[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] xenoprof passive mode overhead


  • To: xen-devel@xxxxxxxxxxxxxxxxxxx
  • From: Ya-Yunn Su <yayunn.su@xxxxxxxxx>
  • Date: Thu, 25 Mar 2010 14:06:26 -0400
  • Delivery-date: Mon, 29 Mar 2010 10:13:10 -0700
  • Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:date:message-id:subject:from:to:content-type; b=muH+GqRv2IHVFoJ69lcN33ZfeugAf4x0jr5py8dYA9xrVC2eFGXerRNMBnka2hZ69J /GGxLFYiHiaoodglZKEO/As1//yxJUO4DnCs17O5qbfT2cJyMfyHWKP2QAnelEuePZ4F WOllkbcidKh+EBWN6HwSd5gubqHZLW012sDbQ=
  • List-id: Xen developer discussion <xen-devel.lists.xensource.com>

Hi all,
  I'd like to use xenoprof passive mode on domain-0 to calculate L2
cache missed for each guest domain.  I setup oprofile daemon as
followed

opcontrol --start-daemon --event=LLC_MISSES:4000
--xen=/boot/xen-3.4.2.gz
--vmlinux=/scratch/xen-3.4.2/build-linux-2.6.18-lttng-xen0_x86_32/vmlinux
--passive-domains=1,2
--passive-images=/scratch/xen-3.4.0/build-linux-2.6.18-xenU_x86_32/vmlinux,/scratch/xen-3.4.0/build-linux-2.6.18-xenU_x86_32/vmlinux

My machine has two cores and I pinned domain-0 to one core and guest
domains to the other.  When I ran oprofile to calculate cache
misses/sec every 5 seconds on domain 0, I noticed that the transient
CPU utilization can be as high as 30%.  This affects the performance
of whatever benchmark I'm running in the guest domain.  Is this 30%
overhead for domain-0 common or am I doing something obviously wrong?
Any suggestion is appreciated.

Thank you,
Ya-Yunn

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.