Dynamic replica management scheme based on cost-deadline discipline in grids그리드에서의 cost-deadline 정책에 근거한 동적 replica 관리 기법

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 410
  • Download : 0
Distributed scientific and engineering applications often require access to large amounts of data (terabytes or petabytes). Access to distributed data is typically as important as access to distributed computational resources. Future applications also require widely distributed access to data (For example, access in many places by many people, virtual collaborative environments, etc.). There are several Data Grid middlewares (SRB, Globus Data Grid Support, SRM, etc.) to attempt to identify, prototype, and evaluate the key technologies required to support such scientific and engineering collaborations. As one of major components of Data Grid middleware, replication can be used to reduce access latency, improve data availability, and increase robustness, scalability and performance for distributed applications. Yet the scientific applications want guarantees that data will be available when they need it - at least with high probability. It is often desirable to create remote read-only copies (replicas) of data elements (files). A well-known technique for improving availability in unreliable systems is replication. If multiple copies of data exist on independent nodes, then the chances of at least one copy being accessible are increased. Aggregate data access performance will also tend to increase, and total network load will tend to decrease, if replicas and requests are reasonably distributed. Yet while replication has advantages, it also has significant costs. We need a mechanism for creating replicas that allows us to meet availability and performance goals. We describe here a mechanism with these properties. In our approach, we define replica location service as a S/W component that determines how many replicas of any file are needed to maintain desired availability and where are the most suitable locations under the given cost (and deadline) discipline. Another service provided in the data grid middleware is replica selection. Replica selection is the ...
Advisors
Youn, Chan-Hyunresearcher윤찬현researcher
Description
한국정보통신대학원대학교 : 공학부,
Publisher
한국정보통신대학교
Issue Date
2003
Identifier
392243/225023 / 020014059
Language
eng
Description

학위논문(석사) - 한국정보통신대학원대학교 : 공학부, 2003, [ vii, 61 p. ]

Keywords

Cost-Deadline; Management Scheme

URI
http://hdl.handle.net/10203/55199
Link
http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=392243&flag=dissertation
Appears in Collection
School of Engineering-Theses_Master(공학부 석사논문)
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