Choosy: Max-Min Fair Sharing for Datacenter Jobs with Constraints
Title | Choosy: Max-Min Fair Sharing for Datacenter Jobs with Constraints |
Publication Type | Conference Paper |
Year of Publication | 2013 |
Authors | Ghodsi, A., Zaharia M., Shenker S. J., & Stoica I. |
Page(s) | 365-378 |
Other Numbers | 3434 |
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. |
URL | https://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 |