Another Phoronix article

[Jeremy Chadwick]

On Sun, Jan 09, 2011 at 03:56:12AM +0100, Ivan Voras wrote:
> On 9 January 2011 03:30, Adam Vande More <amvandemore at gmail.com> wrote:
> > On Sat, Jan 8, 2011 at 5:35 PM, Ivan Voras <ivoras at freebsd.org> wrote:
> >>
> >> In summary: very varied results, some are expected (low parallel write
> >> small IOPS for FreeBSD), some are not (apparently the BSDs have a monopoly
> >> on high-performance gzip :) ).
> >
> > I think might have read the graph backwards or I'm misinterpreting your
> > statement.  The BSD's have a monopoly on low performance gzip meaning they
> > aren't as fast as the Linux counterparts.
> 
> You're right. Which is interesting - it isn't a (simple) compiler
> difference since DragonflyBSD has gcc 4.4 also.
> 
> > That's not much a fs benchmark
> 
> Yes, they keep ignoring that remark.

I agree it's not much of a filesystem benchmark but more of a
"usability" benchmark.  I actually see the legitimacy of using it as a
benchmark point, but shouldn't be classified as a filesystem bench.

Has anyone actually taken the time to profile gzip and libz on BSD to
find out where it spends most of its time compared to Linux?  It could
be something as simple as a write() operation with a buffer size that
isn't optimal (compared to what the underlying filesystem has), or
equally some badly-sized buffers or operations/logic used within libz.
And also remember, FreeBSD's gzip is not pure GNU gzip (see man page).

All I'm trying to say here is that the effort to determine why BSD's
gzip behaves worse than on Linux is something the community or
developers should investigate -- if anything, maybe it's good Phoronix
indirectly brought this to light.

-- 
| Jeremy Chadwick                                   jdc at parodius.com |
| Parodius Networking                       http://www.parodius.com/ |
| UNIX Systems Administrator                  Mountain View, CA, USA |
| Making life hard for others since 1977.               PGP 4BD6C0CB |