Time Division Dominant Resource Allocation in Cloud Environments

  • Author / Creator
    Zhang, Xu
  • Providing service in cloud is a popular trend nowadays. Users request resources from a shared pool in a framework like Hadoop, and the Hadoop fair scheduler controls the progress of tasks from users. In cloud, usually multiple resource types are available. The heterogeneous user demands and competition between users in cloud make the fairness control of multiple-resource allocation much more complicated than the single-resource allocation. A recent work proposed a Dominant Resource Fairness (DRF) method for multiple-resource allocation, which has been included in Hadoop Next-Generation. However, DRF may not lead to desired fairness performance as a result of the indivisibility of the user demands. This thesis proposes a Time Division Allocation (TDA) method with time slots for two users. The TDA method allocates resources over the time slots and each time slot is assigned a different allocation. By adjusting the length of the time slots and the resource allocations, TDA method can achieve a global optimal max-min fairness. A further study shows that two time slots are sufficient to achieve optimality. Besides the analysis about fairness, this thesis also explores the theoretical performance bounds for two users. This analysis illustrates the relationship between user demands and the maximum dominant resource share. The evaluation shows that the TDA method performs better than the DRF method.

  • Subjects / Keywords
  • Graduation date
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.