SAN DIEGO – Virtualization may still be a new concept for many, but it’s apparently been around long enough to need some re-imagining. Or at least according to Arjan van de Ven, a fellow at Intel and director for the firm’s Pathfinding and Advanced Technology, who touted the a composable hypervisor as being a possible re-imagining tool.

“Virtualization is changing and what people need from virtualization is changing,” van de Ven said during a keynote address at this week’s North America Open Source Summit and Embedded Linux Conference in San Diego.

Van de Ven explained that in talking with customers he has seen five areas emerge as needing re-imagining in order to support evolving virtualization plans. These include a platform that is lightweight; one that is fast; something that can support high density workloads; that has quick start up; and one that is secure. However, the degrees of those needs remains in flux.

In terms of being lightweight, he explained that while this requirement came back “over-and-over as a main request,” what that meant differed based on where those virtualized loads were running. Those running on virtual machines (VMs) could support larger workloads than those running in a containerized or functions-as-a-service (FaaS) environment.

Speed is also difficult to nail down, he explained. “Everyone want’s fast, but what does fast mean,” van de Ven said.

Something more tangible is density, which is a significant issue as it pertains to the financial stability of a virtualized environment. “People want to use the same hardware to do more work,” van de Ven said of this requirement. “If Amazon ran each application on its own server it would not be in business.”

A quick boot time is also necessary to support microservices that might only run for a couple of seconds. “The world where you fired up a server or VM and let it run for three months is gone,” he said, explaining that applications need instantaneous support. “If it takes two to three minutes to start you don’t exist.”

And of course, security has become a baseline. “You can’t negotiate security,” he said. “You can negotiate density, you can negotiate start up, but you can’t negotiate security.”

So how can all of these requirements be supported?

Van de Ven explained that a composable hypervisor was one way to deal with these varying needs, pointing to Intel’s work with the recently launched rust-vmm hypervisor.

That open source project provides a set of common hypervisor components developed by contributing vendors that can provide a more secure, higher performance container technology designed for cloud native environments.

Van de Ven described rust-vmm as basically a composable hypervisor made up of different components that can be used as building blocks to support only the services or level of support needed. “You can put them together any way you want,” he explained. “FaaS might only use half or a third. If you want an enterprise hypervisor you can use all.”

He did note that the work around rust-vmm was still ongoing and not yet ready for production environments, but that the concept is needed in order to support where virtualization needs to go.

“Re-imagining virtualization is really about being composable, being domain specific while sharing as many blocks as you can,” van de Ven said.