I thought I had responded previously but if not then my apologies.
Our testing was on a real CentOS deployment and we received the best performance using Docker as opposed to not running Docker because, we understand, Docker distributes access to hardware resources more efficiently than would be the case when using a group of virtual machines.
If you refer to running a CentOS machine as a single complete instance, hence no shared or conflicting requirements from different users then I think you would be correct since this would make the entire machine available without the load created by Docker?
Its all back to the laws of Physics isn't it, the more you ask the machine to do the longer it takes?
Thanks for taking the time to comment.