Gentle persons:
Gene Heskett asked earlier this month on emc-users if "the" server
(presumably he meant git.linuxnc.org since he mentioned an update) were
under attack because he was getting ca 20kb/s transfer speeds instead of
his usual ca 380kb/s.
I wasn't interested in the abnormality (speed seems more or less back to
normal) but in Gene's "usual" 380kb/s. As it happens, I'm seeing roughly 30
percent lower transfer speed, but...
I have crudely benchmarked git.linuxcnc.org against github.com by cloning
the linuxcnc repo from each.
Results:
git clone git://git.linuxcnc.org/git/linuxcnc: elapsed real time: 5m22s
67.31MB transferred, 260KiB/s transfer speed
git clone https:github.com/jepler/linuxcnc-mirror: elapsed real time: 43.5s
72.76MB transferred, 2.7MiB/s transfer speed
Leaving aside why Jeff's repo contains an additional 5MB, this 10-to-1
difference in transfer is substantial. It is true as well for other
linuxcnc mirrors on github.com (try "linuxcnc" in the github searchbox). I
presume the difference is mostly due to faster Internet "pipes" into the
data warehouse(s) which host github and not due to differences in the
servers themselves. Perhaps the use of different transfer protocols---git
versus https---plays a minor role but I would naively expect the git
protocol to be faster.
Have we considered setting up an "official" repository to github.com which
mirrors git.linuxcnc.org? I suggest there should be such and that it should
be contained in a LinuxCNC account. (So far, no one seems to have
registered that user/organization name but there is nothing to prevent
anyone from grabbing it.)
I like this idea for several reasons:
1) redundancy (by design rather than by happenstance)
2) faster throughput for many of us users
3) as well, this could lower the burden on the git.linuxcnc.org host and
its ISP but perhaps this does not matter.
I assume that the sizes of the different repos already on github differ
only because the different github users have added their own branches while
major branches like master, ja3, unified-build-candidate-3 remain
identical, but you know what they say about "assume". Besides,
non-developers using a repo as a read-only resource should not have to
assume or to dig through the commits and/or directory trees to find out
even if they have the know-how. Hence my suggestion.
Regards,
Kent
------------------------------------------------------------------------------
October Webinars: Code for Performance
Free Intel webinars can help you accelerate application performance.
Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from
the latest Intel processors and coprocessors. See abstracts and register >
http://pubads.g.doubleclick.net/gampad/clk?id=60134071&iu=/4140/ostg.clktrk
_______________________________________________
Emc-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-developers
