• DocumentCode
    1925373
  • Title

    Resource manager with multi-core support for parallel desktop

  • Author

    García, José R. ; Lerida, J. Ll ; Hernández, Porfidio

  • Author_Institution
    Comput. Sci. & Oper. Syst. Dept., Univ. Autonoma de Barcelona, Barcelona, Spain
  • fYear
    2009
  • fDate
    Aug. 31 2009-Sept. 4 2009
  • Firstpage
    1
  • Lastpage
    4
  • Abstract
    Due to the fall in the price of multicore processors, today´s non-dedicated clusters tend to include this kind of hardware in their configurations. However, most current resource managers and job schedulers are optimized with the aim of maximizing the throughput in single-core environments. Additionally, job schedulers balance thread loads without considering such aspects as cache affinity, resource contention or other additional criteria. In the same way, general purpose Operating System schedulers will support requirements like the coexistence of soft-real time, best effort or interactive applications are open questions that need to be addressed carefully. For these reasons, new user interfaces, middlewares, and environments that can fill the gap between the user and the new processor architectures are needed. Our CISNE environment has been extended to take into account issues discussed earlier: the mixture of different types of applications considering the multicore features of current desktop environments.
  • Keywords
    parallel processing; resource allocation; scheduling; general purpose operating system; job scheduling; multicore processors; multicore support; parallel desktop; resource management; Environmental management; Hardware; Middleware; Multicore processing; Operating systems; Processor scheduling; Resource management; Throughput; User interfaces; Yarn;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Cluster Computing and Workshops, 2009. CLUSTER '09. IEEE International Conference on
  • Conference_Location
    New Orleans, LA
  • ISSN
    1552-5244
  • Print_ISBN
    978-1-4244-5011-4
  • Electronic_ISBN
    1552-5244
  • Type

    conf

  • DOI
    10.1109/CLUSTR.2009.5289134
  • Filename
    5289134