DragonFlyBSD 3.2.1 vs. Ubuntu Linux Performance
Phoronix: DragonFlyBSD 3.2.1 vs. Ubuntu Linux Performance
At the beginning of this month there was the release of DragonFlyBSD 3.2.1 that claimed a battle for speed against Linux with major improvements for the multi-threaded application performance against Linux. PostgreSQL was the only benchmark cited by the DragonFly camp with the new performance results, so a couple Phoronix tests were carried out.
Very impressive DragonFlyBSD...
Why is it that when Phoronix tests a server-oriented OS, they use benchmarks that are more or less desktop/workstation oriented micro-benchmarks that test a single algorithm?
I mean it's nice that someone actually bothers to do benchmarks but why not do them properly?
It would be more meaningful to test all the major RDBMS systems, and things such as concurrent read, concurrent write, filesystem peformance, blogbench etc.
It's also worth noting that DragonFly didn't have openmp untill the current development version 3.3.
The Himeno Benchmark which is the only one where DragonFly is lacking significantly is interesting though. It's almost as if wrong compiler settings were used or there is simply some feature that DF doesn't have.
Last edited by joe_gunner; 11-18-2012 at 02:42 AM.
It would have been nice if you would have also put FreeBSD 9.1-PRERELEASE into your benchmark parcours, since in the past, it came up to light that DragonFly BSD is now capable of keeping pace with FreeBSD and ScientificLinux. I would have been curious about the difference.
I read your report about the tests, and feel that one cannot draw conclusions
In a statistical test, one would have a few machines with similar configurations and similar disk configurations. One would run the tests you indicated on the 10 or more systems and average out the measurements and then indicate the standard deviation from the average.
And only doing the comparisons on the most souped up machine does not respond well to we who may have dual core computers with 4 gigs of memory, no raid, and 7200rpm disk(s).
What could we expect as results with a quad core AMD or Intel system on a Best Buy machine? Lets keep the machine prices to $500 or less.
From the graphs, I believe that the difference between the Linux and BSD system is less than one percent. The two systems are matched. The decider would be a) support community and b) Quantity of application software.
I know the work you put into setting up and performing the tests and documenting the comparisons, and I believe that you ran your tests multiple times, in order to produce credible results. You followed a good procedure. Please do not take my comments as criticism, I live with quadcore AMD and dualcore INTEL systems, and could not determine how to extrapolate your results to my own two systems.
Leslie in Montreal
they use benchmarks that are more or less desktop
Originally Posted by boucheath
It's not even desktop, many of them are algorithms that don't get used anywhere bar in some scientific university circles.