Divisible workload applications occur in many fields of science and engineering. Although these application can be easily parallelized in a master-worker fashion, they pose several scheduling challenges. Previously proposed scheduling algorithms either distribute work to processors in a single round of work allocation or in multiple rounds. Multi-round algorithms can achieve better overlap of computation and communication but are more difficult to analyze. Consequently, a number of open questions still remain for multi-round scheduling. In this paper we improve on the seminal "multi-installment" algorithm proposed by Bharadwaj. et al. for homogeneous star networks. Their work suffers from two important limitations: (i) communication and computation latencies are assumed to be negligible; and (ii) size of application output data is assumed to be negligible. These two limitations strongly restrict the applicability of multi-installment to real-world platforms and applications. In this paper we remove both limitations.
The authors of these documents have submitted their reports to this technical report series for the purpose of non-commercial dissemination of scientific work. The reports are copyrighted by the authors, and their existence in electronic format does not imply that the authors have relinquished any rights. You may copy a report for scholarly, non-commercial purposes, such as research or instruction, provided that you agree to respect the author's copyright. For information concerning the use of this document for other than research or instructional purposes, contact the authors. Other information concerning this technical report series can be obtained from the Computer Science and Engineering Department at the University of California at San Diego, email@example.com.
[ Search ]