Migration Queue?

as i understand it, the main reason that linode began the xen beta is that xen had significantly less overhead than uml, and linode was considering complete changeover at some point. that may no longer apply.

first of all there has been quite a lot of uml development to try to eliminate overhead. in particular, newer uml kernels can take advantage of virtualization features in the host kernel that in turn rely on virtualization features in the cpu. this is something that xen has always done, and soon the deployed uml kernel will also do it. the new linode host machines all have the necessary cpu features, as well as some of the older hosts. the older that lack this feature in the cpu are scheduled for decommissioning beginning this sunday. this basically means that it is only a matter of time before linode's uml deployments match xen for performance, and the impetus to switch technologies essentially disappears.

second of all, the linode staff has expressed at various times that they're unhappy with xen's stability. the issues are longstanding and so far no one has figured out how to fix them, and so "production level" stability won't be available on a known timeline.

in short, linode will be sticking with uml. as for the single xen beta host, i know of no plan to revert it back to uml.

caveat: i am not a staff member; i only read the tea leaves.