Xen, an open-source project with growing ties to Red Hat, Novell and Hewlett-Packard, is emerging as the leading contender for providing open-source virtualization for the Linux environment.
Red Hat, Novell To Package Xen
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Unlike higher level virtualization mechanisms like FreeBSD jails and Solaris Containers, Xen does not scale across multiple processors. Otherwise, it would seem like an ideal candidate for hosting database servers, available for a select few platforms, on a system which predominantly runs another operating system.
I’m certainly looking forward to the addition of AMD64 support.
Is it totally like Vmware will it allow me to run multiple linux distros/windows/FreeBSd
No. this is not virtualisation unlike vmware.however windows support is being worked upon
“Unlike higher level virtualization mechanisms like FreeBSD jails and Solaris Containers,”
perhaps you got this all wrong. jails are very different from what xen manages to do. it uses a mechanism to host multiple operating systems unlike jails(solaris containers are more similar to UML in Linux – perhaps you want bochs[bochs.sf.net) and the xen is being proposed for inclusion in mainline linux 2.6 to run linux without modifications. however it doesnt have a freebsd port yet though it is known to work on netbsd. plan is to support freebsd and plan 9 OS in the next few revisions
I believe that Xen requires modifications to the kernel to run. Being based from University research the developers have had access to the windows kernel via the MS shared source inititive and have built a version that works, the only problem is last time I checked was that they couldn’t distribute it.
So yes it runs windows… but it isn’t the same kind of application that VMWare Workstation is.
However it will be a concern to VMWare Inc as it will compete for the virtualisation of open source operating systems.
However it will be a concern to VMWare Inc as it will compete for the virtualisation of open source operating systems.
—-
thats putting it very mildly. if this can run windows without modifications out of the box it will just smash vmware and MS virtual PC because its going to be integrated in linux 2.6 in a couple of revisions and already supports netbsd..
I believe that Xen requires modifications to the kernel to run. Being based from University research the developers have had access to the windows kernel via the MS shared source inititive and have built a version that works, the only problem is last time I checked was that they couldn’t distribute it.
If the Xen developers have had access to MS shared source, wouldn’t it be dangerous to include Xen in the kernel?
These developers would be tainted with MS-technology and could open up for lawsuits by Microsoft.
”
If the Xen developers have had access to MS shared source, wouldn’t it be dangerous to include Xen in the kernel?
”
xen is a seperate architecture which can host other operating systems. I dont know whether MS shared source was used. even if it was the linux host is completely independant on the windows host so not likely
From what I understood of the Xen project it requires a modified kernel in the guest (virtual) operating system, apart from kernel support from the host (physical) machine.
So you’d need the modified windows kernel rewritten by the university to run windows in xen. The very same software the uni is forbidden to distribute.
Shared source is like that: you can play around, but in the end your work goes into nothing, since you can’t distribute anything.
Xen is basically a hypervisor technology like the one IBM uses in it’s mainframe and Pseries (AIX) offerings to run multiple OS instances in ‘virtual machines’ or logical partitions.
“A hypervisor is a software layer that implements viriual
machines having the same instruction-set architecture as the
hardware on which the hypervisor executes. Because the virtual machine’s instruction-set architecture is indistinguishable from the bare hardware, software run on a virtual machine cannot tell whether a hypervisor is present.”
So once the virtualization layer is good enough windows should run on it without modification. Of course it’s easier just to patch the kernel which just shows the strenght of open source.
For more info :
“CPU Resource Distribution by POWER Hypervisor and Partition Load Manager” ( http://www.redbooks.ibm.com/abstracts/tips0427.html?Open )
“Hypervisor based fault tolerance” ( http://www.pdos.lcs.mit.edu/6.824/papers/hyperfault95.pdf )
Xen runs WINDOWS XP and Linux.
However Windows XP isnt released yet due to licensing issues. Its a pure microsoft decision if they want Windows XP to run side by side with Xen.
Unlike higher level virtualization mechanisms like FreeBSD jails and Solaris Containers
This is a common misconception among BSD users: Jails and containers are not virtualizations, as you do not run a different instance of the kernel. Inside the jail you are completely tied to the original kernel, or in other words: there is simply no virtualization of the system at all.
Sorry, exclude Solaris Containers.
Re: Anonymous (IP: 61.95.184.—)
No. this is not virtualisation unlike vmware.however windows support is being worked upon
I don’t want to be dragged into a semantic argument. Jails are a mechanism by which a system can be made to function as multiple virtual systems, albeit managed by the same kernel image. As long as you know what I’m talking about there’s no need to debate the semantics of the word “virtualization”, that’s simply a waste of time.
perhaps you got this all wrong. jails are very different from what xen manages to do.
Indeed, however this article was talking about Xen’s use as a means to create virtual server instaces on Linux. While they do talk about using Windows, the implecation, at least in my mind, was running Linux on top of Linux.
it uses a mechanism to host multiple operating systems unlike jails(solaris containers are more similar to UML in Linux
Wrong. Solaris containers lie somewhere inbetween jails and UML. While containers run a seperate kernel instance, these kernel instances communicate directly with each other entirely in kernel context, rather than running as a user mode process. It is in this way that Solaris containers can span multiple processors, rather than being bound to just a single CPU.
Furthermore, the performance of UML is abominable in comparison to Solaris Containers. Since Solaris Containers allow multiple kernel instances all running in kernel context, they surpass the performance of even Xen, putting them more or less on par with FreeBSD jails (which, of course, use a single kernel image)
Anonymous (IP: —.dyn.columbia.edu)
This is a common misconception among BSD users: Jails and containers are not virtualizations, as you do not run a different instance of the kernel. Inside the jail you are completely tied to the original kernel, or in other words: there is simply no virtualization of the system at all.
Yet more trolls for a semantic argument. I would personally love to see whose definition you are using which says that virtualization explicitly requires a seperate kernel instance.
The resources being virtualized are those of the system. While the kernel is the same, the userspace is seperate and isolated. Just because the virtualization is occuring within a single kernel image doesn’t mean that it isn’t virtualization. Jails are substantially more than simply locking down what parts of the system given processes have access to…
Yet more trolls for a semantic argument. I would personally love to see whose definition you are using which says that virtualization explicitly requires a seperate kernel instance.
—-
dude. better have a good idea about what you are talking about.
SEMANTICS are important here…
anyway you lost all credibility with all your bsd jail comparisons
Yet more trolls for a semantic argument. I would personally love to see whose definition you are using which says that virtualization explicitly requires a seperate kernel instance.
Beside your unacceptable tone:
1. In your “definition”, virtually every resource sharing technique would be virtualization, from chroot() to even ordinary time sharing, rendering the word “virtualization” completely useless.
2. Let’s take e.g. the humble (“loose”, so the author) difinition: http://www.kernelthread.com/publications/virtualization/
“virtualization is a framework or methodology of dividing the resources of a computer into multiple execution environments, by applying one or more concepts or technologies such as hardware and software partitioning, time-sharing, partial or complete machine simulation, emulation, quality of service, and many others.”
plus what one might further considers: http://en.wikipedia.org/wiki/Virtualization
Note that both definition more or less contains the feature of being able to specify how to divide the resources, i.e. “concentration” or “logical blocking/partitioning”, not just mere processes floating around.
Or you can refer to:
http://www.acmqueue.com/modules.php?name=Content&pa=showpage&pid=17…
explicitly differing between BSD jail and virtualization.
The resources being virtualized are those of the system. While the kernel is the same, the userspace is seperate and isolated.
But they are certainly not different execution environments, but just different instances of the same execution environment.
Just because the virtualization is occuring within a single kernel image doesn’t mean that it isn’t virtualization.
In the same kernel image, you are unable to provide different execution environments. Also unlike real virtualization, security vulnerabilities in BSD jails would immediately leads to leaks shows clearly that it is not a different execution environment – inacceptable for those security-sensitve application of virtualization:
http://lists.freebsd.org/pipermail/freebsd-security-notifications/2…
Jails are substantially more than simply locking down what parts of the system given processes have access to…
Jails provides absolutely no ability for “concentration” or “logical partitioning” whatsoever, similar to chroot() or e.g. the time sharing itself.
To summarize: I somehow have the feeling that you are attempting to “dilute” the definition of virtualization so much that people cannot see the real difference between real virtual patitions as provided by VMware, Sun N1 Grid, HP/UX vPar, … and BSD jails or chroot(). BSD jails has its advantages, but is not the same.