Building a high-availability opensource datacenter with shared GPU

In recent years, the IT world has made great leaps forward in virtualization. From a situation where each computer was installed under a particular operating system, we've moved on to an era where a single system can host an infinite number of others.

This evolution, based on the sharing of machine resources, has considerably reduced hardware requirements, with the main criterion becoming the availability of processor, memory or disk resources. It has also enabled access to new capabilities, such as high application availability, at much lower cost. In some cases, this can also enable the implementation of machines based on other processors or different operating systems, such as Android on an ARM processor or MacOS.

To take advantage of all these new services, all you need to do is install these physical servers in the form of hypervisors, whose function will be to host virtual machines.