Open Access System for Information Sharing

Login Library

 

Thesis
Cited 0 time in webofscience Cited 0 time in scopus
Metadata Downloads

QoS를 지원하는 클라우드 컴퓨팅 자원 할당 기법

Title
QoS를 지원하는 클라우드 컴퓨팅 자원 할당 기법
Authors
차요중
Date Issued
2014
Publisher
포항공과대학교
Abstract
Cloud computing resource allocation is allocating the tasks that clients need to each cloud computing resources efficiently. We applied the concept ‘utility’. Utility means how much the clients satisfy about the given allocation. Utility value is calculated using the sum of total execution time of tasks and the client’s total cost of using cloud computing resources. In this paper, we applied QoS(Quality of Service) concept by dividing clients into several groups. Clients in same group have similar utility values and clients in group of higher level have better utility values than clients in lower level group. We propose a method to divide the level of a service using level of clients in allocation algorithm. In this paper the proposed allocation algorithm consists of two phases. The first phase is initial allocation phase. In this phase, tasks are allocated to minimize the deviation of utilities in same group and divide the utility values of different groups. The second phase is swap phase. In this phase, allocated tasks are swapped in calculated ratio to reduce the total execution time maintaining the deviation of system. Total utility of the system is improved by second phase by the decrease of total execution time. We compared simulation results with game-theoretic method. The game-theoretic method had randomly diffused utility values, but proposed heuristic swap method had standardized utility values.
URI
http://postech.dcollection.net/jsp/common/DcLoOrgPer.jsp?sItemId=000001679443
http://oasis.postech.ac.kr/handle/2014.oak/2230
Article Type
Thesis
Files in This Item:
There are no files associated with this item.

qr_code

  • mendeley

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Views & Downloads

Browse