Approximation algorithms for scheduling monotonic moldable tasks on multiple platforms

We consider scheduling monotonic moldable tasks on multiple platforms, where each platform contains a set of processors. A moldable task can be split into several pieces of equal size and processed simultaneously on multiple processors. Tasks are not allowed to be processed spanning over platforms....

Full description

Saved in:
Bibliographic Details
Published in:Journal of scheduling Vol. 26; no. 4; pp. 383 - 398
Main Authors: Wu, Fangfang, Jiang, Zhongyi, Zhang, Run, Zhang, Xiandong
Format: Journal Article
Language:English
Published: New York Springer US 01.08.2023
Springer Nature B.V
Subjects:
Algorithms
Approximation
Artificial Intelligence
Business and Management
Calculus of Variations and Optimal Control; Optimization
Cranes
Mathematical analysis
Microprocessors
Operations research
Operations Research/Decision Theory
Optimization
Platforms
Polynomials
Processors
Supply Chain Management
Task scheduling
Moldable tasks
Dual approximation algorithm
Scheduling
Approximation algorithm
Multiple platforms
ISSN:1094-6136, 1099-1425
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We consider scheduling monotonic moldable tasks on multiple platforms, where each platform contains a set of processors. A moldable task can be split into several pieces of equal size and processed simultaneously on multiple processors. Tasks are not allowed to be processed spanning over platforms. This scheduling model has many applications, ranging from parallel computing to the berth and quay crane allocation and the workforce assignment problem. We develop several approximation algorithms aiming at minimizing the makespan. More precisely, we provide a 2-approximation algorithm for identical platforms, a Fully Polynomial Time Approximation Scheme (FPATS) under the assumption of large processor counts and a 2-approximation algorithm for a fixed number of heterogeneous platforms. Most of the proposed algorithms combine a dual approximation scheme with a novel approach to improve the dual approximation algorithm. All results can be extended to the contiguous case, i.e., a task can only be executed by contiguously numbered processors.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1094-6136
1099-1425
DOI:10.1007/s10951-022-00774-2