[libvirt] [PATCH 04/10] test: add test case for resctrl monitor
Wang, Huaqiang
huaqiang.wang at intel.com
Mon Sep 10 19:18:52 UTC 2018
> -----Original Message-----
> From: John Ferlan [mailto:jferlan at redhat.com]
> Sent: Saturday, September 8, 2018 1:14 AM
> To: Wang, Huaqiang <huaqiang.wang at intel.com>; libvir-list at redhat.com
> Cc: Feng, Shaohe <shaohe.feng at intel.com>; Niu, Bing <bing.niu at intel.com>;
> Ding, Jian-feng <jian-feng.ding at intel.com>; Zang, Rui <rui.zang at intel.com>
> Subject: Re: [libvirt] [PATCH 04/10] test: add test case for resctrl monitor
>
>
>
> On 09/07/2018 05:12 AM, Wang, Huaqiang wrote:
> >
> >
> >> -----Original Message-----
> >> From: John Ferlan [mailto:jferlan at redhat.com]
> >> Sent: Wednesday, September 5, 2018 7:59 PM
> >> To: Wang, Huaqiang <huaqiang.wang at intel.com>; libvir-list at redhat.com
> >> Cc: Feng, Shaohe <shaohe.feng at intel.com>; Niu, Bing
> >> <bing.niu at intel.com>; Ding, Jian-feng <jian-feng.ding at intel.com>;
> >> Zang, Rui <rui.zang at intel.com>
> >> Subject: Re: [libvirt] [PATCH 04/10] test: add test case for resctrl
> >> monitor
> >>
> >>
> >>
> >> On 08/27/2018 07:23 AM, Wang Huaqiang wrote:
> >>> Signed-off-by: Wang Huaqiang <huaqiang.wang at intel.com>
> >>> ---
> >>> .../linux-resctrl/resctrl/info/L3_MON/max_threshold_occupancy | 1 +
> >>> .../vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/mon_features | 3
> +++
> >>> tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/num_rmids | 1 +
> >>> tests/vircaps2xmldata/vircaps-x86_64-resctrl.xml | 6 ++++++
> >>> 4 files changed, 11 insertions(+)
> >>> create mode 100644
> >>> tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/max_threshol
> >>> d_
> >>> occupancy create mode 100644
> >>> tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/mon_features
> >>> create mode 100644
> >>> tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/num_rmids
> >>>
> >>
> >> And this would be combined with part of patch2 and patch3
> >>
> >>> diff --git
> >>> a/tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/max_thresh
> >>> ol
> >>> d_occupancy
> >>> b/tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/max_thresh
> >>> ol
> >>> d_occupancy
> >>> new file mode 100644
> >>> index 0000000..77f05e2
> >>> --- /dev/null
> >>> +++ b/tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/max_th
> >>> +++ re
> >>> +++ shold_occupancy
> >>> @@ -0,0 +1 @@
> >>> +270336
> >>> diff --git
> >>> a/tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/mon_featur
> >>> es
> >>> b/tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/mon_featur
> >>> es
> >>> new file mode 100644
> >>> index 0000000..0c57b8d
> >>> --- /dev/null
> >>> +++ b/tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/mon_fe
> >>> +++ at
> >>> +++ ures
> >>> @@ -0,0 +1,3 @@
> >>> +llc_occupancy
> >>> +mbm_total_bytes
> >>> +mbm_local_bytes
> >>
> >> Could/should this list values that aren't prefixed by "llc_" and "mbm_"
> >> to validate your code?
> >
> > Will add some 'fake' features to the list, and make more tests.
> > To be done in next version patch.
> >
> >>
> >> There's only 1 set of data but it's printed twice - that's the reason
> >> for my comment in patch2 about duplication of the same data that is
> unnecessary.
> >> What if there were 10 bank id's, 100? 1000? - lots of waste. Only 2, no big
> deal.
> >>
> >
> > <cache>
> > <bank id='0' level='3' type='both' size='55' unit='MiB' cpus='0-21,44-65'>
> > <control granularity='2816' unit='KiB' type='both' maxAllocs='16'/>
> > <monitor threshold='270336' unit='B' maxAllocs='176'>
> > <feature name='llc_occupancy'/>
> > </monitor>
> > </bank>
> > <bank id='1' level='3' type='both' size='55' unit='MiB' cpus='22-43,66-87'>
> > <control granularity='2816' unit='KiB' type='both' maxAllocs='16'/>
> > <monitor threshold='270336' unit='B' maxAllocs='176'>
> > <feature name='llc_occupancy'/>
> > </monitor>
> > </bank>
> > </cache>
> >
> > Above is the cache capabilites section, dumped from my system through
> > 'virsh capabilities' command.
> > This is a 2-socket E5-2699v4 CPU(22 core with CAT/CMT enabled and CDP
> > disabled) system, as you said, the cache monitor capability is printed more
> than once.
> >
> > And I need to point out that the following cache 'control' element is
> > also printed for twice, it met the same situation you mentioned for cache
> monitor.
> > "<control granularity='2816' unit='KiB' type='both' maxAllocs='16'/>"
> >
>
> I think we covered this earlier... I'm still not really clear on what <control> per
> <bank> really provides other than it's a calculated value based on a few different
> numbers and honestly I had a hard time following that logic all over the place.
>
Understood. cache <control> information might be alerted by the cache size or
something else.
Have proposed new cache monitor layout in previous email's update. Hope it
looks better.
Thanks for review.
Huaqiang
> John
>
> > After you have read my considerations(in the reply to your review of
> > patch 2) for reason why I used current disign, if you still think it
> > not wise to make it duplicated, let's find a proper place to keep the
> > data and eliminate such kind of duplication.
> > We can make any changes to make our design more reasonable in the
> > design stage.
> >
> >
> > Thanks for your kind revew!
> >
> > BR
> > Huaqiang
> >
> >> John
> >>
> >>> diff --git
> >>> a/tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/num_rmids
> >>> b/tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/num_rmids
> >>> new file mode 100644
> >>> index 0000000..1057e9a
> >>> --- /dev/null
> >>> +++ b/tests/vircaps2xmldata/linux-resctrl/resctrl/info/L3_MON/num_rm
> >>> +++ id
> >>> +++ s
> >>> @@ -0,0 +1 @@
> >>> +176
> >>> diff --git a/tests/vircaps2xmldata/vircaps-x86_64-resctrl.xml
> >>> b/tests/vircaps2xmldata/vircaps-x86_64-resctrl.xml
> >>> index 9b00cf0..678fdc9 100644
> >>> --- a/tests/vircaps2xmldata/vircaps-x86_64-resctrl.xml
> >>> +++ b/tests/vircaps2xmldata/vircaps-x86_64-resctrl.xml
> >>> @@ -44,9 +44,15 @@
> >>> <cache>
> >>> <bank id='0' level='3' type='both' size='15' unit='MiB' cpus='0-5'>
> >>> <control granularity='768' min='1536' unit='KiB' type='both'
> >>> maxAllocs='4'/>
> >>> + <monitor threshold='270336' unit='B' maxAllocs='176'>
> >>> + <feature name='llc_occupancy'/>
> >>> + </monitor>
> >>> </bank>
> >>> <bank id='1' level='3' type='both' size='15' unit='MiB' cpus='6-11'>
> >>> <control granularity='768' min='1536' unit='KiB' type='both'
> >>> maxAllocs='4'/>
> >>> + <monitor threshold='270336' unit='B' maxAllocs='176'>
> >>> + <feature name='llc_occupancy'/>
> >>> + </monitor>
> >>> </bank>
> >>> </cache>
> >>> <memory_bandwidth>
> >>>
More information about the libvir-list
mailing list