IT managers well-versed in mainframe virtualization might expect smooth sailing when implementing virtualization tools for x86-based servers. But they're quickly finding unexpected challenges because x86 virtualization is nowhere near as mature as the mainframe virtualization tools that evolved over the past four decades, says a Saugatuck Technology analyst who is researching virtualization.

“The large shops start with the impression that it will be easy, because they think they know what they’re doing,” says Charlie Burns, whose report - The Many Faces of Virtualization: Understanding a New IT Reality - was published this week. “The problem is it’s different. They start to find out pretty quick that the things handled automatically by the virtualization mechanisms in the mainframe either aren’t there, or are less mature and robust.”

Virtualized servers are in some ways harder to manage than the traditional environment in which each server hosts a single application, because they contain a layer of abstraction between the operating system and hardware. When something goes wrong, this layer of abstraction makes it difficult to identify malfunctioning devices, according to Burns.

Beyond troubleshooting limitations, Burns says IT departments will run into difficult management challenges if they attempt to run more than about five operating system images on a single server. Mainframe virtualization tools can comfortably run hundreds of virtual servers on a single piece of hardware, Burns says.

But hypervisor software used for x86 servers today falls short of mainframe virtualization when it comes to balancing the requirements of workloads vs. performance, he says.

IT managers Burns interviewed while doing his research have discovered that there are different rules and best practices for operating virtualized x86-based servers compared to “real” physical servers. If you want a physical server to run faster, you give it more memory, Burns says. But this isn’t always the right move with virtual servers.

“In a virtualized environment, you might get the exact opposite effect,” Burns says. “If you increase the size of the virtual storage amount in that virtual image, you might cause the whole thing to slow down. There are things you relearn for a new server environment.”