“In NetBSD’s sweet spot are organizations looking for a slim, lightweight, highly stable, and capable operating system to run the latest server applications on modest or specialized hardware.” Read the article at ServerWatch.
NetBSD, for When Portability and Stability Matter
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
“Stability and scalability are among NetBSD’s other strengths. Being portable to so many systems demands these capabilities. In a recent scalability benchmark, the current NetBSD release performed better than its BSD kin and nearly as well as the new Linux 2.6 kernel, which topped the rankings. ”
I hate it when so-called journalists post unsupported statements. Does anyone know what scalability benchmark this author is talking about and can post a link to the data and methodology of the test?
That would likely be this article – http://bulk.fefe.de/scalability/
(“Benchmarking BSD and Linux”). I’ve seen this referred to from several places.
In it, with scalability is meant: how CPU load, response times etc. scale with increasing numbers of requests, running processes and the like. In that respect, NetBSD seems to do very decent compared to the competition.
I happen to view scalability also as: how does it adapt to really small (embedded) or really big (many CPU-) systems? In that respect, I’d say Linux 2.6 (running on systems with 64 or more CPU’s) clearly beats NetBSD. On the other hand, the list of NetBSD’s supported hardware platforms is really impressive. BTW: multiprocessing support (SMP) is one of the new/improved features in 2.0.
> That would likely be this article
> – http://bulk.fefe.de/scalability/
In that regards it would be nice to have another test when net2 and free5.3 and linuxkernel2.6.9 will be released.
I emailed the author on that regards.
Thank you. Very interesting reading. It would be even more interesting so see more current data: NetBSD 2.0, FreeBSD 5.3, Linux 2.6, and yes even the clear loser OpenBSD at now 3.6. Maybe Open has fixed things, maybe not.
FreeBSD 5.3 should have most of the debugging code removed and most of the missing features as of 5.1 inserted. NetBSD 2.0 now in release candidate phase should be a good contender now as well. There were also a few changes in the linux kernel between 2.6.0-pre and the current 2.6.9-ish (sometimes they come out in the blink of an eye, heh).
Linux now runs on more platforms than NetBSD does, partly at the embedded end but also at the high end (IA-64, S390, better support of high-end PPC). NetBSD tends to win in supporting older hardware, but the argument that it’s more portable than Linux doesn’t fly any more.
As has been pointed out many times before, this particular scalability test refers to algorithmic as opposed to parallel scalability.
Algorithmic scalability is the easy part. It usually involves code that is localized to one particular subsystem, and the changes made does not have system-wide ramifications. It’s just doing the same old thing more efficiently.
For instance, when the above mentioned test was first published, NetBSD didn’t do too well. They rectified that in a matter of weeks.
Parallel scalability is hard and requires a long-term system-wide effort, and some of the time it’s hard to know if you’re going in the right direction.
When the FreeBSD project branched 5.x over four (?) years ago, they started working on improving SMP (the SMPng project) and multithreading (the KSE project), and those two projects are the main causes for why a production ready 5.x has been so long in coming.
Algorithmic scalability is relatively easy to do right (esp. when you can look at open source systems that has already done the right thing), and is only really interesting when it’s done poorly.
Sorry I am curios. So can you please provide some e.g. url to prove your claim?
I cant see a list of official supported ports on linux.org like i can see on netbsd.org
Regards
Andreas
This may look like nitpicking, but being less ported doesn’t mean being less portable. NetBSD may well still be more portable than any other OS, while it is less ported than Linux, due to much less developpers to do the job, less people interested in a specific platform.
It’s probably easiest to go through the NetBSD list and pick the ones that aren’t supported under Linux. Last time I checked, this was bebox, luna68k, news, next68k, pc532, sun2 and x68k. None of these machines have been manufactured since 1997, and most for some time before that.
In addition to the platforms that both run on, Linux also runs on IA64, Coldfire/Dragonball, cris, m32r, v850, s390, power 4 and various other embedded platforms. And that’s just CPUs – many of these would be counted as more than one platform in the NetBSD world (NetBSD defines platforms that need different kernels as ports – Debian, for instance, defines a port as something that needs an entirely different userspace)
If you only look at processor families, Linux runs on every CPU that NetBSD runs on (well, except for the 68010). It also runs on rather a few others.
I would be willing to bet most of you linux fanboys haven’t even run netbsd for more than a day or so on your home computer. why are you concerned with scalability. NetBSD is teh 0wn.
adam.
that is… on anything BUT your home computer
If you only look at processor families, Linux runs on every CPU that NetBSD runs on (well, except for the 68010). It also runs on rather a few others.
Just a nit, Linux doesn’t run on ns32532 either.
And the Linux definition of “runs” is different than the NetBSD one, at least as far as counting platforms is concerned. In many cases, on those “other” CPUs that Linux “runs” on, it barely runs (e.g. VAX, arm26). On NetBSD those are stable, maintained & mainstream platforms.
On the other hand, NetBSD will not run on anything without an MMU, which Linux has figured out.
>> that is… on anything BUT your home computer
why not on your home (desktop) computer? NetBSD makes a great desktop system and I speak from experience. The only thing it might lack is the ability to run 3d accelerated games (but that is more a driver problem. Take a look here: http://netbsd.org/gallery/in-Action/
I would love to use NetBSD on my laptop, however I can’t. It just won’t recognize my generic pcmcia nic. NetBSD needs a generic driver like Linux’s pcnet32 or whatever its called.
what netbsd provide for desktop/server that linux can’t?
what linux provide for desktop/server that windows can’t?
I agree that NetBSD is good for the home computer. But people arguing here about scalability arent using it on anything that would matter if it could “scale”.
—-
what netbsd provide for desktop/server that linux can’t?
—-
A stable secure system.
> what netbsd provide for desktop/server that linux can’t?
Easy to deploy, easy to update and easy to use (adminning NetBSD is a lot easier than Linux i think). Better manpages. Clean and consistent implementation.
If you use any of the architectures you always have the same system, that is something Linux can’t give you. Example, get yourself a Sparcstation, an AMD64, a PS2 and a Mac. You always have the same version for sed, the same stuff in /usr/share and even the same Kernelversion.
Try the same with Linux and you will see what happens, you end up with four Kernelversions, four versions of sed and so on.
Many people claim that Linux supports more architektures than NetBSD. I don’t think so. What you get is often enough just a Kernel patch and thats it. Usually that patch is for 2.3.9 or 2.4.15 or something like that. Well, i don’t call that supported. Do you?
Once NetBSD is installed it is the same system, no matter what hardware you use. If you find a bug you file a problem-report and the fix will benefit all architectures, thats the good thing when everything is compiled from one sourcetree.
The thirdparty apps are of course the same in NetBSD and Linux, KDE is KDE is KDE.
Important for me, countmail. To enjoy countmail in it’s full glory you have to install NetBSD yourself, of course
.
Karsten
—
THREE HUNDRED SIXTY!
THREE HUNDRED SIXTY NEW MAIL MESSAGES!
HAHAHAHAHA!
tecneeq: I couldn’t agree more!