[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: SprezzOS and high-performance

Miro Moman left as an exercise for the reader:
> I have been searching the entire Sprezzatech website and Wiki trying
> to find out how exactly SprezzOS delivers that promised
> high-performance. The only thing I have actually discovered is that
> it uses ZFS as the default filesystem. I have also found elsewhere
> (Phoronix, I believe) that the SprezzOS implementation of APT should
> be faster than the standard one.
> Other than that, are the foundations of those high-performance gains
> explained somewhere?
> I am currently struggling to compile Funtoo from source with ZFS and
> all kind of optimisations. Am I wasting my time? Would I be better
> off just installing SprezzOS?

 {I've CC'd the debian-derivatives list}

Great questions. There's been no comparative study of speed across SprezzOS,
but I posted some measurements regarding my APT work on the Sprezzatech


The major ways in which I was hoping to speed up the Debian-derived
ecosystem through SprezzOS were:

 a) My own development on Debian tools. I tried to engage the APT team after
 completing this work, but they were very resistant to optimization efforts.
 Consult the APT list for more details. To be fair, I could have played the
 political game here a lot better -- I pretty much dropped completely
 rewritten tools on the APT team, with no real warning. Learned some
 valuable lessons about how to engage a community.

 The RAPTORIAL project's code is still available on github, and there's a
 Wiki page about it here:


 So far as I know, raptorial-file(1), rapt-show-versions(1), and
 rapt-parsechangelog(1) remain wholly suitable, massively accelerated
 drop-in replacements for their corresponding tools. I use them on my
 machine and on machines I run on a daily basis.

 b) Compilation using microarchitecture-targeted optimization, and
 distribution of microarchitecture-specific binary packages. This never got
 off the ground, as SprezzOS 2 development faded out.

 c) Use of aggressive compiler options and alternative compilers. I wasn't
 going to get into this until I had benchmarks showing that it would be
 useful, which have not been obtained.

 d) Tracking upstream as tightly as possible, in the hopes of catching
 performance improvements early, and also watching for regressions.

Only (a) really panned out. I haven't done much work on SprezzOS since this
May (though I upgraded most of our X stack last week), having been
distracted by a book I'm writing (http://nick-black.com/mupcovered-2013-09-13.pdf)
and other things. I'm still running it on my personal workstation, but our
APT traffic has died drastically down, and the Project can be considered in
stasis at best. That said, I'd happily add contributors to the GitHub
projects and wiki, but I'm not holding my breath. =]

Good luck! Hack on!

nick black     http://www.sprezzatech.com -- unix and hpc consulting
to make an apple pie from scratch, you need first invent a universe.

Attachment: signature.asc
Description: Digital signature

Reply to: