Hosting business-critical applications on virtualized servers means that reliability is deeply important. If a hardware failure brings the machine to a halt, all of the services hosted on that machine go doen as well.

Isolating memory errors using the Machine Check Architecture, in theory, means that failures can be contained. In practice, how does it work? According to Principled Technologies it works really well. A June 2011 test report from Principled Technologies shows that forced memory errors on a system using a Intel Xeon E7 and VMware vSphere were caught and isolated to only the VM that suffered the error.
Want to learn more? Check out the test report from Principled Technologies, “Intel and VMware Deliver Reliability: Simulated RAM error isolated to single VM.”