• DocumentCode
    474862
  • Title

    Towards an efficient switch architecture for high-radix switches

  • Author

    Mora, G. ; Flich, J. ; Duato, J. ; Lopez, Pierre ; Baydal, E. ; Lysne, O.

  • Author_Institution
    Tech. Univ. of Valencia, Valencia
  • fYear
    2006
  • fDate
    3-5 Dec. 2006
  • Firstpage
    11
  • Lastpage
    20
  • Abstract
    The interconnection network plays a key role in the overall performance achieved by high performance computing systems, also contributing an increasing fraction of its cost and power consumption. Current trends in interconnection network technology suggest that high-radix switches will be preferred as networks will become smaller (in terms of switch count) with the associated savings in packet latency, cost, and power consumption. Unfortunately, current switch architectures have scalability problems that prevent them from being effective when implemented with a high number of ports. In this paper, an efficient and cost-effective architecture for high-radix switches is proposed. The architecture, referred to as partitioned crossbar input queued (PCIQ), relies on three key components: a partitioned crossbar organization that allows the use of simple arbiters and crossbars, a packet-based arbiter, and a mechanism to eliminate the switch-level HOL blocking. Under uniform traffic, maximum switch efficiency is achieved. Furthermore, switch-level HOL blocking is completely eliminated under hot-spot traffic, again delivering maximum throughput. Additionally, PCIQ inherently implements an efficient congestion management technique that eliminates all the network-wide HOL blocking. On the contrary, the previously proposed architectures either show poor performance or they require significantly higher costs than PCIQ (in both components and complexity).
  • Keywords
    multistage interconnection networks; congestion management technique; high-radix switch architecture; interconnection network; partitioned crossbar input queued; power consumption; switch-level head-of-line block elimination; Computer architecture; Costs; Delay; Energy consumption; High performance computing; Multiprocessor interconnection networks; Packet switching; Switches; Telecommunication traffic; Traffic control; arbiter efficiency; partitioned crossbar; switch organization;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Architecture for Networking and Communications systems, 2006. ANCS 2006. ACM/IEEE Symposium on
  • Conference_Location
    San Jose, CA
  • Print_ISBN
    978-1-59593-580-9
  • Type

    conf

  • Filename
    4579519