Greetings users and developers,
Just put up a feature page for the "NUMA aware KSM support";
In summary,
===========
The KSM service is optimizing shared memory pages across all NUMA nodes. The consequences
is: a shared memory pages (controlled by KSM) to be read from many CPUs across NUMA nodes.
Hence degrading (could totally eliminate) the NUMA performance gain. The optimal behavior
is for KSM feature to optimize memory usage per NUMA nodes. This feature is implemented in
KSM since RHEL 6.5. We want to introduce a new feature to oVirt that will allow host
administrator to control NUMA aware KSM policy.
You're welcome to review the feature page:
http://www.ovirt.org/NumaAwareKsmSupport
The feature page is responding to RFE
https://bugzilla.redhat.com/show_bug.cgi?id=840114
Your comments appreciated.
Thanks
Joy and happiness
Dudi
--
Dudi Maroshi
Software Engineer
Red Hat Enterprise Virtualization project
http://www.redhat.com/en/technologies/virtualization
maroshi(a)redhat.com
--------------------------------
I am perfectly imperfect.
"There is only one absolute truth, and that is: there in no one absolute
truth."