This paper addresses multi-resource fair allocation: a fundamental research topic in cloud computing. To improve resource utilization under well-studied fairness constraints, we propose a new allocation mechanism call...This paper addresses multi-resource fair allocation: a fundamental research topic in cloud computing. To improve resource utilization under well-studied fairness constraints, we propose a new allocation mechanism called Dominant Resource with Bottlenecked Fairness(DRBF), which generalizes Bottleneck-aware Allocation(BAA) to the settings of Dominant Resource Fairness(DRF). We classify users into different queues by their dominant resources. The goals are to ensure that users in the same queue receive allocations in proportion to their fair shares while users in different queues receive allocations that maximize resource utilization subject to well-studied fairness properties such as those in DRF. Under DRBF, no user 1) is worse off sharing resources than dividing resources equally among all users; 2) prefers the allocation of another user; 3) can improve their own allocation without reducing other users' allocations; and(4) can benefit by misreporting their resource demands. Experiments demonstrate that the proposed allocation policy performs better in terms of high resource utilization than does DRF.展开更多
基金financial support of the Oversea Study Program of the Guangzhou Elite Project(GEP)supported by the National Natural Science Foundation of China under Grant 61471173Guangdong Science Technology Project(no:2017A010101027)
文摘This paper addresses multi-resource fair allocation: a fundamental research topic in cloud computing. To improve resource utilization under well-studied fairness constraints, we propose a new allocation mechanism called Dominant Resource with Bottlenecked Fairness(DRBF), which generalizes Bottleneck-aware Allocation(BAA) to the settings of Dominant Resource Fairness(DRF). We classify users into different queues by their dominant resources. The goals are to ensure that users in the same queue receive allocations in proportion to their fair shares while users in different queues receive allocations that maximize resource utilization subject to well-studied fairness properties such as those in DRF. Under DRBF, no user 1) is worse off sharing resources than dividing resources equally among all users; 2) prefers the allocation of another user; 3) can improve their own allocation without reducing other users' allocations; and(4) can benefit by misreporting their resource demands. Experiments demonstrate that the proposed allocation policy performs better in terms of high resource utilization than does DRF.
