Restricted Duplication Based MILP Formulation for Scheduling Task Graphs on Unrelated Parallel Machines

Singh, Jagpreet; Mangipudi, Bhargav; Betha, Sandeep; Auluck, Nitin

doi:10.1109/paap.2012.37

Cited by 7 publications

(3 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While they have shown similarly promising results as the pure branch-and-bound approach, neither method has been shown to have significantly better performance than the other. Some ILP formulations with duplication have been proposed [18], [23], [19]. In [23], duplication of tasks was used in order to increase the reliability of schedules, but its use in reducing schedule lengths was not considered.…”

Section: Related Workmentioning

confidence: 99%

“…All tasks were considered for duplication on all processors, making the complexity very high. This was built on by [19] with a restricted duplication formulation (RESDMILP). This approach restricts the number of copies of each task according to a given parameter, reducing the complexity of the ILP but sacrificing the optimality of the solution.…”

Section: Related Workmentioning

confidence: 99%

“…One factor in making this decision would be the likelihood that allowing duplication will cause a reduction in the length of the optimal schedule for a given task graph. [19] Unrelated heterogeneous processors Restricted Up to 20 tasks 1000 sec limit Tang 2016 [21] Allocation predetermined Predetermined 14-48 tasks 10 min limit Table 2 Comparison of reported successful optimal solving between ILP formulations with duplication.…”

Section: Duplication Benefitmentioning

confidence: 99%

See 2 more Smart Citations

Integrating Task Duplication in Optimal Task Scheduling With Communication Delays

Orr

Sinnen

2020

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

Task scheduling with communication delays is an NP-hard problem. Some previous attempts at finding optimal solutions to this problem have used branch-and-bound state-space search, with promising results. Duplication is an extension to the task scheduling model which allows tasks to be executed multiple times within a schedule, providing benefits to schedule length where this allows a reduction in communication costs. This paper proposes the first approach to state-space search for optimal task scheduling with task duplication.Also presented are new definitions for important standard bounding metrics in the context of duplication. An extensive empirical evaluation shows that the use of duplication significantly increases the difficulty of optimal scheduling, but the proposed approach also gives certainty that a large proportion of task graphs can be scheduled more effectively when duplication is allowed, and permits to quantify the exact advantage.

show abstract