Performance analysis and models for collocated VMs running on multi-core physical machines

Organized by: 

Arnaud Legrand


Jonathan Rouzaud Cornabas


During this presentation, I will first do a scope of my previous and on-going works.

Then, I will present on-going work around the performance evaluation and modeling of Cloud platforms and applications. I will present an in-depth analysis of the performance of collocated VMs. By running my experiments on the Grid’5000 testbed, I was able to evaluate 2 processor families for a total of 6 different processor models. I have systematically explored the effect of collocation by testing all the different VCPU to CPU mapping while taking into account micro-architectural components (shared caches and NUMA nodes). I also explored the effect of multi-core virtual machines. To execute such experiment, I have used and extended a toolkit Execo that performs automatic and reproducible experiments.

Based on these experiments, I evaluate 8 lightweight performance models and observe that the virtual machine performance can be accurately predicted using a model that takes into account the number of VMs on the core and on the related NUMA node (with less than 8\% error). Finally, we validate our models on several processors and on both single and multi-(virtual)-cores VM. Using this model, we can increase the accuracy of the virtualization layer of the general purpose distributed system simulator SimGrid and improve it’s usability to simulate (HPC) Clouds. These results may also be used to improve VM placement algorithms.