As I write this, VMware is wrapping up its annual VMworld conference in Las Vegas, leaving attendees to ponder the significance of what VMware execs describe as the company's bold new direction. According to CEO Paul Maritz, VMware is building nothing less than a "virtual datacenter OS" -- VDC OS for short -- that will eventually make traditional operating systems "all but disappear."
If that sounds like shameless hype, consider the source. With the price tag for basic virtualization functionality now essentially nil, VMware has little choice but to couch its expensive enterprise product line in increasingly grandiose terms. Virtualization itself is old hat. What VMware offers is "virtual infrastructure" -- a fully integrated line of platforms and management tools for virtualized environments.
In some respects, VDC OS is merely the next rung on the VMware hype ladder. After all, if the press is predicting that OS vendors will put you out of business, preemptively proclaiming the death of the traditional OS is a decent tactic. If you look past the marketing, however, VDC OS is actually a fascinating concept. If VMware manages to achieve half of what it promises, it could have significant implications for application developers and customers alike.
An OS for the virtual datacenter
According to VMware execs, VDC OS will not be a product as such. Instead, it is an umbrella concept covering a range of capabilities that VMware will build into the next generation of its Virtual Infrastructure products.
Neither will VDC OS take on all of the responsibilities of existing operating systems, such as Linux, Unix, or Windows. "It has many parallels with an OS, in the sense that it has APIs and services," says Maritz, "but it is not a traditional OS."
Instead, VDC OS is literally an operating system for the virtual datacenter. The idea is to build services and APIs that make it easier to provision and allocate resources for applications in an abstract way.
In VMware's idealized datacenter, applications don't exist in the traditional sense as processes running on specific CPUs, each with its own dedicated memory and storage space. Instead, they are reduced to "application workloads," defined by the malleable but finite quantities of compute resources needed to execute their individual roles.