DragonFly BSD’s Virtual Pace


DragonFly BSD continues its evolution with version 1.8, almost four years after being forked from the FreeBSD 4.x code base. New kernel virtualization features top a long list of system improvements over version 1.6, as the operating system looks ahead to broader commercial adoption.


Most notable on the new feature list is the addition of virtual
kernel support for the DragonFly BSD kernel. The virtual kernel allows for
a “system-in-a-box” running as a userland process. But virtual kernel support is different things to different people, said DragonFly BSD Project Leader Matthew Dillon. “Its uses are approximately the same uses that UML [User Mode Linux] is to Linux.


UML and virtual kernel support is used to create compartmentalized
whole-system environments and to greatly ease the development of kernel code
by reducing the engineering-test cycle time. Dillon noted that DragonFly BSD will use virtual kernel support primarily for reducing the engineering-test cycle time.


“The clustering and cache coherency management subsystems are going to be
very complex, and it is literally impossible to develop them without a
fast-booting virtual kernel environment,” Dillon told internetnews.com.


Though DragonFly BSD now has virtual kernel support, it’s unclear at this
point whether a more robust virtualization effort that includes VMware or
Xen virtual images is imminent.


Dillon said that a number of people in the project have expressed an
interest in doing a virtual image, though it is not on his agenda, which is focused on clustering work. In other words, Dillon isn’t particularly enthusiastic about either VMWare or Xen.


“Hardware virtualization in general seems to be the catch-phrase of the day,
and it will be interesting to see how it develops over time,” Dillon said. “But I don’t see
much practical use for it other then to glue disparate operating systems
together in the consumer realm (e.g. run Linux and Windows and Mac OS on the
same machine at the same time).

“I far prefer virtualization
systems that know they are running in a virtualized environment (closer to
UML than to Xen or VMWare), because such systems can theoretically be
optimized far better in the long run.”


While DragonFly BSD does not yet have a major enterprise following like
other BSDs (most notably FreeBSD) and Linux, the time for broader
commercial adoption may not be very far off.


DragonFly BSD’s user base is currently large enough to keep the project
honest but small enough to allow it to continue development toward the
project’s goals, said Dillon.


“It seems to be slanted more towards home use than the enterprise, at least
for now, but there is a clear demand for the type of
native-to-the-operating-system machine clustering that we intend to
provide,” Dillon said. “We are not a kernel coding group at our core — not
well suited to providing commercial support ourselves, nor would I want to
get into that business.”


“If DragonFly gets to that point, commercial support would be a business
unto itself, like Red Hat, separate from core development,” Dillon added.


With the 1.8 release now available, the project can focus on version
2.0


“DragonFly is entering the second phase of the project, where the actual
cache coherency management and clustering subsystems are finally starting to
be developed,” Dillon explained. “Other aspects of the system are also well
positioned for a big push by other developers, in particular the SMP work.


“I don’t know how much of this will make it into 2.0, but I hope to have at
least one big-ticket item to show off in the 2.0 release.”

News Around the Web