Abstract : Advanced Virtual Machines placement policies are evaluated either using limited scale in-vivo experiments or ad hoc simulator techniques. These validation methodologies are unsatisfactory. First they do not model precisely enough real production platforms (size, workload representativeness, etc.). Second, they do not enable the fair comparison of different approaches.
To resolve these issues, we propose VMPlaceS, a dedicated simulation framework to perform in-depth investigations and fair comparisons of VM placement algorithms. Built on top of SimGrid, our framework provides programming support to ease the implementation of placement algorithms and runtime support dedicated to load injection and execution trace analysis. It supports a large set of parameters enabling researchers to design simulations representative of a large space of real-world scenarios. We also report on a comparison using VMPlaceS of three classes of placement algorithms: centralized, hierarchical and fully-distributed ones.