It's different, and not really a problem as such. If you are running one of the native Xen kernels, you can lock your clock to that of the host (/proc/sys/xen/independent_wallclock = 0) or have it independent (/proc/sys/xen/independent_wallclock = 1). If you are running a paravirt kernel, your clock is independent of the host. Independent clock accuracy is determined by the performance of the crystal oscillator in the host -- hence the drift. Use ntpd to keep your clock accurate.

_________________
/ Peter

Thanks for the info. So if I set independent_wallclock = 0, and my host runs ntpd, everything will be fine, and I don't need to run ntpd myself?

I have a linode in Fremont set up exactly like that and it is accurate to the second.

_________________
/ Peter