Re: Not goading a flamewar...
[Author here]
> who nailed the SMP problem?
The problem is that it's hard to define and measure.
Nowadays, according to benchmarking exercises, for large systems with a lot of CPUs or cores, Dragonfly is faster.
Caveat #1: OTOH, it is still a work in progress, and a lot of stuff isn't finished yet. Hammer2 is not yet usable across a cluster, for instance. It's possible that if it were, it would big a big help in performance.
So, from what I've read, Dragonfly beats FreeBSD _on very large systems_. NetBSD is allegedly far behind but this is not its main goal or a primary area of research.
Caveat #2: However, there's more R&D work in Linux and Linux does better at this stuff.
So Linux > Dragonfly > FreeBSD > NetBSD.
Caveat #3: most people don't have lots and lots of CPU cores and this stuff doesn't really apply to the desktop, where single-threaded performance is most important.
So Linux would probably win.
Caveat #4: ... along with filesystem performance and things, where ZFS wins. ZFS on Linux has unavoidable drawbacks, due to licensing issues. (E.g. the Linux cache and ZFS cache are separate.) Advantage, FreeBSD.
Caveat #5: Back in the day, Solaris scaled way better than Linux. I suspect on very large systems, it still might. If someone made enough money to throw at OpenSolaris, I reckon it could still win on large systems.
So... it's complicated. Depends which part of the market you're interested in, and there are no very clear winners and losers.
Heretical view: me, personally, I'd like to see the various BSDs settle their differences and recombine.
Biting the hand that feeds IT
