Scientific workflows scheduling in cloud
Dissertant
Thesis advisor
University
Princess Sumaya University for Technology
Faculty
King Hussein Faculty for Computing Sciences
Department
Department of Computer Sciences
University Country
Jordan
Degree
Master
Degree Date
2018
English Abstract
Scientific workflows cover different topics such as astronomy, physics, and others.
In such application workflows, formed as Directed Acyclic Graph (DAG), vertices represent tasks with estimated runtime and edges represent dependencies and data flow between tasks.
Recently, with the massive growth of scientific workflows, cloud computing has become a promising computing environment that provides the ability of on-demand resource allocation and pay-as-you-go pricing model.
Normally, the objective of any scientific workflow scheduling algorithm is to minimize the execution cost and/or the execution time.
In addition, to efficiently schedule the scientific workflows, the right amount and type of resources (VMs) to rent should be determined.
Over-renting results in increasing the cost of the execution, while under-renting results in increasing the execution time.
This underlines designing an efficient algorithmic solution that addresses the bi-objective problem of minimizing execution time and cost as a major challenge.
This thesis presents Location-Aware Scheduling (LAS) algorithm, which divides the given workflow into group of partitions with different "demands." A partition demand represents the number of resources (VMs), such that this partition can achieve the lowest possible execution time.
Typically, the number of available VMs will not be enough to satisfy all partitions demands.
Thus, this algorithm works by dividing the resources "fairly" between the partitions.
The fair division aims to divide the resources between the partitions in a manner that no partition envies any other partition’s share of the resources (envy-free).
This division strategy aims to increase the utilization of the resources and address the objective on reducing the execution cost and time.
This is established, since the resources allocated to each partition depend on its requirements, and every partition is satisfied with its allocation.
Experiments conducted to evaluate the performance of LAS algorithm against three other comparable well-known algorithms from the literature.
The comparison was done against HEFT, RDAS and PBWS algorithms.
The results show that the proposed algorithm improves the utilization of the resources by 60% and makespan minimization with 22%.
In addition, in most scenarios, it outperforms the other algorithm in term of cost (over 30%) while achieving reasonable execution time
Main Subjects
Information Technology and Computer Science
Topics
No. of Pages
65
Table of Contents
Table of contents.
Abstract.
Abstract in Arabic.
Chapter One : Introduction.
Chapter Two : Background and related work.
Chapter Three : Location-aware scheduling (LAS) algorithm.
Chapter Four : Evaluation and discussion.
Chapter Five : Conclusions and future works.
References.
American Psychological Association (APA)
al-Dabaybah, Balqis Muhammad. (2018). Scientific workflows scheduling in cloud. (Master's theses Theses and Dissertations Master). Princess Sumaya University for Technology, Jordan
https://search.emarefa.net/detail/BIM-833195
Modern Language Association (MLA)
al-Dabaybah, Balqis Muhammad. Scientific workflows scheduling in cloud. (Master's theses Theses and Dissertations Master). Princess Sumaya University for Technology. (2018).
https://search.emarefa.net/detail/BIM-833195
American Medical Association (AMA)
al-Dabaybah, Balqis Muhammad. (2018). Scientific workflows scheduling in cloud. (Master's theses Theses and Dissertations Master). Princess Sumaya University for Technology, Jordan
https://search.emarefa.net/detail/BIM-833195
Language
English
Data Type
Arab Theses
Record ID
BIM-833195