A study on meta-heuristic scheduling mechanisms for fault-tolerant distributed computing결함 감내형 분산 컴퓨팅을 위한 메타 휴리스틱 스케줄링 메커니즘에 관한 연구

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 763
  • Download : 0
Executing scientific workload is computation-intensive, thus time-consuming, however job completion time is a complex non-linear function of allocated resources and resource capability varies over time unpredictably. Moreover, resources are globally distributed but there is no universal computing platform and most of resources are temporally available and managed by different domain rules. Therefore, uncertainty on the job execution quality particularly arise from resource faults. In this dissertation, we thus concentrate on the two types of resource faults in distributed computing environments such as \emph{failure} and \emph{fragmentation}, which make the balanced and optimal scheduling decision further complicated and deteriorate the reliability of a generated solution. These problems have been known as NP-hard. However the existing heuristics which have been proposed to provide fault tolerance are even static to fluctuated availability, are biased to a specific objective or are strongly dependent on the problem one attempts to deal with. To develop a new heuristic mechanism specialized for the fault tolerant job scheduling over unreliable resources, we build a novel meta-scheduling framework (MSF) based on self-organized genetic algorithm (SOGA) which has a promising ability to search near-optimal solutions based on the concept of Pareto optimality within the reasonable time and space complexity even though failures are highly unpredictable and frequently occur. The ultimate objective of the dissertation is to investigate its effectiveness and superiority on the aforementioned platform-level vulnerabilities such as failure and fragmentation. The primary contributions of the dissertation are as follows. This dissertation firstly proposes a self-organized genetic algorithm as the kernel process of the meta-scheduling framework. The proposed SOGA consists of two processes such as master and agents, where the master process sets the number of subpopulations a...
Advisors
Youn, Chan-Hyunresearcher윤찬현
Description
한국과학기술원 : 정보통신공학과,
Publisher
한국과학기술원
Issue Date
2013
Identifier
566051/325007  / 020065212
Language
eng
Description

학위논문(박사) - 한국과학기술원 : 정보통신공학과, 2013.8, [ x, 165 p. ]

Keywords

Job Scheduling; 과학 워크로드; 메타 휴리스틱; 분산 컴퓨팅; 유전 알고리즘; 자원 휘발성; Resource Brokering; Fault Tolerance; Job Faliure; Resource Fragmentation; Resource Volatility; Genetic Algorithm; Distributed Computing; Meta-Heuristic; Scientific Workload; 잡 스케줄링; 자원 브로커링; 결함 감내; 잡 실패; 자원 단편화

URI
http://hdl.handle.net/10203/197775
Link
http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=566051&flag=dissertation
Appears in Collection
ICE-Theses_Ph.D.(박사논문)
Files in This Item
There are no files associated with this item.

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0