Support a new type of virtual machine in oVirt destined for running a virtual machine with highest possible performance and performance metrics as close to bare metal as possible. By choosing this new High Performance virtual machine type, the virtual machine will be per-configured with a set of suggested and recommended configuration settings for reaching the best efficiency.

Before this feature was implemented in oVirt, configure a virtual machine to run with high performance workloads was not an easy straightforward mission to do and required the user to manually set the virtual machine as such by going over all settings and check what is relevant and how to configure it. In addition, live migration was usually not supported for such virtual machines due to pinning constraints. Furthermore, few required features essential for improving virtual machine performance were not supported at all by oVirt (for example: using huge pages, IO and Emulator threads pinning, CPU cache layer 3 support, USV and other devices disabling, headless mode, enable Multi Queues per Virtual Interface etc).
All these new features and more can now be leveraged to suggest one solution of the best recommended configuration according to virtual machine usage requirements.

We will explain why increasing performance will decrease flexibility of the virtual machine, discuss how to set the virtual machine's devices, pass through host CPU mode, io/cpu/emulator threads pinning typologies, paravirtualized random number generator, virtual NUMA nodes and NUMA pinning topology and more settings. We will also explain how to combine all settings together wisely, what are the requirements from the physical hosts running those virtual machines and what are the limitations. We will also discuss virtual machines' pinned hosts limitations

We will explain out solution for supporting migration for those high performance virtual machines which include pinning configuration, migration constraints to handle and hosts scheduler requirements. In addition we will talk about affinity rules for managing groups of high performance virtual machines and physical hosts.