On Thu, Jun 19, 2014 at 5:06 PM, Stefan Fuhrmann
<stefan.fuhrmann_at_wandisco.com> wrote:
>
> Turn out that the ruby repo is something special
> in that it has very deep histories of relatively few,
> very small files combined with one huge changelog
> file (the latter taking up ~75% of the repo). See
> below for details.
>
> Also, please note that your exports contained
> >500000 files. Using 16MB of cache with that
> project size *may* not be an adequate setup.
> Upping that to insane 256MB (roughly what 1.6
> would use anyway), gives much better numbers.
> However, there is hardly a difference between
> f6 and f7 in these runs.
>
> Here my measurements with svn: under Linux:
>
> null-export svn://localhost/ruby/trunk
> 461 directories
> 4,263 files
> 48,880,945 bytes in files
> 23,783 properties
> 324,660 bytes in properties
>
> F7 packed 20.249s, all default
> F6 packed 20.168s, all default
>
> F7 packed 14.310s, 256MB cache
> F6 packed 15.167s, 256MB cache
>
> F7 packed 13.649s, 256MB cache -c 0
> F6 packed 14.226s, 256MB cache -c 0
>
> F7 packed 13.478s, 256MB cache -c 0 --cache-revprops yes
> F6 packed 13.223s, 256MB cache -c 0 --cache-revprops yes
>
> export svn://localhost/ruby/
> 54,641 directories
> 512,580 files
> 4,558,565,078 bytes in files
> 3,077,645 properties
> 45,809,954 bytes in properties
>
> F7 packed 763.520s, all default
> F6 packed 785.913s, all default
>
> F7 packed 153.520s, 256MB cache
> F6 packed 152.746s, 256MB cache
>
> F7 packed 64.695s, 256MB cache -c 0
> F6 packed 65.505s, 256MB cache -c 0
>
> F7 packed 55.534s, 256MB cache -c 0 --cache-revprops yes
> F6 packed 56.697s, 256MB cache -c 0 --cache-revprops yes
>
> null-log svn://localhost/ruby/
> 46,054 revisions
> 177,176 msg lines
> 0 changes
>
> F7 packed 58.494s, all defaults
> F6 packed 58.852s, all defaults
>
> F7 packed 2.153s, --cache-revprops yes
> F6 packed 2.470s, --cache-revprops yes
>
Can you guys dig a little bit deeper here? Seems the performance
regression that Ivan is seeing deserves some more investigation. With
Ivan testing http: (and file:) on Windows, and Stefan testing svn: on
Linux (with different cache settings) it's hard to see where the
regression comes from.
Maybe Ivan can retest with different cache settings (256MB etc), so we
at least know that it *can* be performant with http: on Windows? Or
maybe Stefan can do more benchmarks with http: (on Linux and on
Windows)?
Right now, the numbers might just as well indicate that f7 is fast on
Linux but slow on Windows. Or that it's only fast with svn: but slow
with http:. That would be "not good".
If f7 is all about better performance, there should be virtually no
performance regressions (especially not for setups with "default
configurations"):
- Whether it's the ruby or the FreeBSD repo.
- Whether it's svn: or http:.
- Whether it's hosted on Linux or on Windows.
- ...
--
Johan
Received on 2014-06-20 10:16:37 CEST