Xen is coming [LWN.net]

Xen is coming

Posted Nov 25, 2004 21:40 UTC (Thu) by alspnost (guest, #2763)
Parent article: Xen is coming

So am I right in thinking that this is a sort of "inverse" UML? It sounds like you build a special Xen host kernel, on which you run multiple vanilla kernels; as opposed to UML, where you build multiple UML guest kernels to run on a vanilla host kernel.

I'm quite interested in this: what I'm trying to do is run multiple 32-bit virtual Linuxes on a 64-bit host kernel on an Opteron server. This is to run a crucial Windoze app via CrossOver on Linux - that part already works fine on bog-standard 32-bit Linuxes. I figured that by virtualising on top of a 64-bit machine, I could effectively use more than 4GB of RAM without any nonsense; what's important is being able to run multiple instances of this app, each with up to 1GB of RAM, but usually not more.

Anyway, ideas on a postcard please, but otherwise, Xen looks intruiging ;-)

(Log in to post comments)

Xen is coming

Posted Nov 26, 2004 10:52 UTC (Fri) by mab (subscriber, #314) [Link]

I'm wondering the same, if you are right it seems like we will be giving Vmware and Microsofts Virtual PC a run for their money on systems that only run Linux. I like it I hope it goes in

Xen is coming

Posted Nov 27, 2004 14:25 UTC (Sat) by Lorenzo (guest, #260) [Link]

Linux on Linux is not interesting to me at all.
Windows on Linux is interesting.

The various Wine configurations or specialized Win9x on Linux virtualizations just don't cut it for me. I need a full x86, and soon I presume x86-64, virtualization so that I can run Win2K and [gakkk] WinXP on Linux.

Don't tell me to get VMWare. Too flippin' expensive.

Xen is coming

Posted Nov 27, 2004 16:42 UTC (Sat) by riel (subscriber, #3142) [Link]

The idea behind Xen is that full virtualisation is complex and often inefficient - so the obvious solution is for the guest OSes to behave and make the job easy for the virtualisation engine.

This strategy, often called paravirtualisation, is a lot more efficient than full virtualisation, but requires special guest kernels. The rest of the guest operating system can remain unchanged. This is a very promising approach to virtualisation, one that should help a lot of Linux developers by allowing them to efficiently run multiple versions of a distribution on the same system, or help with security separation. I am working on integrating Xen into Fedora and hope it will be a good piece of infrastructure for the Fedora developers.