Cannot take my allocation

Enforcing fairness by considering demand and payment in clouds

Gyuyeong Kim, Wonjun Lee

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Cloud data centers provide cost-effective and elastic resource management environments through multi-tenancy. However, we observe that existing policies cause unfair bandwidth allocation. A max-min based policy allocates lower bandwidth to a tenant whose the size of flows is mostly large, and DCTCP falls short of high utilization and minimum performance guarantee. Moreover, commodity clouds are vulnerable to denial of service (DoS) attacks, which can be attempted by a malicious tenant. The aforementioned deficiencies decrease the motivation to use clouds and revenue of cloud providers. To address the problem, we propose Spiderweb, a fair network sharing mechanism for cloud data centers, which allocates bandwidth to tenants in proportion to the amount of demand and payment during oversubscription. Our proposal delivers both a cloud-specialized fair network sharing environment and robustness against DoS attacks. We show that our proposal meets the desirable sharing properties including proportional fairness, performance isolation, and high utilization through analytical allocation scenarios.

Original languageEnglish
Title of host publication2013 4th International Conference on the Network of the Future, NoF 2013
PublisherIEEE Computer Society
DOIs
Publication statusPublished - 2013 Jan 1
Event2013 4th International Conference on the Network of the Future, NoF 2013 - Pohang, Korea, Republic of
Duration: 2013 Oct 232013 Oct 25

Other

Other2013 4th International Conference on the Network of the Future, NoF 2013
CountryKorea, Republic of
CityPohang
Period13/10/2313/10/25

Fingerprint

Bandwidth
Frequency allocation
Costs
Denial-of-service attack

ASJC Scopus subject areas

  • Computer Networks and Communications

Cite this

Kim, G., & Lee, W. (2013). Cannot take my allocation: Enforcing fairness by considering demand and payment in clouds. In 2013 4th International Conference on the Network of the Future, NoF 2013 [6724514] IEEE Computer Society. https://doi.org/10.1109/NOF.2013.6724514

Cannot take my allocation : Enforcing fairness by considering demand and payment in clouds. / Kim, Gyuyeong; Lee, Wonjun.

2013 4th International Conference on the Network of the Future, NoF 2013. IEEE Computer Society, 2013. 6724514.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Kim, G & Lee, W 2013, Cannot take my allocation: Enforcing fairness by considering demand and payment in clouds. in 2013 4th International Conference on the Network of the Future, NoF 2013., 6724514, IEEE Computer Society, 2013 4th International Conference on the Network of the Future, NoF 2013, Pohang, Korea, Republic of, 13/10/23. https://doi.org/10.1109/NOF.2013.6724514
Kim G, Lee W. Cannot take my allocation: Enforcing fairness by considering demand and payment in clouds. In 2013 4th International Conference on the Network of the Future, NoF 2013. IEEE Computer Society. 2013. 6724514 https://doi.org/10.1109/NOF.2013.6724514
Kim, Gyuyeong ; Lee, Wonjun. / Cannot take my allocation : Enforcing fairness by considering demand and payment in clouds. 2013 4th International Conference on the Network of the Future, NoF 2013. IEEE Computer Society, 2013.
@inproceedings{6fcf12a88f44496fb6573e6936be7a9a,
title = "Cannot take my allocation: Enforcing fairness by considering demand and payment in clouds",
abstract = "Cloud data centers provide cost-effective and elastic resource management environments through multi-tenancy. However, we observe that existing policies cause unfair bandwidth allocation. A max-min based policy allocates lower bandwidth to a tenant whose the size of flows is mostly large, and DCTCP falls short of high utilization and minimum performance guarantee. Moreover, commodity clouds are vulnerable to denial of service (DoS) attacks, which can be attempted by a malicious tenant. The aforementioned deficiencies decrease the motivation to use clouds and revenue of cloud providers. To address the problem, we propose Spiderweb, a fair network sharing mechanism for cloud data centers, which allocates bandwidth to tenants in proportion to the amount of demand and payment during oversubscription. Our proposal delivers both a cloud-specialized fair network sharing environment and robustness against DoS attacks. We show that our proposal meets the desirable sharing properties including proportional fairness, performance isolation, and high utilization through analytical allocation scenarios.",
author = "Gyuyeong Kim and Wonjun Lee",
year = "2013",
month = "1",
day = "1",
doi = "10.1109/NOF.2013.6724514",
language = "English",
booktitle = "2013 4th International Conference on the Network of the Future, NoF 2013",
publisher = "IEEE Computer Society",

}

TY - GEN

T1 - Cannot take my allocation

T2 - Enforcing fairness by considering demand and payment in clouds

AU - Kim, Gyuyeong

AU - Lee, Wonjun

PY - 2013/1/1

Y1 - 2013/1/1

N2 - Cloud data centers provide cost-effective and elastic resource management environments through multi-tenancy. However, we observe that existing policies cause unfair bandwidth allocation. A max-min based policy allocates lower bandwidth to a tenant whose the size of flows is mostly large, and DCTCP falls short of high utilization and minimum performance guarantee. Moreover, commodity clouds are vulnerable to denial of service (DoS) attacks, which can be attempted by a malicious tenant. The aforementioned deficiencies decrease the motivation to use clouds and revenue of cloud providers. To address the problem, we propose Spiderweb, a fair network sharing mechanism for cloud data centers, which allocates bandwidth to tenants in proportion to the amount of demand and payment during oversubscription. Our proposal delivers both a cloud-specialized fair network sharing environment and robustness against DoS attacks. We show that our proposal meets the desirable sharing properties including proportional fairness, performance isolation, and high utilization through analytical allocation scenarios.

AB - Cloud data centers provide cost-effective and elastic resource management environments through multi-tenancy. However, we observe that existing policies cause unfair bandwidth allocation. A max-min based policy allocates lower bandwidth to a tenant whose the size of flows is mostly large, and DCTCP falls short of high utilization and minimum performance guarantee. Moreover, commodity clouds are vulnerable to denial of service (DoS) attacks, which can be attempted by a malicious tenant. The aforementioned deficiencies decrease the motivation to use clouds and revenue of cloud providers. To address the problem, we propose Spiderweb, a fair network sharing mechanism for cloud data centers, which allocates bandwidth to tenants in proportion to the amount of demand and payment during oversubscription. Our proposal delivers both a cloud-specialized fair network sharing environment and robustness against DoS attacks. We show that our proposal meets the desirable sharing properties including proportional fairness, performance isolation, and high utilization through analytical allocation scenarios.

UR - http://www.scopus.com/inward/record.url?scp=84894193427&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84894193427&partnerID=8YFLogxK

U2 - 10.1109/NOF.2013.6724514

DO - 10.1109/NOF.2013.6724514

M3 - Conference contribution

BT - 2013 4th International Conference on the Network of the Future, NoF 2013

PB - IEEE Computer Society

ER -