.5339764}
    }
    
  • Hui Li
    Liang-Min Lin
    Jia-Qing Huang
    Hui-Yao AN
    Peng Yi
    Bin-Qiang Wang
    Year: 2009
    A load balancing scheme for two-stage switches with minimum buffers for scalability
    CHINACOM
    IEEE
    DOI: 10.1109/CHINACOM.2009.5339764
Hui Li 1,2,*, Liang-Min Lin1,2, Jia-Qing Huang3, Hui-Yao AN1,2, Peng Yi 4, Bin-Qiang Wang4
  • 1: Key Lab of Integrated Microsystems, Shenzhen Graduate School, Peking University , Shenzhen 518055, China
  • 2: Shanghai Research Institute of Microelectronics, Peking University, Shanghai 201203 China
  • 3: Dept. of Electronics & Info..Eng., Huazhong University of Science & Technology, Wuhan, 430074, China;
  • 4: University of Information Engineering, National Digital Switching Center, Zhengzhou,450002, China;
*Contact email: lih64@szpku.edu.cn

Abstract

In this paper, we propose a novel load-balanced scheme-split aggregated flow (SAF) and design the buffering mechanism for two-stage self-routing switch. By grouping the signal lines with concentrators, the proposed scheme can obtain the statistical multiplex gain and reduce the complexity of computation. The mathematical analysis and simulations show that 100% throughput can be achieved for any admissible traffic pattern. Compared with other load-balancing schemes used crossbar as the basic fabric, this scheme has the distinctive advantages including the lowest complexity buffer of O(N), low queuing delay O(1), and free from the out-of-sequence problem. These properties make it practicably suitable for very large scale switching structures in Next Generation Network (NGN).