Paper Abstract and Keywords |
Presentation |
2017-11-17 10:30
Multi-QoS Compliant Virtual Resource Recommendation Scheme Abu Hena Al Muktadir, Takaya Miyazawa, Pedro Martinez-Julia, Ved P. Kafle, Hiroaki Harai (NICT) NS2017-125 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
This paper proposes MTCRS, a Multi-Target Classification based automatic virtual resource Recommendation Scheme, aiming to reduce resource over-allocation by infrastructure providers (InPs) and to satisfy multiple quality-of-service (QoS) requirements of resource requests by virtual network operators (VNOs). The InP provides abstracted modular virtual resource information (as Capability Units (CUs)) and their usage prices to the subscribing VNOs. In MTCRS, the InP proactively classifies each type of CU (e.g. virtual machine, shared radio access technology, link bandwidth) according to various attributes (e.g. function, capability, location, energy consumption, price etc.) to enable multi-QoS compliant resource allocation. InP employs separate classifier for each type of CU. The classifiers deploy multi-target classification algorithms and are regularly trained with resource availability aware labeled training data for adaptability. MTCRS concurrently uses x classifiers to predict the class of x types of CUs demanded in a VNO request. InP recommends a set of x CUs to VNO, for each processed request. By assessing several algorithms, we determine Bayesian Classifier Chain method to be most suitable for our scheme. We numerically evaluate MTCRS w.r.t. prediction time and resource over-allocation. We achieved an average 0.4 milliseconds prediction time for a new request, and over-allocation of CPU and memory resources are reduced by about 25% and 33%, respectively, with better trained classifiers. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
virtual resource / multi-target classification / proactive resource management / virtual network operators / infrastructure providers / / / |
Reference Info. |
IEICE Tech. Rep., vol. 117, no. 303, NS2017-125, pp. 79-84, Nov. 2017. |
Paper # |
NS2017-125 |
Date of Issue |
2017-11-09 (NS) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
NS2017-125 |
|