Package download speed calculation, spurious results?

  • Done
  • quality assurance status badge
Details
2 participants
  • francismb
  • Ludovic Courtès
Owner
unassigned
Submitted by
francismb
Severity
normal
F
F
francismb wrote on 19 Feb 2023 12:14
(address . bug-guix@gnu.org)
9f94be69-0e66-06ef-a0c5-dc4665d4888b@email.de
Dear Maintainers team,
I've notice that the package download speed (is that the measure?),
seems to have some special edges/cases giving IMHO not realistic values,
see for e.g. the value calculated for 'graphviz' on my current
(~2023.02.19T12:00UTC) pull and upgrade iteration:

guix pull && guix upgrade
Updating channel 'guix' from Git repository at
Authenticating channel 'guix', commits 9edb3f6 to 57495d0 (286 new
commits)...
Building from this channel:
substitute: updating substitutes from 'https://ci.guix.gnu.org'... 100.0%
substitute: updating substitutes from 'https://ci.guix.gnu.org'... 100.0%
substitute: updating substitutes from 'https://ci.guix.gnu.org'... 100.0%
substitute: updating substitutes from 'https://ci.guix.gnu.org'... 100.0%
substitute: updating substitutes from 'https://ci.guix.gnu.org'... 100.0%
config.scm 501B
19KiB/s 00:00 [##################] 100.0%
git.scm 101B
64KiB/s 00:00 [##################] 100.0%
hash.scm 132B
13KiB/s 00:00 [##################] 100.0%
module-import 2KiB
367KiB/s 00:00 [##################] 100.0%
module-import 2KiB
316KiB/s 00:00 [##################] 100.0%
module-import-compiled 1.2MiB
5.3MiB/s 00:00 [##################] 100.0%
module-import-compiled 1.2MiB
6.0MiB/s 00:00 [##################] 100.0%
building
/gnu/store/h3lx1r3canrrfy2km329bzqdhq1l6bz2-compute-guix-derivation.drv...
graphviz-2.49.0-doc 2.0MiB
6.33GiB/s 00:00 [##################] 100.0%
dbus-1.12.20 266KiB
412.0MiB/s 00:00 [##################] 100.0%
fribidi-1.0.9 71KiB
175.0MiB/s 00:00 [##################] 100.0%
avahi-0.8 349KiB
1.8MiB/s 00:00 [##################] 100.0%
gd-2.3.2 124KiB
376KiB/s 00:00 [##################] 100.0%
gettext-minimal-0.21 3.4MiB
8.2MiB/s 00:00 [##################] 100.0%
gnutls-3.7.7 1.6MiB
4.97GiB/s 00:00 [##################] 100.0%
gettext-minimal-0.21-doc 1.0MiB
5.5MiB/s 00:00 [##################] 100.0%
graphite2-1.3.13 90KiB
875KiB/s 00:00 [##################] 100.0%
gts-0.7.6 273KiB
1.8MiB/s 00:00 [##################] 100.0%
guile-avahi-0.4.1 57KiB
323KiB/s 00:00 [##################] 100.0%
guile-git-0.5.2 416KiB
2.9MiB/s 00:00 [##################] 100.0%
guile-gnutls-3.7.11 98KiB
642KiB/s 00:00 [##################] 100.0%
guix-daemon-1.4.0-3.d5fece6 260KiB
820KiB/s 00:00 [##################] 100.0%
icu4c-69.1 9.1MiB
7.0MiB/s 00:01 [##################] 100.0%
libdatrie-0.2.13 38KiB
200KiB/s 00:00 [##################] 100.0%
harfbuzz-2.8.2 813KiB
5.7MiB/s 00:00 [##################] 100.0%
libice-1.0.10 72KiB
519KiB/s 00:00 [##################] 100.0%
libssh-0.9.6 236KiB
2.1MiB/s 00:00 [##################] 100.0%
libsm-1.2.3 41KiB
282KiB/s 00:00 [##################] 100.0%
guile-ssh-0.16.3 217KiB
1.4MiB/s 00:00 [##################] 100.0%
libthai-0.1.28 162KiB
1.1MiB/s 00:00 [##################] 100.0%
guile-ssh-0.16.3-debug 223KiB
1.3MiB/s 00:00 [##################] 100.0%
libxext-1.3.4 33KiB
227KiB/s 00:00 [##################] 100.0%
libxt-1.2.1 181KiB
713KiB/s 00:00 [##################] 100.0%
perl-gettext-1.07 11KiB
142KiB/s 00:00 [##################] 100.0%
libxmu-1.1.3 68KiB
519KiB/s 00:00 [##################] 100.0%
libxpm-3.5.13 50KiB
350KiB/s 00:00 [##################] 100.0%
perl-pod-parser-1.65 75KiB
456KiB/s 00:00 [##################] 100.0%
libxaw-1.0.14 322KiB
1.0MiB/s 00:00 [##################] 100.0%
perl-sgmls-1.1 33KiB
529KiB/s 00:00 [##################] 100.0%
perl-syntax-keyword-try-0.28 29KiB
244KiB/s 00:00 [##################] 100.0%
perl-text-wrapi18n-0.06 5KiB
32KiB/s 00:00 [##################] 100.0%
perl-unicode-linebreak-2019.001 105KiB
853KiB/s 00:00 [##################] 100.0%
perl-xs-parse-keyword-0.06 32KiB
242KiB/s 00:00 [##################] 100.0%
perl-yaml-tiny-1.73 26KiB
177KiB/s 00:00 [##################] 100.0%
pixman-0.40.0 233KiB
612KiB/s 00:00 [##################] 100.0%
po4a-0.68 1.9MiB
10.1MiB/s 00:00 [##################] 100.0%
cairo-1.16.0 644KiB
4.1MiB/s 00:00 [##################] 100.0%
pango-1.48.10 344KiB
1.2MiB/s 00:00 [##################] 100.0%

graphviz-2.49.0 1.2MiB
1.15GiB/s 00:00 [ ] 5.1%
graphviz-2.49.0 1.2MiB
2.2MiB/s 00:01 [##################] 100.0%

substitute: updating substitutes from 'https://ci.guix.gnu.org'... 100.0%

I'm not behind a ~ >100MB connection, and it's curious to get 1.15GiB/s
here, is that somehow the value from some cache? or is a spurious
calculation? (IMHO nothing that releases the packaging unusable thus low
priority, just curiosity, is that a bug?)

Thanks in advance,
francisMB
L
L
Ludovic Courtès wrote on 23 Feb 2023 14:28
(name . francismb)(address . francismb@email.de)(address . 61625@debbugs.gnu.org)
87leko7bpd.fsf@gnu.org
Hi,

francismb <francismb@email.de> skribis:

Toggle quote (12 lines)
> graphviz-2.49.0 1.2MiB
> 1.15GiB/s 00:00 [ ] 5.1%
> graphviz-2.49.0 1.2MiB
> 2.2MiB/s 00:01 [##################] 100.0%
>
> substitute: updating substitutes from 'https://ci.guix.gnu.org'... 100.0%
>
> I'm not behind a ~ >100MB connection, and it's curious to get 1.15GiB/s
> here, is that somehow the value from some cache? or is a spurious
> calculation? (IMHO nothing that releases the packaging unusable thus low
> priority, just curiosity, is that a bug?)

As you can see, the 1.15GiB/s was printed while Guix had downloaded just
5.1% of 1.2 MiB, meaning 61 KiB.

The time it takes to download 61 KiB is very small, which is why you can
end up with surprising values when dividing that by the time it took.

Notice that the final result of 2.2 MiB/s is probably accurate.

There’s not much to do here, except perhaps not displaying anything
until at a bigger chunk of time has elapsed.

Ludo’.
L
L
Ludovic Courtès wrote on 3 Mar 2023 10:46
(name . francismb)(address . francismb@email.de)(address . 61625-done@debbugs.gnu.org)
878rge182n.fsf@gnu.org
Hi,

francismb <francismb@email.de> skribis:

Toggle quote (5 lines)
> thanks for the explanation, then AFAIU, is an scaling/extrapolation to
> seconds effect. I can imagine, that waiting a bit more, just for the
> accuracy here, isn't worth the effort (or complexity). Thus IMHO the
> issue may be closed.

Alright, closing!

Ludo’.
Closed
?