Supervision : Gaël THOMAS
Co-supervision : SENS Pierre
Efficient Virtualization of NUMA Architectures
While virtualization only introduces a negligible overhead on machines with few cores, this is not the case when the number of cores increases. We can find such computers with tens of cores in todays data centers dedicated to the cloud computing, a resource management model which relies on virtualization. These large multicore machines have a complex architecture, called Non Uniform Memory Access (NUMA). Achieving high performance on a NUMA architecture requires to wisely place application threads on the appropriate cores and application data in the appropriate memory bank.
In this thesis, we show how virtualization techniques modify the applications behavior by preventing them to efficiently place their data in memory. We show that the data misplacement leads to a serious performance degradation, up to 700%.
Additionally, we suggest a method which allows the Xen hypervisor to efficiently virtualize NUMA architectures by implementing a set of generic memory placement policies. With an evaluation over a set of 29 applications on a 48-cores machine, we show that the NUMA policies can multiply the performance of 9 applications by more than 2 and decrease the virtualization overhead below 50% for 23 of them.
Defence : 03/08/2018 - Site Jussieu 25-26/105
Jury members :
Rachid Guerraoui, Ecole Polytechnique Fédérale de Lausanne [Rapporteur]
Laurent Réveillère, Université de Bordeaux [Rapporteur]
Emmanuelle Encrenaz, Sorbonne Université
Vivien Quéma, Université Grenoble-Alpes
Pierre Sens, Sorbonne Université
Gaël Thomas, Télécom SudParis
- M. Mosli Bouksiaa, F. TRAHAY, A. Lescouet, G. Voron, R. Dulong, A. Guermouche, E. Brunet, G. Thomas : “Using differential execution analysis to identify thread interference”, IEEE Transactions on Parallel and Distributed Systems, vol. 30 (12), pp. 2866-2878, (Institute of Electrical and Electronics Engineers) (2019)
- G. Voron : “Efficient Virtualization of NUMA Architectures”, thesis, defence 03/08/2018, supervision Thomas, Gaël, rapporteurs : SENS Pierre (2018)
- M. Mosli Bouksiaa, F. TRAHAY, A. Lescouet, G. Voron, A. Guermouche, E. Brunet, G. Thomas : “Identifying thread interference with performance variation”, 14 pages (2017)
- G. Voron, G. Thomas, V. Quéma, P. Sens : “An interface to implement NUMA policies in the Xen hypervisor”, Twelfth European Conference on Computer Systems, EuroSys 2017, Belgrade, Serbia, pp. 15 (2017)
- G. Voron, G. Thomas, P. Sens, V. Quéma : “Optimisation mémoire dans une architecture NUMA : comparaison des gains entre natif et virtualisé”, Conférence en Parallélisme, Architecture et Système, (COMPAS'15), Lille, France (2015)