VKernel progress update - 8 Jan 2006 (milestone reached!)

Matthew Dillon dillon at apollo.backplane.com
Wed Jan 10 15:33:26 PST 2007


:Will a virtual kernel be able to run another virtual kernel as one of
:its user processes?

    There are only minor roadblocks to being able to do so.  Once I get
    the VM system stabilized it shouldn't take long to make that work.

    Performance will suffer greatly until some optimizations are made,
    though.  In particular, the system call path, copyin/copyout paths,
    and page fault paths.

    The biggest limitation to these environments is simply going to be
    available system memory.  If you tell a virtual kernel that it has
    256M of main memory, then it will only really be happy if it can
    actually use 256M of main memory without thrashing the real system.

    If you create a virtual kernel inside a virtual kernel, the memory
    gets subdivided so the parent virtual kernel would need a rather
    significant reservation in order for the child to be able to do anything
    meaningful.  At least there isn't any duplication of data in memory
    (once the VK disk starts using direct I/O to prevent it from being
    cached by the parent kernel).

					-Matt
					Matthew Dillon 
					<dillon at backplane.com>





More information about the Kernel mailing list