Ali Ghodsi and Matei Zaharia and Benjamin Hindman and Andrew Konwinski and Scott Shenker and Ion Stoica

EECS Department, University of California, Berkeley

Technical Report No. UCB/EECS-2011-18

March 6, 2011

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2011/EECS-2011-18.pdf

We consider the problem of fair resource allocation in a system containing different resource types, where each user may have different demands for each resource. To address this problem, we propose Dominant Resource Fairness (DRF), a generalization of max-min fairness to multiple resource types. We show that DRF, unlike other possible policies, satisfies several highly desirable properties. First, DRF incentivizes users to share resources, by ensuring that no user is better off if resources are equally partitioned among them. Second, DRF is strategy-proof, as a user cannot increase her allocation by lying about her requirements. Third, DRF is envy-free, as no user would want to trade her allocation with that of another user. Finally, DRF allocations are Pareto efficient, as it is not possible to improve the allocation of a user without decreasing the allocation of another user. We have implemented DRF in the Mesos cluster resource manager, and show that it leads to better throughput and fairness than the slot-based fair sharing schemes in current cluster schedulers.

---

BibTeX citation:

@techreport{Ghodsi:EECS-2011-18,
    Author= {Ghodsi, Ali and Zaharia, Matei and Hindman, Benjamin and Konwinski, Andrew and Shenker, Scott and Stoica, Ion},
    Title= {Dominant Resource Fairness: Fair Allocation of Multiple Resource Types},
    Year= {2011},
    Month= {Mar},
    Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2011/EECS-2011-18.html},
    Number= {UCB/EECS-2011-18},
    Abstract= {We consider the problem of fair resource allocation in a system containing different resource types, where each user may have different demands for each resource. To address this problem, we propose Dominant Resource Fairness (DRF), a generalization of max-min fairness to multiple resource types. We show that DRF, unlike other possible policies, satisfies several highly desirable properties. First, DRF incentivizes users to share resources, by ensuring that no user is better off if resources are equally partitioned among them. Second, DRF is strategy-proof, as a user cannot increase her allocation by lying about her requirements. Third, DRF is envy-free, as no user would want to trade her allocation with that of another user. Finally, DRF allocations are Pareto efficient, as it is not possible to improve the allocation of a user without decreasing the allocation of another user. We have implemented DRF in the Mesos cluster resource manager, and show that it leads to better throughput and fairness than the slot-based fair sharing schemes in current cluster schedulers.},
}

EndNote citation:

%0 Report
%A Ghodsi, Ali 
%A Zaharia, Matei 
%A Hindman, Benjamin 
%A Konwinski, Andrew 
%A Shenker, Scott 
%A Stoica, Ion 
%T Dominant Resource Fairness: Fair Allocation of Multiple Resource Types
%I EECS Department, University of California, Berkeley
%D 2011
%8 March 6
%@ UCB/EECS-2011-18
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2011/EECS-2011-18.html
%F Ghodsi:EECS-2011-18