On 3 Jun 2000, David Wragg wrote:
> James Sutherland <jas88@cam.ac.uk> writes:
> > This is, IMHO, quite an attractive idea: a fully hot-swappable system,
> > where any failed component can be replaced without any downtime.
>
> Many big systems (historically and presently) with some degree of
> hardware fault-tolerance have also provided for partitioning, either
> through software, hardware, or a bit of both. One reason is that even
> if you eliminate downtime due to hardware problems, you are still left
> with downtime due to software; partitioning adds a lot of flexibility
> in testing and deploying changes such as kernel bug fixes.
Indeed; partitioning is rather a useful feature in many ways...
> Much of the kernel support needed for hot-swappable processors and
> memory is the same as the support needed to cooperate with some kind
> of hypervisor layer that manages software partitioning.
Not quite - this applies to dynamic partitioning. Static partitioning
would present the "client" OS with a perfectly normal situation: x CPUs, y
Gb of RAM. Both dynamic partitioning and hot-swap hardware would be
useful, though, and would tend to go together.
> So rather than putting full support for hot swappable processors and
> memory into the kernel, it might be better to add minimal support to
> the kernel, and then develop a separate hypervisor layer to sit
> underneath the kernel.
Perhaps; there are certainly good reasons for wanting a hypervisor of some
sort. This is a rather bigger undertaking than providing a syscall to make
a CPU enter "cli; hlt", though...
James.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Jun 07 2000 - 21:00:17 EST