Virtualized workloads are known to require carefully crafted configuration and tuning, both at the host and at the guest level, if good performance are to be achieved. But, considering that this comes at the price of lack of both flexibility, efficiency and complex management, is it really worth it?
On the other end, modern Operatins System offer mechanisms for handling complex workloads dynamically, efficiently and without the need of cumbersome manual setup. But are the performance good enough?
This talk will try to provide some possible answers to the above questions, by showing the results of an extensive analysis of different combinations and levels of automatic and manual tuning of a virtualization server, when some of the most typical workloads and load conditions are considered.