About this list Date view Thread view Subject view Author view Attachment view

From: Herbert Poetzl (herbert_at_13thfloor.at)
Date: Tue 22 Mar 2005 - 11:36:17 GMT


On Tue, Mar 22, 2005 at 12:10:55PM +0800, Tony Mobily IMAP wrote:
> Hi,
>
> >>The Leap from Virtual Host to Virtual Machine
> >>XEN, virtual machine monitor-1
> >>Vserver <--- !!!
> >
> >Linux-VServer ;)
>
> Woooos... sorry!

in the age of vservers you have to be precise ;)

> >>I have already assigned the first one. For XEN, I already have an
> >>article. Please have a look here:
> >>
> >>http://www.freesoftwaremagazine.com/free_issues/issue_04/xen/
> >>(login: author, password:tnmafs)
> >
> >I skimmed over it, and I think it depends on what
> >you plan to do .. because:
> >
> >about 70% of the stuff 'explained' in this article
> >is absolutely true for QEMU, UML, Xen and Linux-VServer
> >they basically differ in the level of virtualization
> >in the following way:
> >
> > - QEMU (like VMware) allows to run arbitrary software
> > (including a kernel) for one arch on the same or
> > even a different arch by utilizing emulated hardware
> > and binary translation ...
> > - UML or Xen allows to run an operating system compiled
> > for the UML/arch or Xen/arch architecture on a (normal
> > or specialized) system of the same architecture, by
> > 'handling' special/restricted stuff ...
> > - Linux-VServer (like solaris slices) allow you to run
> > distributions of linux (all userspace) in an isolated
> > environment, sharing as much resources as possible,
> > nevertheless separated enough to run almost unmodified
> > copies of real servers
>
> OK.
>
> >so if you plan to make one big article about virtualization
> >in general with three or more subsections to introduce the
> >different techniques (QEMU, XEN, Linux-VServer), then I
> >have no problem writing some stuff for that ...
>
> OK. that would have been the optimal choice, if we didn't have the Xen
> article already...
>
> >OTOH, if the Xen article is published this way, it doesn't
> >make much sense to repeat all the virtualization stuff in
> >a Linux-VServer article, and writing an article about
> >Linux-VServer stating: "does basically the same, just a
> >little faster/more resource efficient, but you are not
> >able to have your own kernel (per guest)" is not really
> >a good way to do it ...
>
> True. However, I see Linux-VServer as a different "branch" of
> virtualization, where there is only one kernel running. The article on
> Xen (which I must admit is not very objective) focuses on
> virtualization, where Linux-VServer concentrates on "separation" (only
> one kernel is running).

let's call it 'isolation' (of course, separation leads
to isolation, or was it the other way round ;)

> Here is my idea:
>
> * We publish the article on Xen, taking out this part (which is a
> little too nasty and objective):

hmm, greetings from Freud?

> --------------------------------------------
> It is important not to confuse OS virtualization with so-called
> ?application virtualization?, a software technique that in effect
> ?bundles? all processes, threads and application related state for each
> different application hosted by an OS, into a virtual container.
> Application virtualization software vendors, such as Trigence, attempt
> to provide balanced performance to each virtual container, by applying
> application-specific policies to the OS scheduler. This achieves few of
> the benefits of true OS virtualization, the least of which is its
> inability to take advantage of new hardware features for
> virtualization, and consequently is not considered a serious contender
> in the data center
> ----------------------------------------------
>
> * I tell Eddy (who is writing the first article) to say, right at the
> end:

why not make some contact between 'Eddy' and myself
and we'll talk a little about the different things,
I have no problem with a Xen specific point of view
and as I said, they are pretty orthogonal IMHO ...

> QEMU, UML, Xen and Linux-VServer basically differ in the level of
> virtualization
> in the following way:
>
> - QEMU (like VMware) allows to run arbitrary software
> (including a kernel) for one arch on the same or
> even a different arch by utilizing emulated hardware
> and binary translation ...
> - UML or Xen allows to run an operating system compiled
> for the UML/arch or Xen/arch architecture on a (normal
> or specialized) system of the same architecture, by
> 'handling' special/restricted stuff ...
> - Linux-VServer (like solaris slices) allow you to run
> distributions of linux (all userspace) in an isolated
> environment, sharing as much resources as possible,
> nevertheless separated enough to run almost unmodified
> copies of real servers
>
> * You write an article dedicated to Linux-VServer, which represents in
> a way "the other side of virtualization" - we can call it "separation"
> not to confuse our readers. I will tell Eddy to use the word
> "separation" as well.
>
> So, the Focus will contain the following articles:
>
> 1) Introduction by Eddy (covering the bit on QEMU, UML, Xen and
> Linux-VServer )
>
> 2) XEN (taking out the "nasty" part). This article will only cover
> "virtualization"
>
> 3) Linux-VServer. This article will only cover "separation". You can
> say *whatever* you like (things like "Complete virtualization is a
> waste of resources", and so on :-) ). You could also explain how the
> separation is acheived technically.

> Would you be happy with this?

well, I'm not very happy about bashing other technologies,
especially if they are free and not directly related,
I would more prefer to paint a good picture of what is
possible with each technology ...

> >especially as I think QEMU, UML, Xen and Linux-VServer
> >are somewhat orthogonal, which can easily be seen if you
> >consider that:
> >
> > - Linux-VServer can run on UML and probably Xen
> > - UML and QEMU can be run on Linux-VServer and Xen
> > - Linux-VServer and probably Xen can be run in QEMU
> >
> >other combinations are thinkable too, and you could even
> >build several levels like this extreme example:
> >
> >PPC64 running QEMU emulating an i386 running a Linux
> >system which has one UML instance running Linux-VServer
> >which starts a QEMU (on i386) emulating some ARM ;)
>
> AAAAAAAAAAAHHHHHHHHHH!!!!!!!!!!!

but not that unrealistic ...

> >well, if you want to give some history/background, about
> >the development, then please go ahead, that's nothing I
> >could write objectively about .. btw, a good idea would be
> >to pay a visit to our irc channel (#vserver @ irc.oftc.net)
>
> OK, I will try :-)

great!

> >>Please let me know as soon as possible if you would be willing to
> >>write
> >>this article! (Grisha, in cc:, offered help as well!). The deadline is
> >>a little tight (10th of April), but a short article would be fine as
> >>well.
> >
> >I guess the community can do that easily ...
> >
> >btw, I would like to forward this email to the mailing list
> >(vserver_at_list.linux-vserver.org) if you do not have any
> >strong objections ...
>
> I don't have any objections at all. However, I think it would be better
> to decide what the Focus' outline should be, or things can get a little
> confusing (too much input can be dangerous if the foundations are not
> set...)

okay, so this is cc-ed to the ml then ;)

best,
Herbert

> Bye!
>
> Merc.
>
> ------------------------------------------------------------------------
> ----
> Tony Mobily
> Author of "Hardening Apache" (Apress)
> "...this book can save you..." -- Mitchell Pirtle, PHP Magazine 05/2004
_______________________________________________
Vserver mailing list
Vserver_at_list.linux-vserver.org
http://list.linux-vserver.org/mailman/listinfo/vserver


About this list Date view Thread view Subject view Author view Attachment view
[Next/Previous Months] [Main vserver Project Homepage] [Howto Subscribe/Unsubscribe] [Paul Sladen's vserver stuff]
Generated on Tue 22 Mar 2005 - 11:36:44 GMT by hypermail 2.1.3