> On Wed, 29 Jun 2005, Magenheimer, Dan (HP Labs Fort Collins) wrote: > > Also unlike Xen/x86, Linux/ia64 with CONFIG_XEN enabled runs > > successfully on Xen, but also run natively on hardware (called > > "transparent paravirtualization"). > > That's neat indeed. > > > This should be appealing to Linux/ia64 distributors as different > > kernels need not be shipped for running on Xen or running native. > > What's the performance impact for using transparanent > paravirtualization? > Is it sufficiently low that distributors really can consider using a > single kernel for both native and Xen "hosting"?
I ran a test over the weekend comparing kernel.org 2.6.12 compiled for a zx1 (rx2600) vs the same with transparent paravirtualization enabled. The test is compiling linux-2.6.12 (several times... I interrupted the second run) in single-user mode. As you can see, the performance impact is "in the noise". If someone wants to run a benchmark that can be measured with better precision than 1 part in 1000, please let me know. Dan ===== transparent paravirtualization [EMAIL PROTECTED] djm]$ egrep '^real|^user|^sys' *048 05.07.11.071048:real 18m42.709s 05.07.11.071048:user 18m26.588s 05.07.11.071048:sys 0m25.431s 05.07.11.071048:real 18m33.756s 05.07.11.071048:user 18m27.412s 05.07.11.071048:sys 0m25.480s 05.07.11.071048:real 18m34.987s 05.07.11.071048:user 18m28.532s 05.07.11.071048:sys 0m25.700s 05.07.11.071048:real 18m35.458s 05.07.11.071048:user 18m28.976s 05.07.11.071048:sys 0m25.502s native 2.6.12 [EMAIL PROTECTED] djm]$ egrep '^real|^user|^sys' *753 05.07.10.180753:real 18m43.805s 05.07.10.180753:user 18m27.958s 05.07.10.180753:sys 0m25.181s 05.07.10.180753:real 18m34.283s 05.07.10.180753:user 18m27.864s 05.07.10.180753:sys 0m25.414s 05.07.10.180753:real 18m34.513s 05.07.10.180753:user 18m28.247s 05.07.10.180753:sys 0m25.362s - To unsubscribe from this list: send the line "unsubscribe linux-ia64" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
