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

Re: [libvirt] [PATCH 0/5 v3] Add NUMA memory tuning support



于 2011年06月20日 13:30, Daniel Veillard 写道:
On Fri, Jun 17, 2011 at 06:22:53PM +0800, Osier Yang wrote:
Hi,

This series incorporated previous feedbacks received on the
XML schema, especially on syntax of "nodeset". Now using same
syntax of "cpuset" of "vcpu". And doing conversion in qemu
driver internally, as libnuma uses different bitmask format
to represent NUMA nodes.

The XML is like:
   <numatune>
     <memory mode="strict" nodeset="0-10,^5"/>
   </numatune>

Value of "mode" can be either of "strict", "preferred", or
"interleave".

One may specify nodemask which is out of range, for this
case, libvirt will report soft warnings instead of hard
error. (Kernel is silent as long as one of the set bit is
valid, E.g. For a host with 2 NUMA nodes, kernel will be
silent for nodemask "0101010101", as first "bit" which
represents the first NUMA node is set). The reason for
reporting soft warnings instead of hard error is one may
want to define the memory tuning policy priorly with a node
doesn't exist yet, but may come soon.

Any feedback is appreciated.

   At this point I think this patchset is ripe for consumption,

   ACK, please push,

    thanks !

Daniel


Thanks, COMMITED.

Regards
Osier


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