Choosy: Max-Min Fair Sharing for Datacenter Jobs with Constraints

TitleChoosy: Max-Min Fair Sharing for Datacenter Jobs with Constraints
Publication TypeConference Paper
Year of Publication2013
AuthorsGhodsi, A., Zaharia M., Shenker S. J., & Stoica I.
Page(s)365-378
Other Numbers3434
Abstract

Max-Min Fairness is a flexible resource allocation mechanismused in most datacenter schedulers. However, an increasingnumber of jobs have hard placement constraints,restricting the machines they can run on due to specialhardware or software requirements. It is unclear how todefine, and achieve, max-min fairness in the presence ofsuch constraints. We propose Constrained Max-Min Fairness(CMMF), an extension to max-min fairness that supportsplacement constraints, and show that it is the only policysatisfying an important property that incentivizes usersto pool resources. Optimally computing CMMF is challenging,but we show that a remarkably simple online scheduler,called Choosy, approximates the optimal scheduler well.Through experiments, analysis, and simulations, we show

Acknowledgment

This work was partially supported by funding provided to ICSI through National Science Foundation grant CNS: 1161813 (“CSR: Medium: Limiting Manipulation in Data Centers and the Cloud”). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors or originators and do not necessarily reflect the views of the National Science Foundation.

URLhttps://www.icsi.berkeley.edu/pubs/networking/ICSI_choosymaxmin13.pdf
Bibliographic Notes

Proceedings of the 2013 EuroSys Conference, pp. 365-378, Prague, Czech Republic

Abbreviated Authors

A. Ghodsi, M. Zaharia, S. Shenker and I. Stoica

ICSI Research Group

Networking and Security

ICSI Publication Type

Article in conference proceedings