Cleaning up user confusion about different virt modes in Xen

Cleaning up user confusion about different virt modes in Xen

[Lars Kurth]

Folks,

have a look at 
http://brendangregg.com/blog/2014-05-07/what-color-is-your-xen.html ... 
It's a well made case about the confusion our terminology has created in 
user-land. Admittedly, it was a lot worse before we came up with the 
virtualization spectrum. But since then ARM created extra complexity.

The article makes some very good points, and a sensible solution, but 
also throws up new problems. Maybe it is time to clean up the confusion 
in terminology, use it consistently and fix our documentation.

But before we consider this, I wanted put the article up for debate

Best Regards
Lars

Re: Cleaning up user confusion about different virt modes in Xen

[George Dunlap]

On Thu, May 8, 2014 at 12:02 PM, Lars Kurth <lars.kurth@xen.org> wrote:
> Folks,
>
> have a look at
> http://brendangregg.com/blog/2014-05-07/what-color-is-your-xen.html ... It's
> a well made case about the confusion our terminology has created in
> user-land. Admittedly, it was a lot worse before we came up with the
> virtualization spectrum. But since then ARM created extra complexity.
>
> The article makes some very good points, and a sensible solution, but also
> throws up new problems. Maybe it is time to clean up the confusion in
> terminology, use it consistently and fix our documentation.
>
> But before we consider this, I wanted put the article up for debate

Yeah; my original purpose in writing the articles was to help people
understand what PVH mode was going to be.  But I think that my
original post was a bit developer-centric: I like knowing about all
the gory details behind things, but customers want things to Just
Work.

The thing is, from a customer point of view, there are still only 2
choices: HVM or PV.  If you boot a Linux guest in HVM mode, it will
automatically use as much PV as it is capable of; the user doesn't
need to know anything to take advantage of it.  The main thing is just
to know that we continue to make improvements to Linux, so that newer
versions should be faster than older ones. :-)

I think long-term, PVH will actually end up the default no matter
what, unless specified otherwise.  Our goal is that the only
difference between HVM and PVH on the Xen side will be whether there
is a qemu running, emulating the motherboard, PCI devices, &c.  If you
boot a kernel in HVM mode, and it's PVH-capable, it may be able to
request qemu exit and switch into full PVH mode.  And when booting PV,
the toolstack will check to see if the kernel is PVH-capable, and if
so, will simply boot in PVH mode unless otherwise specified.

So from the user's point of view, there will still basically 2 modes
you need to know: HVM (starting with emulated hardware), and PV (no
emulated hadware).  Everything else will be "advanced configuration
options" for people who are really keen. :-)

 -George

Re: Cleaning up user confusion about different virt modes in Xen

[Lars Kurth]

No views then?
In this case, I think it makes sense to start using Brendan's 
terminology in wiki docs
Also see http://brendangregg.com/blog/2014-05-09/xen-feature-detection.html
Lars

On 08/05/2014 12:02, Lars Kurth wrote:
> Folks,
>
> have a look at 
> http://brendangregg.com/blog/2014-05-07/what-color-is-your-xen.html 
> ... It's a well made case about the confusion our terminology has 
> created in user-land. Admittedly, it was a lot worse before we came up 
> with the virtualization spectrum. But since then ARM created extra 
> complexity.
>
> The article makes some very good points, and a sensible solution, but 
> also throws up new problems. Maybe it is time to clean up the 
> confusion in terminology, use it consistently and fix our documentation.
>
> But before we consider this, I wanted put the article up for debate
>
> Best Regards
> Lars
>
>