On Thu, Dec 02, 2004 at 12:11:47PM -0500, Paul Iadonisi wrote: > 1. Uh ... so where are the Xen control tools? I suspect these are > forthcoming, I just wanted to double check. Correct > 2. On a somewhat related note, Red Hat at one time bundled a UML kernel > but later stopped (even in an errata kernel for that same release). > What were the reasons for this? At the time UML wasn't in the base kernel and it got very hard to maintain. > 3. I know some basic differences between UML and Xen, but I'm wondering > about the differences in mindshare. How much 'industry' buy-in does > each have? I note that Xen has some support of both HP Labs and Intel > Research Cambridge. A lot of people are very excited about Xen right now. Virtualisation is a big thing and you need the tools to do it well. > 4. Any idea if it will show up in a future version of RHEL as well? No idea myself > 5. On the technical side ... are the changes to the domain0 kernel > running on the raw hardware non-intrusive enough that it may one > day become feasible and/or desirable to just ship the standard > kernel as a Xen-enabled kernel? Not in the short term although getting to the stage where you do "insmod virtualisation" is obviously desirable. Likewise the Xen people really want Xen not to do non virtualisation work that could be done in the guest0 kernel. > 6. What is the likelihood that Xen will be included in Linus' kernel > someday? Has anyone solicited Linus' opinion on Xen? I'm curious The Xen folks are currently submitting the small set of patches needed to support Xen as arch/xen. Several of them have been through a couple of revisions already and some are now accepted upstream. > end-user perspective ... if you really want to run a bunch of Xen > domains within a UML ... I suppose there are worse ways to torture > yourself ;-). UML and Xen solve different problems. Xen is an extremely efficient secure virtualisation system, UML is a fantastic research, debugging and prototyping system but is unlikely ever to be as efficient. > 7. What do people think of the idea of porting Anaconda to run under Xen so > that you can install a full release of Fedora Core (or RHEL) as an > unprivileged guest the way you would normally install the OS? I'm not > exactly volunteering, but I did *almost* have the installer booting within > UML and start to install Red Hat Linux. It bombed out at some point. Can't > remember where. I did it all without hacking any python, IIRC. "Send patches" Its obviously a key part of the thing that you can provision Xen guests efficiently, manage them, update them and the like. Stateless linux obviously helps there but anaconda on Xen guests > guest0 also has value.