Jes Sorensen wrote: > Zhang, Xiantao wrote: >> Hi, Jes >> Could you help to verify the patch on you mainframe? As you said, >> when guests configures with more that 200G ram, it will cost 20 >> minutes to initilize qemu. In this patch, I changed the logic of >> memory allocation, and I think time spent on initilization stage >> should be reduce to less than 1min. If so, it should resolve the >> issue you met. > > Xiantao, > > This patch is pretty amazing, I can now get QEMU to start booting in a > few seconds with a 200GB setup, instead of the 20+ minutes it took > before. Without this patch, the 200G memory is really allocated for this guest at starting stage, but you know, this is unnecessary, because guest only uses very little memory(<1G) at boot time. So we can have the rest memory allocated until they are really used by guest. With this patch, regardless of user assigning how much memory to guest, only ~1G memory is really allocated to this guest. Therefore, guests with 200G memory and with 1G memory should cost same time when they boot up. You also can use free command in host to check it, when guest boots up, you may see freed memory in host only reduces ~1G. > I did see one time where I got what looked like a CPU deadlock, but I > have only seen it once, so it doesn't have to be because of this > patch. Without this patch, I also met the issue once. So it may have nothing to do with the patch, but anyway, it should be a pentential issue, and we need to keep an eye on it. > Well done! > > Cheers, > Jes -- To unsubscribe from this list: send the line "unsubscribe kvm-ia64" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
