On Thu, May 22, 2025 at 12:55:05AM +0000, Zhijian Li (Fujitsu) wrote:
>
>
> On 22/05/2025 05:04, Dr. David Alan Gilbert wrote:
> > * Peter Xu (pet...@redhat.com) wrote:
> >> On Wed, May 21, 2025 at 08:43:37AM +0000, Zhijian Li (Fujitsu) wrote:
> >> [...]
> >>>> After this change, sample output (default, no "-a" specified):
> >>>>
> >>>> Status: postcopy-active
> >>>> Time (ms): total=40504, setup=14, down=145
> >>>> RAM info:
> >>>> Bandwidth (mbps): 6102.65
> >>>> Sizes (KB): psize=4, total=16777992
> >>>> transferred=37673019, remain=2136404,
> >>>> precopy=3, multifd=26108780, postcopy=11563855
> >>>> Pages: normal=9394288, zero=600672, rate_per_sec=185875
> >>>> Others: dirty_syncs=3, dirty_pages_rate=278378, postcopy_req=4078
> >>>
> >>> (Feel free to ignore my comment if you have reached a consensus.)
> >>>
> >>> Putting multiple fields in one line is a true need for human reading?
> >>
> >> It definitely helps me but I agree that can be subjective. So I'm happy to
> >> collect opinions.
> >>
> >> So my above layout was trying to leverage more on screens where width is
> >> bigger than the height (which is pretty much the default).
> >
> > I think perhaps the problem with the on-one-line layout is that the grouping
> > is wrong; grouping by unit probably doesn't make sense.
> >
> > So it makes sense to me to have:
> > Sizes: psize=4/KB
> > Transfer: total=16777992 kB transferred=37673019 kB remain=11563855 kB
> > Pages: normal=9394288 zero=600672
> > Page rates: transferred=185875/s dirtied=278378/s
> > Other: dirty_sync=3 postcopy_req=4078
>
>
> Oh, I vote this !!, more clear to me.
I followed up with Dave's idea, but then added all entries into it, below.
Status: postcopy-active
Time (ms): total=40504, setup=14, down=145
RAM info:
Throughput (Mbps): 6102.65
Sizes (KiB): pagesize=4, total=16777992
Transfers (KiB): transferred=37673019, remain=2136404
Channels (KiB): precopy=3, multifd=26108780, postcopy=11563855
Page Types: normal=9394288, zero=600672
Page Rates (pps): transfer_rate=185875, dirty_rate=278378
Others: dirty_syncs=3, postcopy_req=4078
Logically I should have moved "Throughput" out, because that should also
include all other things (non-ram iterators, device states). But currently
it's an entry under info->ram.. so I kept it there.
It also has the "total" in "Sizes" to make the next line shorter
(meanwhile, "total" is also a constant size like "psize"), the hope is it's
still close enough to read when reading "Transfers" on the next line.
I also provided further indents to "Channels" and "Page Types" because they
should be taken as sub-class of "Transfer".
How is this? Since we're at it, I can send a follow up patch after we
reach a consensus (I may also include that in another series where I'll
further add things into HMP; I'm looking at making blocktime to report page
latencies too).
--
Peter Xu
