For years, Inttra, an e-commerce logistics provider to the world's largest cargo-shipping organizations, has been using virtualization on its back-end IBM mainframe and Citrix Systems servers in a secure environment. Now the Parsippany, N.J., company primarily uses IBM blade servers running virtual Linux machines. VMware's virtualization technology on an Intel platform powers this New Data Center infrastructure.
John Debenedette, Inttra's vice president of IT, says he believed he could keep a virtualized data-center environment secure while emulating established best practices. He's not ready, however, to risk running virtual Web servers outside his DMZ. Nor is he ready to allow virtual machines on the endpoints, which are harder to control.
"You can follow best practices on all of your virtual machines. But at the end of the day, you're putting a lot of trust in the virtual-machine platform layer itself," Debenedette says. "This layer - also called the hypervisor, the virtual kernel or virtual-machine monitor - sits between the hardware and all its device drivers, including the operating system, which puts it in a very authoritative position."
Security watchers have not confirmed any exploits at this layer; but virtual-machine-aware malware, such as RedPill, and virtual-machine rootkits, such as BluePill, are common. Debenedette rightfully frets about this new platform layer: It's a vector into which virtual-machine malware writers are trying to break, experts say.
In this virtual environment, effective security best practices are sorely needed. In addition to physical machines, virtual machines must be managed and secured. Network defenses must be tuned to watch for rogue traffic on them. And the virtual-machine layer must be built safely and defended from up-and-coming forms of attackware.
Virtual-machine best practices
In a survey of 707 Network World readers conducted in June, 36% of respondents - 250 respondents - said they realize virtualization has increased security risk. Of those, slightly more than half had deployed firewalls and segmented critical networks into virtual LANs, and another half had included virtual-machine traffic-awareness in their intrusion-detection sensors.
One-third of respondents seemed to grasp that the virtualization platform layer itself is vulnerable. The others did not believe virtual-machine platform vendors need to make security integral to their products.
Clearly, many enterprises are failing to apply even the most basic security policies for protecting their virtual servers.
Topping off that dangerous misstep, organizations are experiencing rogue and unmanaged virtual-machine creep -- the very thing virtualization tries to relieve in the hardware realm, consultants to Fortune 500 companies say.
"The problem is collectively known as virtual-machine sprawl," says Anil Desai, consultant and author of The Definitive Guide to Virtual Platform Management. "If virtual machines are built without IT's knowledge, it's tough even to know they exist on the network," he says.
Consultants report a widespread problem at client sites: "Software developers, intranet users, even users on data-center servers with too much privilege, are setting up virtual machines without IT's knowledge because they're easy to deploy and help get certain jobs done," Desai says.
Inttra's Debenedette says he doesn't understand this phenomenon. Any organization worth its salt should have locked down its data centers according to best practices, which would make actions such as launching a new virtual server something that would trigger alarms, he says. Enforcement of those best practices is what ultimately cuts down on scope creep.