On Tue, Dec 12, 2006 at 02:18:01PM +0200, Pasi Kärkkäinen wrote:
On Tue, Dec 12, 2006 at 12:54:31PM +0300, Nikita V. Youshchenko wrote:
What about idea not to *add* pae versions of xen kernels, but to *replace*
non-pae versions with pae versions?
Rationale:
- not increase in archive size or linux-2.6 package build time,
- this will improve compatimility with other distros (consider scenario
when running FC or RHEL in domU; these distros do ship only pae xen
kernels, according to
http://lists.alioth.debian.org/pipermail/pkg-xen-devel/2006-December/000998.html)
http://fedoraproject.org/wiki/FedoraXenQuickstartFC6
Any x86_64, or ia64 CPU is supported for running para-virtualized guests.
To run i386 guests requires a CPU with the PAE extension.
So FC6 is PAE only, and latest update to FC5 also switched to PAE xen
kernel.
RHEL5 will be PAE only too.
I don't know what Suse uses..
https://www.redhat.com/archives/rhelv5-beta-list/2006-December/msg00068.html
the decision to only support PAE capable hosts was made at the end of the
FC5 cycle. The majority (if not all) of server in customer
datacenters/environments are PAE capable today and the only edge case for
non-PAE support would have been older laptops which do not yet have PAE
capable processors. It also would have been an additional burden for QA/QE
to test/certify older non-PAE capable servers. As the use case for Xen is
certainly geared towards servers and not laptops this made a lot of sense.
-- Pasi
^
. .
Linux
/-\
Choice.of.the
.Next.Generation.