ERA

Download the full-sized PDF of Scheduing problems on Stars and PathsDownload the full-sized PDF

Analytics

Share

Permanent link (DOI): https://doi.org/10.7939/R3NV99Q8S

Download

Export to: EndNote  |  Zotero  |  Mendeley

Communities

This file is in the following communities:

Graduate Studies and Research, Faculty of

Collections

This file is in the following collections:

Theses and Dissertations

Scheduing problems on Stars and Paths Open Access

Descriptions

Other title
Subject/Keyword
approximation algorithms, flow-shop scheduling, min-sum edge colouring
Type of item
Thesis
Degree grantor
University of Alberta
Author or creator
Golestanian, Arnoosh
Supervisor and department
Salavatipour, Mohammad R (Computing Science)
Examining committee member and department
Friggstad, Zachary (Computing Science)
Lin, Guohui (Computing Science)
Department
Department of Computing Science
Specialization

Date accepted
2017-09-28T09:21:13Z
Graduation date
2017-11:Fall 2017
Degree
Master of Science
Degree level
Master's
Abstract
In this thesis, we consider scheduling problems in which jobs need to be processed through a (shared) network of machines according to their given paths. Formally, we are given a graph $G(V, E)$ where the edges $E$ represent the machines. We are also given a set of jobs $J$ and a path of edges for each of them showing that the job needs to be processed on those edges in that order. Each job can be moved to the next machine only if it has been fully processed by all the previous machines in the path. Moreover, each job takes $1$ unit of time to be processed on each machine and once a machine starts processing a job, it cannot process any other jobs. Our goal is to find a schedule with minimum total completion time. We consider two closely related scheduling problems, Star Scheduling with Unit Processing time (SSUP) and Generalized Path scheduling with Unit processing time (GPUP). As it is clear from their terminology, in SSUP the network of machines is a star and in GPUP it is a path. The most important contribution of this thesis is a $1.796$-approximation for the SSUP problem, described in Chapter 2. To achieve this we partition jobs into smaller subsets, in each subset, the number of jobs that share a machine is less than a specific number which is increasing geometrically. We also prove that for the case that jobs cannot have a delay in the middle of their processing, the problem is APX-hard. In Chapter 3, we consider GPUP problem and prove that in polynomial time one can find a schedule which minimizes the total completion time. Our analysis for the algorithm is new and is much simpler than the previous analysis.
Language
English
DOI
doi:10.7939/R3NV99Q8S
Rights
This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for the purpose of private, scholarly or scientific research. 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.
Citation for previous publication
Friggstad, Zachary, Arnoosh Golestanian, Kamyar Khodamoradi, Christopher Martin, Mirmahdi Rahgoshay, Mohsen Rezapour, Mohammad R. Salavatipour, and Yifeng Zhang. ``Scheduling problems over network of machines." APPROX 2017.

File Details

Date Uploaded
Date Modified
2017-09-28T15:21:13.525+00:00
Audit Status
Audits have not yet been run on this file.
Characterization
File format: pdf (PDF/A)
Mime type: application/pdf
File size: 686658
Last modified: 2017:11:08 17:49:50-07:00
Filename: Golestanian_Arnoosh_201709_MSc.pdf
Original checksum: 5d817d6a9418a7d8571be61e4d6bde81
Activity of users you follow
User Activity Date