• DocumentCode
    2497430
  • Title

    Scheduling Traffic Matrices On General Switch Fabrics

  • Author

    Wu, Xiang ; Prakash, Amit ; Mohiyuddin, Marghoob ; Aziz, Adnan

  • fYear
    2006
  • fDate
    23-25 Aug. 2006
  • Firstpage
    87
  • Lastpage
    92
  • Abstract
    A traffic matrix is an |S| times |T| matrix M, where Mij is a non-negative integer encoding the number of packets to be transferred from source i to sink j. Chang et al. (2001) have shown how to efficiently compute an optimum schedule for transferring packets from sources to sinks when the sources and sinks are connected via a rearrangeable fabric such as crossbar. We address the same problem when the switch fabric is not rearrangeable. Specifically, we (1) prove that the optimum scheduling problem is NP-hard for general switch fabrics, (2) identify a sub-class of fabrics for which the problem is polynomial-time solvable, and (3) develop a heuristic for the general case
  • Keywords
    encoding; packet switching; polynomial matrices; scheduling; telecommunication traffic; NP-hard; crossbar; general switch fabrics; nonnegative integer encoding; polynomial-time solvable; scheduling traffic matrices; traffic matrix; Bipartite graph; Fabrics; Matrix decomposition; Multicore processing; Polynomials; Processor scheduling; Switches;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    High-Performance Interconnects, 14th IEEE Symposium on
  • Conference_Location
    Stanford, CA
  • ISSN
    1550-4794
  • Print_ISBN
    0-7695-2654-3
  • Type

    conf

  • DOI
    10.1109/HOTI.2006.22
  • Filename
    1690203