On Thu, 2007-10-18 at 11:22 +0900, Isaku Yamahata wrote:
> On Wed, Oct 17, 2007 at 10:03:14AM -0600, Alex Williamson wrote:
> > * Increasing Xen page size (discussed a little at OLS)
>
> What motivates this? I don't know the discussion.
> I'm guessing the reason behind it.
> - catching up linux deafult page size chage
> - removing current pseudo physical addrese size limit
> Why not moving from 3 level p2m table to 4 level p2m table?
> anyother reason?
I think the main factors were address space and performance. A 4
level table might be an easier way to deal with the address space side.
> Some random ideas. I hope they will help starting discussions.
>
> * NUMA emu (something like Linux x86_64 NUMA_EMU)
> This would be convenient for developing NUMA support on
> non-NUMA box.
I like this idea, at least as a development feature.
> * scalability
> While some hard limit is there as Alex already mentioned,
> Does the current implementation of other part scale as we want?
> If no, which subsystem to rewrite?
Yes, another good area for investigation. I'm not aware of any real
data in this space.
> * debugging tools for xen vmm itself/tools stack/guest OS
> Currently we have Kexec/Kdump, xm dump-core, xenitp, and
> misc tools. Do we want any others?
> e.g. nested Xen or something like Linux 'crash /dev/kcore',
> Linux LTT, systemtap...
> * fast hypercall using gate page
> * super page/hugetlbfs support
> * PMU virtualization
Good, I added all these.
> * performance isolation
Can you describe what you mean by this?
I updated the wiki, feel free to make modification and add stuff there
on your own. Thanks,
Alex
--
Alex Williamson HP Open Source & Linux Org.
_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel
|
Home